Re: array support patch phase 1 patch

This patch is starting to get large enough that it is a challenge to
keep in sync with cvs, and it is reasonably complete as a package, so I
was hoping it could be reviewed and committed as "array support phase
1". The most notable missing item is documentation, but if possible I'd
like to defer that until a few more items are complete (see
"Yet-to-come" below).

If a more complete package (e.g. want to wait for items 1-5 of
"Yet-to-come") is preferred, please let me know and I'll keep plugging
along.

Note that the gram.y changes are a bit ugly, but I struggled with
getting anything less ugly to work.

As is, it passes all regression tests (make installcheck), and covers
the following:
----------------------------------------------------------------------
1. Support for polymorphic functions, accepting and returning ANYARRAY
and ANYELEMENT datatypes that are "tied" to each other and resolved to
an actual type at runtime. This also includes the ability to define
aggregates using the polymorphic functions.

2. Array handling functions:
- singleton_array(ANYELEMENT) returns ANYARRAY
- array_push(ANYARRAY, ANYELEMENT) returns ANYARRAY
- array_accum(ANYARRAY, ANYELEMENT) returns ANYARRAY
- array_assign(ANYARRAY, int, ANYELEMENT) returns ANYARRAY
- array_subscript(ANYARRAY, int) returns ANYELEMENT

3. Grammar and underlying support for the following (examples):

create table foo(f1 integer ARRAY);
create table foo(f1 integer ARRAY[]);
create table foo(f1 integer ARRAY[x]);
create table foo(f1 integer ARRAY[][]); (and more [] etc)
create table foo(f1 integer ARRAY[x][y]); (and more [] etc)

select ARRAY[1,2,3];
select ARRAY[[1,2,3],[4,5,6]];
select ARRAY[ARRAY[1,2,3],ARRAY[4,5,6]];
etc up to 6 dimensions

select ARRAY(select oid from pg_class order by relname);

Yet-to-come:
---------------------------------
1. Functions:
- str_to_array(str TEXT, delim TEXT) returns TEXT[]
- array_to_str(array ANYARRAY, delim TEXT) returns TEXT

2. Grammar:
select ARRAY[1,2] || 3;
select ARRAY[1,2] || ARRAY[3,4];
select ARRAY[[1,2],[3,4]] || 5;
select ARRAY[[1,2],[3,4]] || [5,6]

3. Documentation update:
Update "User's Guide"->"Data Types"->"Arrays" documentation
create a new section: "User's Guide"->
"Functions and Operators"->
"Array Functions and Operators"

4. PL/pgSQL support for polymorphic types

5. SQL function support for polymorphic types.

6. Move as much of contrib/array into backend as makes sense (I haven't
looked too close yet), including migration to use polymorphic
semantics.

7. Move as much of contrib/intarray into backend as makes sense (I
haven't looked too close yet), including migration to use
polymorphic semantics (therefore make work on other than int
where possible).

8. Additional documentation and regression test updates

Thanks,

Joe

> This patch is starting to get large enough that it is a challenge to
> keep in sync with cvs, and it is reasonably complete as a package, so I
> was hoping it could be reviewed and committed as "array support phase
> 1". The most notable missing item is documentation, but if possible I'd
> like to defer that until a few more items are complete (see
> "Yet-to-come" below).
>
> If a more complete package (e.g. want to wait for items 1-5 of
> "Yet-to-come") is preferred, please let me know and I'll keep plugging
> along.
>
> Note that the gram.y changes are a bit ugly, but I struggled with
> getting anything less ugly to work.
>
> As is, it passes all regression tests (make installcheck), and covers
> the following:

I can confirm that this patch applies cleanly, compiles and passes all
regression tests on FreeBSD/alpha.

Chris

Joe Conway <mail(at)joeconway(dot)com> writes:
> This patch is starting to get large enough that it is a challenge to
> keep in sync with cvs, and it is reasonably complete as a package, so I
> was hoping it could be reviewed and committed as "array support phase
> 1".

I looked over this patch a little bit, mostly at the anyarray/anyelement
changes; I didn't study the ArrayExpr stuff (except to notice that yeah,
the grammar change is very ugly; can't it be made independent of the
number of levels?).

I like the anyarray/anyelement stuff with the exception of the
actualRetType field added to FunctionCallInfoData. In the first place,
that's the wrong place for such data; FunctionCallInfoData should only
contain data that will change for each call, which resolved type won't.
This should go into FmgrInfo, perhaps, or maybe we could treat it as a
call-context item (though I think we might have conflicts with existing
uses of the context link). A bigger gripe is that passing only
actualRetType isn't future-proof. It doesn't help functions that take
anyelement but return a fixed type; they won't know what the argument type
is. And it won't scale if we end up adding anyarray2/anyelement2/etc to
allow multiple generic parameter types; at most one of the actual
parameter types could be made visible to the function. I see your concern
here, but I think the only adequate solution is to make sure the function
can learn the types of all its arguments, as well as the assigned return
type. (Of course it could deduce the latter from the former, but we may
as well save it the effort, especially since it would have to repeat
parse-time catalog lookups in many interesting cases.)

A notion that I've toyed with more than once is to allow a function
to get at the parse tree representation for its call (ie, the FuncExpr
or OpExpr node). If we did that, a function could learn all these
types by examining the parse tree, and it could learn other things
besides. A disadvantage is that functions would have to be ready to
cope with all the wooliness of parse trees. It'd probably be bright
to make some convenience functions "get my return type", "get type
of my n'th argument" to localize this logic as much as possible.

In short then, what about passing an Expr* link in FmgrInfo as a
substitute for actualRetType? (An additional advantage of this way
is that it's obvious whether or not the information has been provided,
which it would not be for, say, functions invoked via DirectFunctionCallN.
The patch as it stands fails silently if a polymorphic function is called
from a call site that doesn't provide the needed info.)

regards, tom lane

Tom Lane wrote:
> I looked over this patch a little bit, mostly at the anyarray/anyelement
> changes; I didn't study the ArrayExpr stuff (except to notice that yeah,
> the grammar change is very ugly; can't it be made independent of the
> number of levels?).

I struggled with it for quite a while, but then again, I can't claim
yacc grammar as a particularly strong point. I kept running into
reduce/reduce errors with anything completely recursive, or when it did
work, I found myself restricted to one level of nesting. I'll try again
-- any suggestions for something similar to study?

> parameter types could be made visible to the function. I see your concern
> here, but I think the only adequate solution is to make sure the function
> can learn the types of all its arguments, as well as the assigned return
> type. (Of course it could deduce the latter from the former, but we may
> as well save it the effort, especially since it would have to repeat
> parse-time catalog lookups in many interesting cases.)

I came to the same conclusion with ArrayExpr/ArrayExprState -- you can
always deduce the array type from the element type and vice-versa, but
it seemed costly to do in the executor when it can be done just once in
the parser.

> A notion that I've toyed with more than once is to allow a function
> to get at the parse tree representation for its call (ie, the FuncExpr
> or OpExpr node). If we did that, a function could learn all these
> types by examining the parse tree, and it could learn other things
> besides. A disadvantage is that functions would have to be ready to
> cope with all the wooliness of parse trees. It'd probably be bright
> to make some convenience functions "get my return type", "get type
> of my n'th argument" to localize this logic as much as possible.

OK -- I'll take a look at that option.

> In short then, what about passing an Expr* link in FmgrInfo as a
> substitute for actualRetType? (An additional advantage of this way
> is that it's obvious whether or not the information has been provided,
> which it would not be for, say, functions invoked via DirectFunctionCallN.
> The patch as it stands fails silently if a polymorphic function is called
> from a call site that doesn't provide the needed info.)

Sounds good to me -- thanks for the quick feedback!

A question -- I was thinking that we will need to allow one array type
to be coerced into another in this brave new world (e.g.
ARRAY[1,2,3]::float8[] results in "ERROR: Cannot cast type integer[] to
double precision[]"). My plan was to check the source and target
datatypes in can_coerce_type() and coerce_type() to see if both are
array types. If so, use the element types to determine whether we can
coerce, and loop over the array to coerce, etc.

There are a few issues that I can think of with this (and undoubtedly
you will have some others ;-)):
- this would add at least one cache lookup to every call to
can_coerce_type
- if we pass by can_coerce_type with an array, we'd need to do the same
cache lookup in coerce_type unless we modify the api of the related
functions to pass along the "isarray" status of the datatype
- currently the only reliable(?) method to determine if a
datatype is a varlena array is to check for '_' as the first
character of the type name -- it seems wrong to depend on that
forever.

Any suggestions? I was toying with the idea that an isarray attribute
should be added to pg_type -- thoughts on that?

Joe

Joe Conway <mail(at)joeconway(dot)com> writes:
> Any suggestions? I was toying with the idea that an isarray attribute
> should be added to pg_type -- thoughts on that?

Don't think you need it to go in that direction: the property of having
nonzero typelem indicates that something is an array.

The real issue is whether we should add a link to let you find the array
type given the base type (without relying on the '_' naming convention).
I'm undecided about that...

regards, tom lane

Tom Lane wrote:
> Joe Conway <mail(at)joeconway(dot)com> writes:
>>Any suggestions? I was toying with the idea that an isarray attribute
>>should be added to pg_type -- thoughts on that?
>
> Don't think you need it to go in that direction: the property of having
> nonzero typelem indicates that something is an array.

But I don't think of the following as arrays, at least not in the same
sense of _text, etc.:

regression=# select typname from pg_type where typelem != 0 and typname
not like '\\_%';
typname
------------
name
int2vector
oidvector
point
lseg
box
line
(7 rows)

Joe

Joe Conway <mail(at)joeconway(dot)com> writes:
> Tom Lane wrote:
>> Don't think you need it to go in that direction: the property of having
>> nonzero typelem indicates that something is an array.

> But I don't think of the following as arrays, at least not in the same
> sense of _text, etc.:

If you want to reject fixed-length array types for this purpose, you can
also insist on typlen < 0.

regards, tom lane

Tom Lane wrote:
> I looked over this patch a little bit, mostly at the anyarray/anyelement
> changes; I didn't study the ArrayExpr stuff (except to notice that yeah,
> the grammar change is very ugly; can't it be made independent of the
> number of levels?).
[...snip...]
> In short then, what about passing an Expr* link in FmgrInfo as a
> substitute for actualRetType? (An additional advantage of this way
> is that it's obvious whether or not the information has been provided,
> which it would not be for, say, functions invoked via DirectFunctionCallN.
> The patch as it stands fails silently if a polymorphic function is called
> from a call site that doesn't provide the needed info.)

Here's a new patch that fixes the grammar ugliness (amazing how easily I
got that to work after a week of not looking at it!) and adds an
(Expr *) link in FmgrInfo. It passes all regression tests. If there are
no objections, please apply.

(as phase 1 -- see:
http://archives.postgresql.org/pgsql-patches/2003-03/msg00191.php
for what the patch covers and the planned "yet-to-come").

Thanks,

Joe

Joe Conway wrote:
> Here's a new patch that fixes the grammar ugliness (amazing how easily I
> got that to work after a week of not looking at it!) and adds an
> (Expr *) link in FmgrInfo. It passes all regression tests. If there are
> no objections, please apply.
>

I just found a huge bogosity in this patch (ExecEvalArray)-- I must have
been tired, and my test cases were obviously too simple ;(

I'm working on a fix right now, but I guess I ought to add some good
test cases to the regression suite before submitting again. Please
retract this patch for the moment, and I'll send in the fix with
regression test adjustments in the next few days.

Sorry if I wasted anyone's time!

Joe

Joe Conway wrote:
> I just found a huge bogosity in this patch (ExecEvalArray)-- I must have
> been tired, and my test cases were obviously too simple ;(
>
> I'm working on a fix right now, but I guess I ought to add some good
> test cases to the regression suite before submitting again. Please
> retract this patch for the moment, and I'll send in the fix with
> regression test adjustments in the next few days.

Here's an updated patch that fixes the above mentioned issues. I also
added significantly more functionality.

It needs a good review, but everything seems to work, it passes all
regression tests**, and I think it is ready to apply (I needed to change
some of the regression tests, so someone please take a close look at
those too).

**except horology due to DST, and for some reason that I don't yet
understand, /expected/misc.out loses my changes every time I update from
CVS -- is misc.out autogenerated?

It covers the following:
----------------------------------------------------------------------
1. Support for polymorphic functions, accepting and returning ANYARRAY
and ANYELEMENT datatypes that are "tied" to each other and resolved
to an actual type at runtime. This also includes the ability to
define aggregates using the polymorphic functions.

2. Array handling functions:
- singleton_array(ANYELEMENT) returns ANYARRAY
- array_append(ANYARRAY, ANYELEMENT) returns ANYARRAY
- array_prepend(ANYELEMENT, ANYARRAY) returns ANYARRAY
- array_accum(ANYARRAY, ANYELEMENT) returns ANYARRAY
- array_assign(ANYARRAY, int, ANYELEMENT) returns ANYARRAY
- array_subscript(ANYARRAY, int) returns ANYELEMENT
- array_cat(ANYARRAY, ANYARRAY) returns ANYARRAY

3. Grammar and underlying support for the following (examples):

create table foo(f1 integer ARRAY);
create table foo(f1 integer ARRAY[]);
create table foo(f1 integer ARRAY[x]);
create table foo(f1 integer ARRAY[][]); (and more [] etc)
create table foo(f1 integer ARRAY[x][y]); (and more [] etc)

select ARRAY[1,2,3];
select ARRAY[[1,2,3],[4,5,6]];
select ARRAY[ARRAY[1,2,3],ARRAY[4,5,6]];
...etc up to 6 dimensions

select ARRAY[1,2] || 3;
select 0 || ARRAY[1,2];
select ARRAY[1,2] || ARRAY[3,4];
select ARRAY[1,2] || ARRAY[[3,4],[5,6]];
select ARRAY[[1,2],[3,4]] || ARRAY[5,6];
select ARRAY(select oid from pg_class order by relname);

select ARRAY[[1,2,3],[4,5,6]]::float8[];
select ARRAY[[1,2,3],[4,5,6]]::text[];
select CAST(ARRAY[[1,2,3],[4,5,6]] AS float8[]);

4. Duplicated contrib/array functionality (and then some) in the
backend using polymorphic functions and operators. Also cleared
out the datatype specific "=" operators. Now there is just one
polymorphic one. If the patch is applied, I think contrib/array can
be safely retired. Examples:

SELECT ARRAY[1,2,3] *= 2 AS "TRUE";
SELECT ARRAY[1,2,3] *<> 4 AS "TRUE";
SELECT ARRAY[1,2,3] *< 2 AS "TRUE";
SELECT ARRAY[1,2,3] *> 2 AS "TRUE";
SELECT ARRAY[1,2,3] *<= 1 AS "TRUE";
SELECT ARRAY[1,2,3] *>= 3 AS "TRUE";
SELECT ARRAY['abc','def','qwerty'] *~ 'wer' AS "TRUE";
SELECT ARRAY['abc','def','qwerty'] *!~ 'xyz' AS "TRUE";
SELECT ARRAY['abc','def','qwerty'] *~~ '_wer%' AS "TRUE";
SELECT 'hello' *~~ ARRAY['abc','def','_ell%'] AS "TRUE";
SELECT ARRAY['abc','def','qwerty'] *!~~ '%xyz%' AS "TRUE";

5. Side note: I added ANYARRAY1 and ANYELEMENT1 in this version. I
needed ANYARRAY1 for polymorphic array coercion, but did not add
support for it to be used in any other way. I added ANYELEMENT1
only for symmetry, and did not use it at all.

Still needs to be done (in roughly this order):
----------------------------------------------------------------------
1. Functions:
- str_to_array(str TEXT, delim TEXT) returns TEXT[]
- array_to_str(array ANYARRAY, delim TEXT) returns TEXT

2. Documentation update:
Update "User's Guide"->"Data Types"->"Arrays" documentation
create a new section: "User's Guide"->
"Functions and Operators"->
"Array Functions and Operators"

4. PL/pgSQL support for polymorphic types

5. SQL function support for polymorphic types.

6. Move as much of contrib/intarray into backend as makes sense,
including migration to use polymorphic semantics (therefore make
work on other than int where possible). Note: this may not happen
for 7.4 as it looks to be fairly involved, at least at first glance.

7. Additional documentation and regression test updates

----------------------------------------------------------------------

Hopefully I haven't forgotten anything. If so, be gentle ;-)

Regards,

Joe

Joe Conway <mail(at)joeconway(dot)com> writes:
> **except horology due to DST, and for some reason that I don't yet
> understand, /expected/misc.out loses my changes every time I update from
> CVS -- is misc.out autogenerated?

Yeah. You need to back-patch src/test/regress/output/misc.source ...

regards, tom lane

Joe Conway writes:

> select ARRAY[1,2,3];
> select ARRAY[[1,2,3],[4,5,6]];
> select ARRAY[ARRAY[1,2,3],ARRAY[4,5,6]];
> ...etc up to 6 dimensions

Why only 6?

> 4. Duplicated contrib/array functionality (and then some) in the
> backend using polymorphic functions and operators.

> SELECT ARRAY[1,2,3] *= 2 AS "TRUE";
> SELECT ARRAY[1,2,3] *<> 4 AS "TRUE";

Couldn't this kind of operation be handled more cleanly (at least
semantically speaking), if we provide a function that converts an array to
a set and then use standard set searching operations? For example,

SELECT 2 IN TABLE(ARRAY[1,2,3]);

> 5. Side note: I added ANYARRAY1 and ANYELEMENT1 in this version.

Doing what?

--
Peter Eisentraut peter_e(at)gmx(dot)net

Joe Conway kirjutas E, 07.04.2003 kell 05:12:
> Joe Conway wrote:
> It covers the following:
> ----------------------------------------------------------------------
> 1. Support for polymorphic functions, accepting and returning ANYARRAY
> and ANYELEMENT datatypes that are "tied" to each other and resolved
> to an actual type at runtime. This also includes the ability to
> define aggregates using the polymorphic functions.
>
> 2. Array handling functions:
> - singleton_array(ANYELEMENT) returns ANYARRAY
> - array_append(ANYARRAY, ANYELEMENT) returns ANYARRAY
> - array_prepend(ANYELEMENT, ANYARRAY) returns ANYARRAY
> - array_accum(ANYARRAY, ANYELEMENT) returns ANYARRAY
> - array_assign(ANYARRAY, int, ANYELEMENT) returns ANYARRAY
> - array_subscript(ANYARRAY, int) returns ANYELEMENT
> - array_cat(ANYARRAY, ANYARRAY) returns ANYARRAY
>

How hard would it be to add

array_eq, array_ne, array_gt, array_le and corresponding operators

SELECT ARRAY[1,2,3] = ARRAY[1,2,3]; # --> TRUE
SELECT ARRAY[1,2,3] < ARRAY[1,2,3]; # --> FALSE
SELECT ARRAY[1,2,3] <= ARRAY[1,2,3]; # --> TRUE
SELECT ARRAY[1,2,3] > ARRAY[1,2,3]; # --> FALSE
SELECT ARRAY[1,2,3] >= ARRAY[1,2,3]; # --> TRUE

I'd assume them to behave like string comparisons, i.e shorter subarray
is smaller:

SELECT ARRAY[1,2] < ARRAY[1,2,3]; # --> FALSE

Support for sorting and b-tree indexing could be nice too.

> Still needs to be done (in roughly this order):
> ----------------------------------------------------------------------
> 1. Functions:
> - str_to_array(str TEXT, delim TEXT) returns TEXT[]
> - array_to_str(array ANYARRAY, delim TEXT) returns TEXT
>
> 2. Documentation update:
> Update "User's Guide"->"Data Types"->"Arrays" documentation
> create a new section: "User's Guide"->
> "Functions and Operators"->
> "Array Functions and Operators"
>
>
> 4. PL/pgSQL support for polymorphic types

Where should one start to add PL/Python support for polymorphic types ?

> 6. Move as much of contrib/intarray into backend as makes sense,
> including migration to use polymorphic semantics (therefore make
> work on other than int where possible). Note: this may not happen
> for 7.4 as it looks to be fairly involved, at least at first glance.

What about moving contrib/intagg into backend ?

(And converting it into ANYagg on the way ;)

--------------------
Hannu

Peter Eisentraut <peter_e(at)gmx(dot)net> writes:
> Joe Conway writes:
>> ...etc up to 6 dimensions

> Why only 6?

See uses of MAXDIM. If you feel that 6 isn't enough, I wouldn't have
a problem with raising MAXDIM to 10 or so. I don't think it's worth
trying to eliminate the limit completely; that would add palloc overhead
to every array operation, for a feature people are unlikely to use.

>> 4. Duplicated contrib/array functionality (and then some) in the
>> backend using polymorphic functions and operators.

> Couldn't this kind of operation be handled more cleanly (at least
> semantically speaking), if we provide a function that converts an array to
> a set and then use standard set searching operations? For example,
> SELECT 2 IN TABLE(ARRAY[1,2,3]);

Not sure about that. Is there any guidance in the SQL 200x spec about
what they expect people to actually *do* with the ARRAY[] syntax?

I'm currently working over the patch, but am not going to commit this
part (yet), since I have some problems with the implementation anyway.

>> 5. Side note: I added ANYARRAY1 and ANYELEMENT1 in this version.

> Doing what?

I'm hoping to avoid committing those, as they don't seem to be
completely implemented in this patch. They may be required to
represent the behavior of array_coerce() though. Not sure yet.

regards, tom lane

Peter Eisentraut wrote:
> Joe Conway writes:
>> ...etc up to 6 dimensions
>
> Why only 6?

As Tom mentioned MAXDIM is currently set to 6. I can't imagine many
people using anything over 3 or maybe 4 dimensions.

>>4. Duplicated contrib/array functionality (and then some) in the
>> backend using polymorphic functions and operators.
>
>> SELECT ARRAY[1,2,3] *= 2 AS "TRUE";
>> SELECT ARRAY[1,2,3] *<> 4 AS "TRUE";
>
> Couldn't this kind of operation be handled more cleanly (at least
> semantically speaking), if we provide a function that converts an array to
> a set and then use standard set searching operations? For example,
>
> SELECT 2 IN TABLE(ARRAY[1,2,3]);

I thought about that too. It wouldn't be general enough to handle other
operators though, so I decided to stick with the already somewhat
established contrib/array operators. It sounds like Tom has some
concerns with those anyway, so in the meantime I'll take another look at
SQL200x to see if this is covered somewhere.

>>5. Side note: I added ANYARRAY1 and ANYELEMENT1 in this version.
>
> Doing what?

As I said:
" I needed ANYARRAY1 for polymorphic array coercion, but did not add
support for it to be used in any other way. I added ANYELEMENT1
only for symmetry, and did not use it at all."

The problem lies in that fact that more than one reference to ANYARRAY
in the argument list of a function implies that the said arrays are all
the same data type. This doesn't work for a coercion function where you
need two array arguments, of arbitrary types, but that are not the same
as each other. I could not see any other clean way to achieve this.

I specifically did not add any other support for these data types
because there is not yet a strong case that user defined functions need
this capability. Of course if there is a strong feeling that they should
have full fledged support similar to ANYARRAY and ANYELEMENT, I'll
gladly add it.

Joe

Joe Conway <mail(at)joeconway(dot)com> writes:
>>> 5. Side note: I added ANYARRAY1 and ANYELEMENT1 in this version.
>>
>> Doing what?

> The problem lies in that fact that more than one reference to ANYARRAY
> in the argument list of a function implies that the said arrays are all
> the same data type. This doesn't work for a coercion function where you
> need two array arguments, of arbitrary types, but that are not the same
> as each other. I could not see any other clean way to achieve this.

What I'm currently thinking of is declaring array_coerce to take and
return ANYARRAY (which will make it a no-op if someone tries to invoke
it by hand). The coercion code will have to force the output type to
the desired thing after building the FuncExpr node. But the other way
needs special-case code in parse_coerce too, and it doesn't do something
reasonable if array_coerce is invoked by hand.

regards, tom lane

Hannu Krosing wrote:
> How hard would it be to add
>
> array_eq, array_ne, array_gt, array_le and corresponding operators
>
> SELECT ARRAY[1,2,3] = ARRAY[1,2,3]; # --> TRUE
> SELECT ARRAY[1,2,3] < ARRAY[1,2,3]; # --> FALSE
> SELECT ARRAY[1,2,3] <= ARRAY[1,2,3]; # --> TRUE
> SELECT ARRAY[1,2,3] > ARRAY[1,2,3]; # --> FALSE
> SELECT ARRAY[1,2,3] >= ARRAY[1,2,3]; # --> TRUE
>
> I'd assume them to behave like string comparisons, i.e shorter subarray
> is smaller:
>
> SELECT ARRAY[1,2] < ARRAY[1,2,3]; # --> FALSE
>
> Support for sorting and b-tree indexing could be nice too.

I thought briefly about this, but it wasn't immediately clear what the
semantics ought to be in all cases. I've also spent literally all my
available "hacking" time for the last several weeks just to get to the
patch submitted. I'd like to see at least some of it committed before I
take on anything new ;-)

If you want to propose in detail how these would behave - including:
- different length arrays
- different dimension arrays
- is comparison by ordinal position, or is each element compared to
all elements of the other side, e.g. is
(ARRAY[1,2,3] < ARRAY[2,3,4]) TRUE or FALSE?
If you compare by ordinal position TRUE, but in the latter case
FALSE

> Where should one start to add PL/Python support for polymorphic types ?

Not entirely sure. In plperl, pltcl, and plr there is a section of code
in the compile function that looks something like:

/* Disallow pseudotype result, except VOID */
if (typeStruct->typtype == 'p')
{
if (procStruct->prorettype == VOIDOID)
/* okay */ ;
else if (procStruct->prorettype == TRIGGEROID)
{
free(prodesc->proname);
free(prodesc);
elog(ERROR, "plperl functions cannot return type %s"
"\n\texcept when used as triggers",
format_type_be(procStruct->prorettype));

And something similar for arguments. But I don't at quick glance see
something like this in plpython.

This needs to be changed to allow types ANYARRAYOID and ANYELEMENTOID.
Then you need to do the right thing with the arguments/return type. For
example, from plr:

#define GET_PROARGS(pronargs_, proargtypes_) \
do { \
int i; \
pronargs_ = procStruct->pronargs; \
for (i = 0; i < pronargs_; i++) \
{ \
if (procStruct->proargtypes[i] == ANYARRAYOID || \
procStruct->proargtypes[i] == ANYELEMENTOID) \
{ \
proargtypes_[i] = get_expr_argtype(fcinfo, i); \
if (proargtypes_[i] == InvalidOid) \
proargtypes_[i] = procStruct->proargtypes[i]; \
} \
else \
proargtypes_[i] = procStruct->proargtypes[i]; \
} \
} while (0)

This grabs the parser determined runtime datatype from the expression
node that has been added to FmgrInfo.

Last thing I can think of is an important safety tip. If plpython caches
the compiled function, you need to take care to invalidate it if the
return or argument types change from call to call.

> What about moving contrib/intagg into backend ?
> (And converting it into ANYagg on the way ;)
>

As I said earlier, all in good time ;-) One question about that though
-- how is intagg different from array_accum? (which is already
polymorphic and in the submitted patch)

Joe

Joe Conway kirjutas T, 08.04.2003 kell 19:55:
> Hannu Krosing wrote:
> > SELECT ARRAY[1,2] < ARRAY[1,2,3]; # --> FALSE
> >
> > Support for sorting and b-tree indexing could be nice too.
>
> I thought briefly about this, but it wasn't immediately clear what the
> semantics ought to be in all cases. I've also spent literally all my
> available "hacking" time for the last several weeks just to get to the
> patch submitted. I'd like to see at least some of it committed before I
> take on anything new ;-)
>
> If you want to propose in detail how these would behave - including:

I like the "compare like strings" approach

> - different length arrays

shorter is smaller if all elements are same

> - different dimension arrays

not comparable, just as array elemant is not comparable with array

> - is comparison by ordinal position, or is each element compared to
> all elements of the other side,

neither, pairs of elements at same positions are compared until first !=
or one array ends, then the shorter array the array with smaller element
value is considered smaller.

> e.g. is
> (ARRAY[1,2,3] < ARRAY[2,3,4]) TRUE or FALSE?
> If you compare by ordinal position TRUE, but in the latter case
> FALSE

all the following should also be TRUE

(ARRAY[1,2,3] < ARRAY[2,1,1])
(ARRAY[1,2,3] < ARRAY[1,2,3,4])
(ARRAY[] < ARRAY[1,2,3,4])
(ARRAY[1,2,3,4] < ARRAY[1,2,4,3])

> > Where should one start to add PL/Python support for polymorphic types ?
>
> Not entirely sure. In plperl, pltcl, and plr there is a section of code
> in the compile function that looks something like:
>
> ....

Thanks, I'll take a look.

> > What about moving contrib/intagg into backend ?
> > (And converting it into ANYagg on the way ;)
> >
>
> As I said earlier, all in good time ;-) One question about that though
> -- how is intagg different from array_accum? (which is already
> polymorphic and in the submitted patch)

Sorry, didn't notice array_accum.

intagg has functions for doing it both ways (array->table and table ->
array).

-----------------
Hannu

Joe Conway <mail(at)joeconway(dot)com> writes:
> Here's an updated patch that fixes the above mentioned issues. I also
> added significantly more functionality.

I've applied much but not all of this. Some comments:

I didn't apply any of the aggregate changes, because I'm unsure of how
we actually want to do that. AFAICT you've set it up so that the
polymorphism of the transition function is hidden within the aggregate,
and one must still create a separate aggregate for each input datatype.
Couldn't we fix things so that a polymorphic transition function leads
to a polymorphic aggregate? I'm not totally sure that makes sense, but
it seems worth thinking about.

Also I didn't put in the bool_op stuff. That seemed pretty messy; in
particular I didn't care for looking at the operator names to decide
what to do. Another problem is that the actual lookup of the scalar
operators would be schema search path dependent. I'd feel more
comfortable with something that created a tighter binding of the array
operators to the underlying scalar operators. Not sure how to do it,
though.

There are a number of remaining annoyances in array type assignment and
coercion. Here's one:

create table t1 (p point[]);
insert into t1 values(array['(3,4)','(4,5)','(6,7)']);
ERROR: column "p" is of type point[] but expression is of type text[]
You will need to rewrite or cast the expression
Instead you have to write
insert into t1 values(array['(3,4)'::point,'(4,5)','(6,7)']);

I'm not sure if there's any way around this, but it certainly reduces the
usefulness of untyped literals. Right offhand it seems like maybe we could
invent an UNKNOWNARRAY type so as to postpone the decision about what the
array's type should be --- but I'm worried about possible side-effects.
I think there are places where the existence of "unknown[]" might allow
the type resolution logic to follow paths we don't want it to follow.

I think there are a lot of places where binary-compatible element types
and domain element types will not be treated reasonably. We need to think
about where and how to handle that.

Another problem is:

regression=# select distinct array(select * from text_tbl) from foo;
ERROR: Unable to identify an equality operator for type text[]

This DISTINCT query would fail anyway of course, for lack of a '<'
operator for arrays, but it seems like we ought to be able to find the
polymorphic '=' operator.

We really ought to reimplement array_eq to be less bogus: instead of bit
equality it ought to be applying the element type's equality operator.

I rearranged the way that parse_coerce and function/operator overloading
resolution handle arrays. It seems cleaner to me, but take a look and
see what you think.

One thing I didn't like about the implementation of many of these functions
is that they were willing to repeat catalog lookups on every call. I fixed
array_type_coerce to cache lookup results, please see if you can apply the
technique elsewhere.

The selectivity estimation functions you'd assigned to the bool_ops operators
seemed like pretty bogus choices. I am not sure we can use scalar estimators
for these at all --- perhaps a new set of estimators needs to be written.

> create table foo(f1 integer ARRAY);
> create table foo(f1 integer ARRAY[]);
> create table foo(f1 integer ARRAY[x]);
> create table foo(f1 integer ARRAY[][]); (and more [] etc)
> create table foo(f1 integer ARRAY[x][y]); (and more [] etc)

AFAICT, SQL99 specifies the syntax "typename ARRAY [ intconst ]" and
nothing else. I do not see a reason to accept ARRAY without [],
ARRAY with empty [], ARRAY with multiple [], etc if the spec doesn't
make us do so. The existing Postgres syntax without the word ARRAY
gets the job done just fine.

regards, tom lane

Tom Lane wrote:
> I've applied much but not all of this. Some comments:

Thanks for the review and cleanup!

> I didn't apply any of the aggregate changes, because I'm unsure of how
> we actually want to do that. AFAICT you've set it up so that the
> polymorphism of the transition function is hidden within the aggregate,
> and one must still create a separate aggregate for each input datatype.
> Couldn't we fix things so that a polymorphic transition function leads
> to a polymorphic aggregate? I'm not totally sure that makes sense, but
> it seems worth thinking about.

I went back and forth on that issue myself more than once while I was
working on this. I'll take another look at making aggregate polymorphic
as well.

> Also I didn't put in the bool_op stuff. That seemed pretty messy; in
> particular I didn't care for looking at the operator names to decide
> what to do. Another problem is that the actual lookup of the scalar
> operators would be schema search path dependent. I'd feel more
> comfortable with something that created a tighter binding of the array
> operators to the underlying scalar operators. Not sure how to do it,
> though.

But the lookup would be schema search path dependent if we were given
two scalars, so I don't see this as any different. Would it be better to
use the same operators as the scalars ("=", "<>", ...etc)? It makes
sense to me that "array = element" should apply the "=" operator for the
element data type, across all of the array elements. Maybe this takes us
back to Peter's suggestion:
expression IN (array)
expression NOT IN (array)
expression operator ANY (array)
expression operator SOME (array)
(expression) operator (array)
(expression) operator ALL (array)

In this case operator is taken as the appropriate operator for the
expression type as both left and right arguments, and array is treated
the same as a one column subquery.

> There are a number of remaining annoyances in array type assignment and
> coercion. Here's one:
>
> create table t1 (p point[]);
> insert into t1 values(array['(3,4)','(4,5)','(6,7)']);
> ERROR: column "p" is of type point[] but expression is of type text[]
> You will need to rewrite or cast the expression
> Instead you have to write
> insert into t1 values(array['(3,4)'::point,'(4,5)','(6,7)']);
>
> I'm not sure if there's any way around this, but it certainly reduces the
> usefulness of untyped literals. Right offhand it seems like maybe we could
> invent an UNKNOWNARRAY type so as to postpone the decision about what the
> array's type should be --- but I'm worried about possible side-effects.
> I think there are places where the existence of "unknown[]" might allow
> the type resolution logic to follow paths we don't want it to follow.
>
> I think there are a lot of places where binary-compatible element types
> and domain element types will not be treated reasonably. We need to think
> about where and how to handle that.

Yeah, I think this is what led me to the hack in parse_coerce that you
didn't like ;-), but I'm not sure how to better handle it. I'll think a
bit more on it.

>
> Another problem is:
>
> regression=# select distinct array(select * from text_tbl) from foo;
> ERROR: Unable to identify an equality operator for type text[]
>
> This DISTINCT query would fail anyway of course, for lack of a '<'
> operator for arrays, but it seems like we ought to be able to find the
> polymorphic '=' operator.
>
> We really ought to reimplement array_eq to be less bogus: instead of bit
> equality it ought to be applying the element type's equality operator.

OK. I'll look at these issues again. Should I also look to implement:
array <> array
array > array
array < array
array >= array
array <= array

as Hannu suggested?

> I rearranged the way that parse_coerce and function/operator overloading
> resolution handle arrays. It seems cleaner to me, but take a look and
> see what you think.

OK.

> One thing I didn't like about the implementation of many of these functions
> is that they were willing to repeat catalog lookups on every call. I fixed
> array_type_coerce to cache lookup results, please see if you can apply the
> technique elsewhere.

OK.

> The selectivity estimation functions you'd assigned to the bool_ops operators
> seemed like pretty bogus choices. I am not sure we can use scalar estimators
> for these at all --- perhaps a new set of estimators needs to be written.

I figured you wouldn't like that, but I was at a loss as to how to
approach something better. I'll take a shot at writing new estimators
based on the scalar ones.

> AFAICT, SQL99 specifies the syntax "typename ARRAY [ intconst ]" and
> nothing else. I do not see a reason to accept ARRAY without [],
> ARRAY with empty [], ARRAY with multiple [], etc if the spec doesn't
> make us do so. The existing Postgres syntax without the word ARRAY
> gets the job done just fine.

I had in my notes this (I think from SQL200x):

6.1 <data type>
<collection type> ::= <array type> | <multiset type>
<array type> ::=
<data type> ARRAY [ <left bracket or trigraph>
<unsigned integer>
<right bracket or trigraph> ]

So I thought
typename ARRAY [ intconst ]
typename ARRAY
were both per spec. As far as "typename ARRAY [ ]", ARRAY with multiple
[], etc goes, I only included them for consistency, but I don't care if
we leave them out either.

As always, a thorough review and a lot to think about! Thanks for the
help. It will be at least a couple days before I can pick this up again
(my development system lost a cpu fan yesterday, so I'm taking this as
an opportunity to upgrade ;-)), but I'll get on it as soon as I can.

Joe

Joe Conway <mail(at)joeconway(dot)com> writes:
> Tom Lane wrote:
>> Also I didn't put in the bool_op stuff. That seemed pretty messy; in
>> particular I didn't care for looking at the operator names to decide
>> what to do. Another problem is that the actual lookup of the scalar
>> operators would be schema search path dependent. I'd feel more
>> comfortable with something that created a tighter binding of the array
>> operators to the underlying scalar operators. Not sure how to do it,
>> though.

> But the lookup would be schema search path dependent if we were given
> two scalars, so I don't see this as any different.

But it is different, because the lookup happens at run time not at parse
time. In particular consider rules or views that might be executed with
current search paths completely unrelated to what was used when they
were defined. You've now got a situation where the rule/view author
does not have control of what code is being executed on his behalf.
Not good.

(Come to think of it, much the same complaint might be laid for PL and
SQL functions; maybe it'd be better to associate a schema search path
with a function when it's created?)

Perhaps we could insist that the relevant scalar operators be found in
the same schema the array operator lives in? Safe but restrictive...

> Would it be better to
> use the same operators as the scalars ("=", "<>", ...etc)? It makes
> sense to me that "array = element" should apply the "=" operator for the
> element data type, across all of the array elements.

It seems a little surprising to me. "IN" and "=" mean quite different
things, and I'd not expect them to be represented by the same operator name.

> Maybe this takes us
> back to Peter's suggestion:
> expression IN (array)
> expression NOT IN (array)
> expression operator ANY (array)
> expression operator SOME (array)
> (expression) operator (array)
> (expression) operator ALL (array)

There's a lot to be said for that, if we can think of a way to do it.
I believe this way would force us to integrate the operations into the
parser, though, rather than just throwing a few polymorphic functions
at the problem. It's probably a lot more work :-(

> OK. I'll look at these issues again. Should I also look to implement:
> array <> array
> array > array
> array < array
> array >= array
> array <= array

> as Hannu suggested?

Not sure about that. A point that should have been made in that
conversation is that SQL already defines "row comparison" operations
that are supposed to act in much the same fashion Hannu suggested.
We don't currently implement those correctly, but we should.

Consider also that the minute we have the above functions, people will
expect to be able to build indexes on array columns and sort by array
columns. The bool_ops code as it stands can't effectively support that,
because these contexts do not have expression parsetrees hanging around
to pass to fn_expr.

regards, tom lane

Tom Lane wrote:
> I've applied much but not all of this. Some comments:

Here (finally!) is a doc patch to go with the previously applied array
support changes. If there are no objections, please apply.

Thanks,

Joe

Joe Conway wrote:
> Tom Lane wrote:
>> I've applied much but not all of this. Some comments:
>
> Here (finally!) is a doc patch to go with the previously applied array
> support changes. If there are no objections, please apply.

Attached is an updated version of the doc patch along with the addition
of two new functions, as previously discussed:

- str_to_array(str TEXT, delim TEXT) returns TEXT[]
- array_to_str(array ANYARRAY, delim TEXT) returns TEXT

The patch is fairly large because:

1. the documentation part of it is large
2. I moved some static functions around and made them accessible so
that I could use them in implementing the two functions.

It passes all regression tests, and as mentioned above includes
documentation.

If there are no objections, please apply.

Thanks,

Joe

Joe Conway wrote:
> Attached is an updated version of the doc patch along with the addition
> of two new functions, as previously discussed:
>
> - str_to_array(str TEXT, delim TEXT) returns TEXT[]
> - array_to_str(array ANYARRAY, delim TEXT) returns TEXT
>
> The patch is fairly large because:
>
> 1. the documentation part of it is large
> 2. I moved some static functions around and made them accessible so
> that I could use them in implementing the two functions.
>
> It passes all regression tests, and as mentioned above includes
> documentation.
>
> If there are no objections, please apply.

The attached is an update to the one above, with the following additions
(recommended by Tom here -
http://archives.postgresql.org/pgsql-patches/2003-04/msg00030.php):

1. array_eq() now applies the element type's equality operator on an
element-by-element basis.
2. applied Tom's cache lookup results caching technique in arrayfuncs.c,
array_userfuncs.c, and in the new functions in varlena.c.

Still passes all regression tests. No additional doc updates needed for
this one. If there are no objections, please apply.

Thanks,

Joe

Joe Conway writes:

> Attached is an updated version of the doc patch along with the addition
> of two new functions, as previously discussed:
>
> - str_to_array(str TEXT, delim TEXT) returns TEXT[]
> - array_to_str(array ANYARRAY, delim TEXT) returns TEXT

Please use "string". No need to introduce a new abbreviation into the
world of function names.

I'm not happy with all these nonstandard new function names being
advertised. They can serve as internal implementations of standard
operators and other constructs, but it seems pointless to publicly add
redundant and nonstandard interfaces where the standard ones work
perfectly well.

In particular, array_subscript is not needed, users should simply use the
standard subscripting mechanism. array_prepend, array_cat, array_append
should not be used; instead the standard operator || should be used.
array_accum is not needed because || plus coalesce do the same thing
standardly. singleton_array strikes me as unnecessary as well.

Also note that "multidimensional array" is the correct spelling and
preferred wording (not "multi-dimensional array" or "multi-dimesion array"
[sic] or "multiple dimension array").

--
Peter Eisentraut peter_e(at)gmx(dot)net

Peter Eisentraut wrote:
> Joe Conway writes:
>>Attached is an updated version of the doc patch along with the addition
>>of two new functions, as previously discussed:
>>
>> - str_to_array(str TEXT, delim TEXT) returns TEXT[]
>> - array_to_str(array ANYARRAY, delim TEXT) returns TEXT
>
> Please use "string". No need to introduce a new abbreviation into the
> world of function names.

It's not new. We already have strpos and substr. But I'm not opposed to
using string_to_array and array_to_string if there's a consensus WRT
those names.

> Also note that "multidimensional array" is the correct spelling and
> preferred wording (not "multi-dimensional array" or "multi-dimesion array"
> [sic] or "multiple dimension array").

"multi-dimensional" is not incorrect by my dictionary, and most existing
examples of the word sprinkled throughout the source tree use this
spelling. Are you proposing that I change all of those too? Again, I'm
not opposed if there is general agreement, but we should at least be
consistent throughout.

In any case I'll correct "multi-dimesion" -- thanks for catching that
(new version of patch attached with spelling correction). I don't grep
"multiple dimension array" anywhere, so I'm not sure where you got that one.

Joe

# grep -r "multidimensional" *
contrib/cube/README.cube:CUBE, representing multidimensional cubes.
src/backend/utils/adt/arrayfuncs.c:
* successively in a multidimensional array.

# grep -r "multi-dimensional" *
contrib/cube/cube.sql.in:COMMENT ON TYPE cube IS 'multi-dimensional cube
''(FLOAT-1, FLOAT-2, ..., FLOAT-N), (FLOAT-1, FLOAT-2, ..., FLOAT-N)''';
doc/FAQ: single dimension. R-trees can handle multi-dimensional data. For
doc/FAQ_german: werden. R-Trees können multi-dimensionale Daten
abhandeln. Ein
doc/src/FAQ/FAQ.html: multi-dimensional data. For example, if an
R-tree index can be
doc/src/FAQ/FAQ_german.html: werden. R-Trees können
multi-dimensionale Daten abhandeln. Ein
doc/src/sgml/array.sgml: defined as variable-length multi-dimensional
arrays. Arrays of any
doc/src/sgml/array.sgml: Note that with this syntax, multi-dimensional
arrays must have matching
doc/src/sgml/array.sgml: multi-dimensional arrays.
doc/src/sgml/array.sgml: arrays, but <function>array_cat</function>
supports multi-dimensional arrays.
doc/src/sgml/array.sgml: use comma.) In a multi-dimensional array,
each dimension (row, plane,
doc/src/sgml/array.sgml: When representing multi-dimensional arrays,
the keyword
src/backend/utils/adt/array_userfuncs.c: * Form a
multi-dimensional array given one starting element.
src/interfaces/ecpg/preproc/preproc.y:
mmerror(PARSE_ERROR, ET_ERROR, "No multi-dimensional array support for
simple data types");
src/interfaces/ecpg/preproc/preproc.y:
mmerror(PARSE_ERROR, ET_ERROR, "No multi-dimensional array support for
simple data types");
src/interfaces/ecpg/preproc/preproc.y:
mmerror(PARSE_ERROR, ET_ERROR, "No multi-dimensional
array support for simple data types");
src/interfaces/ecpg/preproc/type.c:
yyerror("internal error, found multi-dimensional array\n");
src/interfaces/ecpg/preproc/variable.c:
mmerror(PARSE_ERROR, ET_FATAL, "No multi-dimensional array support");
src/interfaces/ecpg/preproc/variable.c:
mmerror(PARSE_ERROR, ET_FATAL, "No multi-dimensional array support");
src/interfaces/ecpg/preproc/variable.c: mmerror(PARSE_ERROR,
ET_FATAL, "No multi-dimensional array support");
src/interfaces/ecpg/preproc/variable.c: mmerror(PARSE_ERROR,
ET_FATAL, "No multi-dimensional array support");
src/interfaces/ecpg/preproc/variable.c:
mmerror(PARSE_ERROR, ET_FATAL, "No multi-dimensional array support for
structures");
src/interfaces/ecpg/preproc/variable.c:
mmerror(PARSE_ERROR, ET_FATAL, "No multi-dimensional array support for
simple data types");
src/interfaces/ecpg/preproc/preproc.c:
mmerror(PARSE_ERROR, ET_ERROR, "No multi-dimensional array support for
simple data types");
src/interfaces/ecpg/preproc/preproc.c:
mmerror(PARSE_ERROR, ET_ERROR, "No multi-dimensional array support for
simple data types");
src/interfaces/ecpg/preproc/preproc.c:
mmerror(PARSE_ERROR, ET_ERROR, "No multi-dimensional
array support for simple data types");

Joe Conway <mail(at)joeconway(dot)com> writes:
> Peter Eisentraut wrote:
>> Also note that "multidimensional array" is the correct spelling

> "multi-dimensional" is not incorrect by my dictionary,

The first two online dictionaries I checked (Merriam-Webster and
Cambridge) list *only* "multi-dimensional". IMHO the word is not
common enough in normal use to have lost its hyphen. You can get
away with not hyphenating in technical contexts where it's more
common, but that does not make the hyphenated form less correct.

regards, tom lane

Tom Lane writes:

> The first two online dictionaries I checked (Merriam-Webster and
> Cambridge) list *only* "multi-dimensional". IMHO the word is not
> common enough in normal use to have lost its hyphen. You can get
> away with not hyphenating in technical contexts where it's more
> common, but that does not make the hyphenated form less correct.

In my educated opinion, prefixes that are not words by themselves are
never hyphenated (except possibly to avoid ambiguity). This applies to
pre-, post-, non-, un-, in-, uni-, bi-, re- and so on, and there is no
grammatical reason why it shouldn't apply to multi- as well. The rule
that once a word becomes common enough it can be closed up only applies to
compounds where each part is an independent word. webster.com agrees with
this and also the dead tree dictionary I have here. There's also a more
detailed explanation of this here: <http://www.bartleby.com/64/84.html>.

--
Peter Eisentraut peter_e(at)gmx(dot)net

Peter Eisentraut wrote:
> Tom Lane writes:
>
>>The first two online dictionaries I checked (Merriam-Webster and
>>Cambridge) list *only* "multi-dimensional". IMHO the word is not
>>common enough in normal use to have lost its hyphen. You can get
>>away with not hyphenating in technical contexts where it's more
>>common, but that does not make the hyphenated form less correct.
>
> In my educated opinion, prefixes that are not words by themselves are
> never hyphenated (except possibly to avoid ambiguity). This applies to
> pre-, post-, non-, un-, in-, uni-, bi-, re- and so on, and there is no
> grammatical reason why it shouldn't apply to multi- as well. The rule
> that once a word becomes common enough it can be closed up only applies to
> compounds where each part is an independent word. webster.com agrees with
> this and also the dead tree dictionary I have here. There's also a more
> detailed explanation of this here: <http://www.bartleby.com/64/84.html>.
>

I see both forms in various references as being correct, but I think we
should consistently use one or the other. I'll make the change, but I'd
need a definitive answer on which way to go. Any other opinions out there?

Joe

Joe Conway wrote:
> Joe Conway wrote:
>
>> Attached is an updated version of the doc patch along with the
>> addition of two new functions, as previously discussed:
>>
>> - str_to_array(str TEXT, delim TEXT) returns TEXT[]
>> - array_to_str(array ANYARRAY, delim TEXT) returns TEXT
>>
>> The patch is fairly large because:
>>
>> 1. the documentation part of it is large
>> 2. I moved some static functions around and made them accessible so
>> that I could use them in implementing the two functions.
>>
>> It passes all regression tests, and as mentioned above includes
>> documentation.
>>
>> If there are no objections, please apply.
>
>
> The attached is an update to the one above, with the following additions
> (recommended by Tom here -
> http://archives.postgresql.org/pgsql-patches/2003-04/msg00030.php):
>
> 1. array_eq() now applies the element type's equality operator on an
> element-by-element basis.
> 2. applied Tom's cache lookup results caching technique in arrayfuncs.c,
> array_userfuncs.c, and in the new functions in varlena.c.
>
> Still passes all regression tests. No additional doc updates needed for
> this one. If there are no objections, please apply.
>

This attached patch includes all of the above, with the following changes:

1. array_type_coerce() fixes discussed off list last weekend while the
list was mostly down.
2. empty array element and slice assignment, and empty array
concatenation as recently discussed on HACKERS.
3. Aggregates can now use polymorphic functions and be polymorphic
themselves (examples below)
4. removes array_accum function and documentation per recent discussions

I specifically did not change the spelling of the term
multi-dimensional/multidimensional -- I'd suggest we agree on one way
and be consistent, but I'd really rather not hold up this patch for that
as it is getting tough to stay in sync with cvs (had a backend function
disappear from under me just today).

I also left singleton_array, array_assign, and array_subscript in place
and documented. I'd still like to see them stay (there was a request on
GENERAL this afternoon that could have used singleton_array as probably
the most simple solution, although undoubtedly they could have kludged
it by concatenating to an empty array). But, again, I'd rather see this
patch committed, so if need be, please rip them out.

Here are some examples of the polymorphic aggregates:

create table t(f1 int, f2 int[], f3 text);
insert into t values(1,array[1],'a');
insert into t values(1,array[11],'b');
insert into t values(1,array[111],'c');
insert into t values(2,array[2],'a');
insert into t values(2,array[22],'b');
insert into t values(2,array[222],'c');
insert into t values(3,array[3],'a');
insert into t values(3,array[3],'b');

CREATE AGGREGATE myagg2
(
BASETYPE = int4,
SFUNC = array_append,
STYPE = int4[],
INITCOND = '{}'
);

CREATE AGGREGATE myagg3
(
BASETYPE = anyelement,
SFUNC = array_append,
STYPE = anyarray,
INITCOND = '{}'
);

CREATE AGGREGATE myagg4
(
BASETYPE = anyarray,
SFUNC = array_cat,
STYPE = anyarray,
INITCOND = '{}'
);

regression=# select f3, myagg2(f1) from t group by f3;
f3 | myagg2
----+---------
b | {1,2,3}
a | {1,2,3}
c | {1,2}
(3 rows)

regression=# select f3, myagg3(f1) from t group by f3;
f3 | myagg3
----+---------
b | {1,2,3}
a | {1,2,3}
c | {1,2}
(3 rows)

regression=# select f1, myagg3(f3) from t group by f1;
f1 | myagg3
----+---------
1 | {a,b,c}
3 | {a,b}
2 | {a,b,c}
(3 rows)

regression=# select f3, myagg4(f2) from t group by f3;
f3 | myagg4
----+-----------------
b | {{11},{22},{3}}
a | {{1},{2},{3}}
c | {{111},{222}}
(3 rows)

And assignment to an empty array:

regression=# insert into t values(4,'{}','d');
INSERT 2000865 1
regression=# update t set f2[2][4][6] = 42 where f1 = 4;
UPDATE 1
regression=# select * from t where f1 = 4;
f1 | f2 | f3
----+----------+----
4 | {{{42}}} | d
(1 row)

regression=# select array_dims(f2) from t where f1 = 4;
array_dims
-----------------
[2:2][4:4][6:6]
(1 row)

All 90 regression tests passed. More doc updates are needed based on
these changes, but I'd like to do those after the feature freeze if I
may. Please apply.

thanks,

Joe

Tom Lane wrote:
> Joe Conway <mail(at)joeconway(dot)com> writes:
>>Maybe this takes us
>>back to Peter's suggestion:
>> expression IN (array)
>> expression NOT IN (array)
>> expression operator ANY (array)
>> expression operator SOME (array)
>> (expression) operator (array)
>> (expression) operator ALL (array)
>
> There's a lot to be said for that, if we can think of a way to do it.
> I believe this way would force us to integrate the operations into the
> parser, though, rather than just throwing a few polymorphic functions
> at the problem. It's probably a lot more work :-(

The attached implements the middle 2 examples above. The IN and NOT IN
examples would be ambiguous due to the fact that we already accept
"IN (list_of_expressions)" and "NOT IN (list_of_expressions)".

Specifically implemented:
expression1 operator ANY (expression2)
expression1 operator SOME (expression2)
expression1 operator ALL (expression2)

If expression2 is scalar, it simplifies down to the same as doing:
expression1 operator ANY | SOME | ALL
(select expression2)

If expression2 is an array, the array elements are tested as if you did
this:
expression1 operator ANY | SOME | ALL
(select element_a
union all
select element_b...
union all
select element_n)

Here are a few examples:

create table tse(f1 int, f2 int[], f3 text[]);
insert into tse values(1,array[69,42,54], array['g','d','e']);
insert into tse values(2,array[1,2,3], array['x','y','z']);
insert into tse values(3,array[2,99,0], array['Tom','Bruce']);
insert into tse values(4,array[1,1,1], array['a','a','a']);
insert into tse values(5,array[5,6,7], array['a','b','c']);

regression=# select * from tse where 1 = any (f2);
f1 | f2 | f3
----+---------+---------
2 | {1,2,3} | {x,y,z}
4 | {1,1,1} | {a,a,a}
(2 rows)

regression=# select * from tse where 1 = all (f2);
f1 | f2 | f3
----+---------+---------
4 | {1,1,1} | {a,a,a}
(1 row)

regression=# select * from tse where 'a' != any (f3);
f1 | f2 | f3
----+------------+-------------
1 | {69,42,54} | {g,d,e}
2 | {1,2,3} | {x,y,z}
3 | {2,99,0} | {Tom,Bruce}
5 | {5,6,7} | {a,b,c}
(4 rows)

regression=# select * from tse where 'a' != all (f3);
f1 | f2 | f3
----+------------+-------------
1 | {69,42,54} | {g,d,e}
2 | {1,2,3} | {x,y,z}
3 | {2,99,0} | {Tom,Bruce}
(3 rows)

regression=# select * from tse where 'y' < some (f3);
f1 | f2 | f3
----+---------+---------
2 | {1,2,3} | {x,y,z}
(1 row)

The patch passes all regression tests. It is independent of the phase2
patch submitted earlier this week.

I am already planning to update the array related docs one more time
after the feature freeze, so I'd like to wait until then to document
this. If there are no objections, please apply.

Thanks,

Joe

Tom Lane wrote:
> Another problem is:
>
> regression=# select distinct array(select * from text_tbl) from foo;
> ERROR: Unable to identify an equality operator for type text[]
>
> This DISTINCT query would fail anyway of course, for lack of a '<'
> operator for arrays, but it seems like we ought to be able to find the
> polymorphic '=' operator.

The attached small patch addresses the above concern (finding
polymorphic "=" operator) by introducing a new function,
IsPolymorphicCoercible().

Now, as predicted, I get:

regression=# select distinct array(select f1 from tse);
ERROR: Unable to identify an ordering operator for type integer[]
Use an explicit ordering operator or modify the query

If there are no objections, please apply.

Next question this begs is, should I work on a '<' operator for arrays?
And if so, how is the behavior defined? Hannu suggested
element-by-element, analogous to character-by-character text string
comparison. Comments?

Thanks,

Joe

Joe Conway <mail(at)joeconway(dot)com> writes:
> + /* last check -- polymorphic types? */
> + if (IsPolymorphicCoercible(arg1 , arg2))
> + return optup;

That is surely not what you intended. The test must be whether arg1 and
arg2 are (separately) coercible to the operator's two input types.
Moreover, the test must not be symmetric, any more than
IsBinaryCoercible is. You can coerce int[] to ANYARRAY but not vice
versa.

A bigger problem is that I doubt this will actually work. Most of the
places that call compatible_oper will then proceed to call the function
from specialized code that does not bother with consing up an expression
tree --- so a polymorphic function is going to fail anyway...

regards, tom lane

Tom Lane wrote:
> That is surely not what you intended. The test must be whether arg1 and
> arg2 are (separately) coercible to the operator's two input types.
> Moreover, the test must not be symmetric, any more than
> IsBinaryCoercible is. You can coerce int[] to ANYARRAY but not vice
> versa.

Dooh! Yeah, I can see that now.

>
> A bigger problem is that I doubt this will actually work. Most of the
> places that call compatible_oper will then proceed to call the function
> from specialized code that does not bother with consing up an expression
> tree --- so a polymorphic function is going to fail anyway...

Well, not necessarily in the case of array_type-to-ANYARRAY. In that
case the element type information in the array itself gives the function
all the context it needs (if it looks there, which in the case of
array_eq at least it does). Maybe it makes sense to only allow the
array_type-to-ANYARRAY case?

Joe

Tom Lane wrote:
> That is surely not what you intended. The test must be whether arg1 and
> arg2 are (separately) coercible to the operator's two input types.
> Moreover, the test must not be symmetric, any more than
> IsBinaryCoercible is. You can coerce int[] to ANYARRAY but not vice
> versa.
>
> A bigger problem is that I doubt this will actually work. Most of the
> places that call compatible_oper will then proceed to call the function
> from specialized code that does not bother with consing up an expression
> tree --- so a polymorphic function is going to fail anyway...
>

How's this attempt?

Couple of issues it raised:

1.) In conjunction with the phase2 patch, which changes array_eq to do
an element-by-element equality comparison, I discovered that type
"aclitem" has no equality operator, and apparently nothing
binary-coercible either. Should I invent "=" for aclitem, or have
array_eq fall back to its original byte-by-byte comparison if it doesn't
find an equality operator?

2.) I noticed that if I start the postmaster without redirecting to a
logfile, I see double error messages in psql. Is this intended? Maybe
due to some conf setting?

[postgres(at)dev postgres]$ pg_ctl start
LOG: database system was shut down at 2003-06-01 08:58:18 PDT
IN: StartupXLOG (xlog.c:2510)
LOG: checkpoint record is at 0/5BCDA4
IN: StartupXLOG (xlog.c:2538)
LOG: redo record is at 0/5BCDA4; undo record is at 0/0; shutdown TRUE
IN: StartupXLOG (xlog.c:2558)
LOG: next transaction id: 475; next oid: 17079
IN: StartupXLOG (xlog.c:2562)
postmaster successfully started
[postgres(at)dev postgres]$ LOG: database system is ready
IN: StartupXLOG (xlog.c:2819)

[postgres(at)dev postgres]$ psql template1
Welcome to psql 7.4devel, the PostgreSQL interactive terminal.

Type: \copyright for distribution terms
\h for help with SQL commands
\? for help on internal slash commands
\g or terminate with semicolon to execute query
\q to quit

template1=# select distinct array(select f1 from tse);
ERROR: Unable to identify an ordering operator for type integer[]
Use an explicit ordering operator or modify the query
IN: ordering_oper (parse_oper.c:202)
ERROR: Unable to identify an ordering operator for type integer[]
Use an explicit ordering operator or modify the query

Joe

Joe Conway <mail(at)joeconway(dot)com> writes:
>> A bigger problem is that I doubt this will actually work. Most of the
>> places that call compatible_oper will then proceed to call the function
>> from specialized code that does not bother with consing up an expression
>> tree --- so a polymorphic function is going to fail anyway...

> Well, not necessarily in the case of array_type-to-ANYARRAY. In that
> case the element type information in the array itself gives the function
> all the context it needs (if it looks there, which in the case of
> array_eq at least it does). Maybe it makes sense to only allow the
> array_type-to-ANYARRAY case?

Yeah, I think we could get away with that. It might be appropriate to
put that single special case into IsBinaryCoercible, instead of allowing
it only for the compatible_oper case. I can't recall offhand what else
uses IsBinaryCoercible ...

regards, tom lane

Joe Conway <mail(at)joeconway(dot)com> writes:
> 1.) In conjunction with the phase2 patch, which changes array_eq to do
> an element-by-element equality comparison, I discovered that type
> "aclitem" has no equality operator, and apparently nothing
> binary-coercible either. Should I invent "=" for aclitem, or have
> array_eq fall back to its original byte-by-byte comparison if it doesn't
> find an equality operator?

The former I would think.

> 2.) I noticed that if I start the postmaster without redirecting to a
> logfile, I see double error messages in psql. Is this intended?

Yeah, if postmaster stderr is coming to your console ...

regards, tom lane

Tom Lane wrote:
> Yeah, I think we could get away with that. It might be appropriate to
> put that single special case into IsBinaryCoercible, instead of allowing
> it only for the compatible_oper case. I can't recall offhand what else
> uses IsBinaryCoercible ...
>

OK -- here is that version. Certainly simplifies it, and seems
appropriate given the comments in IsBinaryCoercible.

Next question: should I roll the three array related patches floating
around (phase2, phase3, polycoerce) into one big patch again? It's
difficult for me to continue to make progress without doing that.

Joe

Joe Conway <mail(at)joeconway(dot)com> writes:
> Next question: should I roll the three array related patches floating
> around (phase2, phase3, polycoerce) into one big patch again? It's
> difficult for me to continue to make progress without doing that.

If it's easier at your end.

> + /* Last of the fast-paths: check for matching polymorphic arrays */
> + if (targettype == ANYARRAYOID)
> + if (get_element_type(srctype) != InvalidOid)
> + return true;

This doesn't seem quite right; won't the second test succeed for NAME
and other fixed-length-array types? You need to restrict it to varlena
arrays, I'd think.

regards, tom lane

Tom Lane wrote:
>>+ /* Last of the fast-paths: check for matching polymorphic arrays */
>>+ if (targettype == ANYARRAYOID)
>>+ if (get_element_type(srctype) != InvalidOid)
>>+ return true;
>
>
> This doesn't seem quite right; won't the second test succeed for NAME
> and other fixed-length-array types? You need to restrict it to varlena
> arrays, I'd think.
>

get_element_type already does -- in fact, I think that part was you're
innovation because originally I had get_typelem() which didn't make that
distinction ;-)

/*
* get_element_type
*
* Given the type OID, get the typelem (InvalidOid if not an array
* type).
*
* NB: this only considers varlena arrays to be true arrays;
* InvalidOid is returned if the input is a fixed-length array type.
*/

Joe

Joe Conway wrote:
> Next question this begs is, should I work on a '<' operator for arrays?
> And if so, how is the behavior defined? Hannu suggested
> element-by-element, analogous to character-by-character text string
> comparison. Comments?

I created an equality operator for aclitem, and then all was well until
I finished the above (i.e. I created array-array operators for "<>",
"<", ">", "<=", and ">=" based on the text comparison analogy). Now on
ANALYZE (and therefore on initdb) I get this:

template1=# analyze;
ERROR: Unable to identify an ordering operator for type aclitem
Use an explicit ordering operator or modify the query

This is where I ran into the problem with an equality operator earlier.
Problem this time is that an ordering operator makes no sense for
aclitems -- does it? Any thoughts on how to work around this?

Joe

Tom Lane wrote:
> Joe Conway <mail(at)joeconway(dot)com> writes:
>>Next question: should I roll the three array related patches floating
>>around (phase2, phase3, polycoerce) into one big patch again? It's
>>difficult for me to continue to make progress without doing that.
>
> If it's easier at your end.

Attached is the "super combo" array patch ;-)

It includes the following previously sent patches:
array-phase2.19.patch
array-phase3.05.patch
array-polycoerce.3.patch

It also includes the following additional changes:
1) aclitem type now has an equality operator
2) examine_attribute() now checks for the element type's equality and
ordering operators when dealing with a varlena array type
3) There are now "<>, ">", "<", "<=", and ">=" operators for
array-to-array comparison, using semantics similar to text's
character-to-character comparison.
4) There is also now a btarraycmp() function and other required
btree support. BTree indexes can now be created on varlena arrays.

Below are some examples of the new (since earlier patches)
functionality. CVS tip from this morning plus this patch compiles
cleanly and passes all 90 regression tests on my dev box. If there are
no objections, please apply.

Thanks,

Joe

Examples:
------------------------------------------
create table tse(f1 int, f2 int[], f3 text[]);
insert into tse values(1,array[69,42,54], array['g','d','e']);
insert into tse values(2,array[1,2,3], array['x','y','z']);
insert into tse values(3,array[2,99,0], array['Tom','Bruce']);
insert into tse values(4,array[1,1,1], array['a','a','a']);
insert into tse values(5,array[5,6,7], array['a','b','c']);

regression=# select distinct array(select f1 from tse);
?column?
-------------
{1,2,3,4,5}
(1 row)

regression=# select * from tse where f2 < array[1,2,3,4];
f1 | f2 | f3
----+---------+---------
2 | {1,2,3} | {x,y,z}
4 | {1,1,1} | {a,a,a}
(2 rows)

regression=# select * from tse where f2 < array[1,2,3];
f1 | f2 | f3
----+---------+---------
4 | {1,1,1} | {a,a,a}
(1 row)

regression=# select * from tse where f2 <= array[1,2,3];
f1 | f2 | f3
----+---------+---------
2 | {1,2,3} | {x,y,z}
4 | {1,1,1} | {a,a,a}
(2 rows)

regression=# select * from tse where f2 > array[1,2,3];
f1 | f2 | f3
----+------------+-------------
1 | {69,42,54} | {g,d,e}
3 | {2,99,0} | {Tom,Bruce}
5 | {5,6,7} | {a,b,c}
(3 rows)

regression=# select * from tse where f2 >= array[1,2,3];
f1 | f2 | f3
----+------------+-------------
1 | {69,42,54} | {g,d,e}
2 | {1,2,3} | {x,y,z}
3 | {2,99,0} | {Tom,Bruce}
5 | {5,6,7} | {a,b,c}
(4 rows)

regression=# select * from tse where f2 != array[1,2,3];
f1 | f2 | f3
----+------------+-------------
1 | {69,42,54} | {g,d,e}
3 | {2,99,0} | {Tom,Bruce}
4 | {1,1,1} | {a,a,a}
5 | {5,6,7} | {a,b,c}
(4 rows)

regression=# select * from tse where f2 = array[1,2,3];
f1 | f2 | f3
----+---------+---------
2 | {1,2,3} | {x,y,z}
(1 row)

regression=# select * from tse where f3 >= array['Tom'] and f3 < array['a'];
f1 | f2 | f3
----+----------+-------------
3 | {2,99,0} | {Tom,Bruce}
(1 row)

regression=# create index arridx1 on tse(f2);
CREATE INDEX
regression=# set enable_seqscan to off;
SET
regression=# explain analyze select * from tse where f2 = array[1,2,3];
QUERY PLAN
-----------------------------------------------------------------------
Index Scan using arridx1 on tse (cost=0.00..4.68 rows=2 width=68)
(actual time=0.07..0.07 rows=1 loops=1)
Index Cond: (f2 = '{1,2,3}'::integer[])
Total runtime: 0.15 msec
(3 rows)

(I assume this includes all your outstanding unapplied array patches.)

Your patch has been added to the PostgreSQL unapplied patches list at:

http://momjian.postgresql.org/cgi-bin/pgpatches

I will try to apply it within the next 48 hours.

---------------------------------------------------------------------------

Joe Conway wrote:
> Tom Lane wrote:
> > Joe Conway <mail(at)joeconway(dot)com> writes:
> >>Next question: should I roll the three array related patches floating
> >>around (phase2, phase3, polycoerce) into one big patch again? It's
> >>difficult for me to continue to make progress without doing that.
> >
> > If it's easier at your end.
>
> Attached is the "super combo" array patch ;-)
>
> It includes the following previously sent patches:
> array-phase2.19.patch
> array-phase3.05.patch
> array-polycoerce.3.patch
>
> It also includes the following additional changes:
> 1) aclitem type now has an equality operator
> 2) examine_attribute() now checks for the element type's equality and
> ordering operators when dealing with a varlena array type
> 3) There are now "<>, ">", "<", "<=", and ">=" operators for
> array-to-array comparison, using semantics similar to text's
> character-to-character comparison.
> 4) There is also now a btarraycmp() function and other required
> btree support. BTree indexes can now be created on varlena arrays.
>
> Below are some examples of the new (since earlier patches)
> functionality. CVS tip from this morning plus this patch compiles
> cleanly and passes all 90 regression tests on my dev box. If there are
> no objections, please apply.
>
> Thanks,
>
> Joe
>
>
> Examples:
> ------------------------------------------
> create table tse(f1 int, f2 int[], f3 text[]);
> insert into tse values(1,array[69,42,54], array['g','d','e']);
> insert into tse values(2,array[1,2,3], array['x','y','z']);
> insert into tse values(3,array[2,99,0], array['Tom','Bruce']);
> insert into tse values(4,array[1,1,1], array['a','a','a']);
> insert into tse values(5,array[5,6,7], array['a','b','c']);
>
> regression=# select distinct array(select f1 from tse);
> ?column?
> -------------
> {1,2,3,4,5}
> (1 row)
>
> regression=# select * from tse where f2 < array[1,2,3,4];
> f1 | f2 | f3
> ----+---------+---------
> 2 | {1,2,3} | {x,y,z}
> 4 | {1,1,1} | {a,a,a}
> (2 rows)
>
> regression=# select * from tse where f2 < array[1,2,3];
> f1 | f2 | f3
> ----+---------+---------
> 4 | {1,1,1} | {a,a,a}
> (1 row)
>
> regression=# select * from tse where f2 <= array[1,2,3];
> f1 | f2 | f3
> ----+---------+---------
> 2 | {1,2,3} | {x,y,z}
> 4 | {1,1,1} | {a,a,a}
> (2 rows)
>
> regression=# select * from tse where f2 > array[1,2,3];
> f1 | f2 | f3
> ----+------------+-------------
> 1 | {69,42,54} | {g,d,e}
> 3 | {2,99,0} | {Tom,Bruce}
> 5 | {5,6,7} | {a,b,c}
> (3 rows)
>
> regression=# select * from tse where f2 >= array[1,2,3];
> f1 | f2 | f3
> ----+------------+-------------
> 1 | {69,42,54} | {g,d,e}
> 2 | {1,2,3} | {x,y,z}
> 3 | {2,99,0} | {Tom,Bruce}
> 5 | {5,6,7} | {a,b,c}
> (4 rows)
>
> regression=# select * from tse where f2 != array[1,2,3];
> f1 | f2 | f3
> ----+------------+-------------
> 1 | {69,42,54} | {g,d,e}
> 3 | {2,99,0} | {Tom,Bruce}
> 4 | {1,1,1} | {a,a,a}
> 5 | {5,6,7} | {a,b,c}
> (4 rows)
>
> regression=# select * from tse where f2 = array[1,2,3];
> f1 | f2 | f3
> ----+---------+---------
> 2 | {1,2,3} | {x,y,z}
> (1 row)
>
> regression=# select * from tse where f3 >= array['Tom'] and f3 < array['a'];
> f1 | f2 | f3
> ----+----------+-------------
> 3 | {2,99,0} | {Tom,Bruce}
> (1 row)
>
> regression=# create index arridx1 on tse(f2);
> CREATE INDEX
> regression=# set enable_seqscan to off;
> SET
> regression=# explain analyze select * from tse where f2 = array[1,2,3];
> QUERY PLAN
> -----------------------------------------------------------------------
> Index Scan using arridx1 on tse (cost=0.00..4.68 rows=2 width=68)
> (actual time=0.07..0.07 rows=1 loops=1)
> Index Cond: (f2 = '{1,2,3}'::integer[])
> Total runtime: 0.15 msec
> (3 rows)

> Index: doc/src/sgml/array.sgml
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/doc/src/sgml/array.sgml,v
> retrieving revision 1.25
> diff -c -r1.25 array.sgml
> *** doc/src/sgml/array.sgml 13 Mar 2003 01:30:26 -0000 1.25
> --- doc/src/sgml/array.sgml 1 Jun 2003 15:41:15 -0000
> ***************
> *** 9,15 ****
>
> <para>
> <productname>PostgreSQL</productname> allows columns of a table to be
> ! defined as variable-length multidimensional arrays. Arrays of any
> built-in type or user-defined type can be created.
> </para>
>
> --- 9,15 ----
>
> <para>
> <productname>PostgreSQL</productname> allows columns of a table to be
> ! defined as variable-length multi-dimensional arrays. Arrays of any
> built-in type or user-defined type can be created.
> </para>
>
> ***************
> *** 60,73 ****
> </programlisting>
> </para>
>
> <note>
> <para>
> ! A limitation of the present array implementation is that individual
> ! elements of an array cannot be SQL null values. The entire array can be set
> ! to null, but you can't have an array with some elements null and some
> ! not. Fixing this is on the to-do list.
> </para>
> </note>
> </sect2>
>
> <sect2>
> --- 60,133 ----
> </programlisting>
> </para>
>
> + <para>
> + A limitation of the present array implementation is that individual
> + elements of an array cannot be SQL null values. The entire array can be set
> + to null, but you can't have an array with some elements null and some
> + not.
> + </para>
> + <para>
> + This can lead to surprising results. For example, the result of the
> + previous two inserts looks like this:
> + <programlisting>
> + SELECT * FROM sal_emp;
> + name | pay_by_quarter | schedule
> + -------+---------------------------+--------------------
> + Bill | {10000,10000,10000,10000} | {{meeting},{""}}
> + Carol | {20000,25000,25000,25000} | {{talk},{meeting}}
> + (2 rows)
> + </programlisting>
> + Because the <literal>[2][2]</literal> element of
> + <structfield>schedule</structfield> is missing in each of the
> + <command>INSERT</command> statements, the <literal>[1][2]</literal>
> + element is discarded.
> + </para>
> +
> + <note>
> + <para>
> + Fixing this is on the to-do list.
> + </para>
> + </note>
> +
> + <para>
> + The <command>ARRAY</command> expression syntax may also be used:
> + <programlisting>
> + INSERT INTO sal_emp
> + VALUES ('Bill',
> + ARRAY[10000, 10000, 10000, 10000],
> + ARRAY[['meeting', 'lunch'], ['','']]);
> +
> + INSERT INTO sal_emp
> + VALUES ('Carol',
> + ARRAY[20000, 25000, 25000, 25000],
> + ARRAY[['talk', 'consult'], ['meeting', '']]);
> + SELECT * FROM sal_emp;
> + name | pay_by_quarter | schedule
> + -------+---------------------------+-------------------------------
> + Bill | {10000,10000,10000,10000} | {{meeting,lunch},{"",""}}
> + Carol | {20000,25000,25000,25000} | {{talk,consult},{meeting,""}}
> + (2 rows)
> + </programlisting>
> + Note that with this syntax, multi-dimensional arrays must have matching
> + extents for each dimension. This eliminates the missing-array-elements
> + problem above. For example:
> + <programlisting>
> + INSERT INTO sal_emp
> + VALUES ('Carol',
> + ARRAY[20000, 25000, 25000, 25000],
> + ARRAY[['talk', 'consult'], ['meeting']]);
> + ERROR: Multiple dimension arrays must have array expressions with matching dimensions
> + </programlisting>
> + Also notice that string literals are single quoted instead of double quoted.
> + </para>
> +
> <note>
> <para>
> ! The examples in the rest of this section are based on the
> ! <command>ARRAY</command> expression syntax <command>INSERT</command>s.
> </para>
> </note>
> +
> </sect2>
>
> <sect2>
> ***************
> *** 132,142 ****
> </programlisting>
>
> with the same result. An array subscripting operation is always taken to
> ! represent an array slice if any of the subscripts are written in the
> ! form
> <literal><replaceable>lower</replaceable>:<replaceable>upper</replaceable></literal>.
> A lower bound of 1 is assumed for any subscript where only one value
> ! is specified.
> </para>
>
> <para>
> --- 192,221 ----
> </programlisting>
>
> with the same result. An array subscripting operation is always taken to
> ! represent an array slice if any of the subscripts are written in the form
> <literal><replaceable>lower</replaceable>:<replaceable>upper</replaceable></literal>.
> A lower bound of 1 is assumed for any subscript where only one value
> ! is specified; another example follows:
> ! <programlisting>
> ! SELECT schedule[1:2][2] FROM sal_emp WHERE name = 'Bill';
> ! schedule
> ! ---------------------------
> ! {{meeting,lunch},{"",""}}
> ! (1 row)
> ! </programlisting>
> ! </para>
> !
> ! <para>
> ! Additionally, we can also access a single arbitrary array element of
> ! a one-dimensional array with the <function>array_subscript</function>
> ! function:
> ! <programlisting>
> ! SELECT array_subscript(pay_by_quarter, 2) FROM sal_emp WHERE name = 'Bill';
> ! array_subscript
> ! -----------------
> ! 10000
> ! (1 row)
> ! </programlisting>
> </para>
>
> <para>
> ***************
> *** 147,153 ****
> WHERE name = 'Carol';
> </programlisting>
>
> ! or updated at a single element:
>
> <programlisting>
> UPDATE sal_emp SET pay_by_quarter[4] = 15000
> --- 226,248 ----
> WHERE name = 'Carol';
> </programlisting>
>
> ! or using the <command>ARRAY</command> expression syntax:
> !
> ! <programlisting>
> ! UPDATE sal_emp SET pay_by_quarter = ARRAY[25000,25000,27000,27000]
> ! WHERE name = 'Carol';
> ! </programlisting>
> !
> ! <note>
> ! <para>
> ! Anywhere you can use the <quote>curly braces</quote> array syntax,
> ! you can also use the <command>ARRAY</command> expression syntax. The
> ! remainder of this section will illustrate only one or the other, but
> ! not both.
> ! </para>
> ! </note>
> !
> ! An array may also be updated at a single element:
>
> <programlisting>
> UPDATE sal_emp SET pay_by_quarter[4] = 15000
> ***************
> *** 160,165 ****
> --- 255,268 ----
> UPDATE sal_emp SET pay_by_quarter[1:2] = '{27000,27000}'
> WHERE name = 'Carol';
> </programlisting>
> +
> + A one-dimensional array may also be updated with the
> + <function>array_assign</function> function:
> +
> + <programlisting>
> + UPDATE sal_emp SET pay_by_quarter = array_assign(pay_by_quarter, 4, 15000)
> + WHERE name = 'Bill';
> + </programListing>
> </para>
>
> <para>
> ***************
> *** 169,175 ****
> value currently has 4 elements, it will have five elements after an
> update that assigns to <literal>array[5]</>. Currently, enlargement in
> this fashion is only allowed for one-dimensional arrays, not
> ! multidimensional arrays.
> </para>
>
> <para>
> --- 272,278 ----
> value currently has 4 elements, it will have five elements after an
> update that assigns to <literal>array[5]</>. Currently, enlargement in
> this fashion is only allowed for one-dimensional arrays, not
> ! multi-dimensional arrays.
> </para>
>
> <para>
> ***************
> *** 179,184 ****
> --- 282,367 ----
> </para>
>
> <para>
> + An array can also be enlarged by using the functions
> + <function>array_prepend</function>, <function>array_append</function>,
> + or <function>array_cat</function>. The first two only support one-dimensional
> + arrays, but <function>array_cat</function> supports multi-dimensional arrays.
> + Some examples:
> +
> + <programlisting>
> + SELECT array_prepend(1, ARRAY[2,3]);
> + array_prepend
> + ---------------
> + {1,2,3}
> + (1 row)
> +
> + SELECT array_append(ARRAY[1,2], 3);
> + array_append
> + --------------
> + {1,2,3}
> + (1 row)
> +
> + SELECT array_cat(ARRAY[1,2], ARRAY[3,4]);
> + array_cat
> + ---------------
> + {{1,2},{3,4}}
> + (1 row)
> +
> + SELECT array_cat(ARRAY[[1,2],[3,4]], ARRAY[5,6]);
> + array_cat
> + ---------------------
> + {{1,2},{3,4},{5,6}}
> + (1 row)
> +
> + SELECT array_cat(ARRAY[5,6], ARRAY[[1,2],[3,4]]);
> + array_cat
> + ---------------------
> + {{5,6},{1,2},{3,4}}
> + </programlisting>
> +
> + <function>array_prepend</function> and <function>array_append</function>
> + work with a one-dimensional array and a single element to be pushed on
> + to the beginning or end of the array, respectively. The array is extended
> + in the direction of the push. Hence, by pushing onto the beginning of an
> + array with a one-based subscript, a zero-based subscript array is created:
> +
> + <programlisting>
> + SELECT array_dims(t.f) FROM (SELECT array_prepend(1, ARRAY[2,3]) AS f) AS t;
> + array_dims
> + ------------
> + [0:2]
> + (1 row)
> + </programlisting>
> +
> + <function>array_cat</function> works with either two
> + <replaceable>n</>-dimension arrays, or an <replaceable>n</>-dimension
> + and an <replaceable>n+1</> dimension array. In the former case, the two
> + <replaceable>n</>-dimension arrays become outer elements of an
> + <replaceable>n+1</> dimension array. In the latter, the
> + <replaceable>n</>-dimension array is added as either the first or last
> + outer element of the <replaceable>n+1</> dimension array.
> + </para>
> +
> + <para>
> + A final method of enlarging arrays is through the concatenation operator,
> + <command>||</command>, which works exactly as <function>array_cat</function>
> + does.
> + <programlisting>
> + SELECT ARRAY[1,2] || ARRAY[3,4];
> + ?column?
> + ---------------
> + {{1,2},{3,4}}
> + (1 row)
> +
> + SELECT ARRAY[5,6] || ARRAY[[1,2],[3,4]];
> + ?column?
> + ---------------------
> + {{5,6},{1,2},{3,4}}
> + (1 row)
> + </programlisting>
> + </para>
> +
> + <para>
> The syntax for <command>CREATE TABLE</command> allows fixed-length
> arrays to be defined:
>
> ***************
> *** 194,199 ****
> --- 377,392 ----
> </para>
>
> <para>
> + An alternative syntax for one-dimensional arrays may be used.
> + <structfield>pay_by_quarter</structfield> could have been defined as:
> + <programlisting>
> + pay_by_quarter integer ARRAY[4],
> + </programlisting>
> + This syntax may <emphasis>only</emphasis> be used with the integer
> + constant to denote the array size.
> + </para>
> +
> + <para>
> Actually, the current implementation does not enforce the declared
> number of dimensions either. Arrays of a particular element type are
> all considered to be of the same type, regardless of size or number
> ***************
> *** 292,298 ****
> for the array's element type. (Among the standard data types provided
> in the <productname>PostgreSQL</productname> distribution, type
> <literal>box</> uses a semicolon (<literal>;</>) but all the others
> ! use comma.) In a multidimensional array, each dimension (row, plane,
> cube, etc.) gets its own level of curly braces, and delimiters
> must be written between adjacent curly-braced entities of the same level.
> You may write whitespace before a left brace, after a right
> --- 485,491 ----
> for the array's element type. (Among the standard data types provided
> in the <productname>PostgreSQL</productname> distribution, type
> <literal>box</> uses a semicolon (<literal>;</>) but all the others
> ! use comma.) In a multi-dimensional array, each dimension (row, plane,
> cube, etc.) gets its own level of curly braces, and delimiters
> must be written between adjacent curly-braced entities of the same level.
> You may write whitespace before a left brace, after a right
> ***************
> *** 300,305 ****
> --- 493,564 ----
> is not ignored, however: after skipping leading whitespace, everything
> up to the next right brace or delimiter is taken as the item value.
> </para>
> +
> + <para>
> + As illustrated earlier in this chapter, arrays may also be represented
> + using the <command>ARRAY</command> expression syntax. This representation
> + of an array value consists of items that are interpreted according to the
> + I/O conversion rules for the array's element type, plus decoration that
> + indicates the array structure. The decoration consists of the keyword
> + <command>ARRAY</command> and square brackets (<literal>[</> and
> + <literal>]</>) around the array values, plus delimiter characters between
> + adjacent items. The delimiter character is always a comma (<literal>,</>).
> + When representing multi-dimensional arrays, the keyword
> + <command>ARRAY</command> is only necessary for the outer level. For example,
> + <literal>'{{"hello world", "happy birthday"}}'</literal> could be written as:
> + <programlisting>
> + SELECT ARRAY[['hello world', 'happy birthday']];
> + array
> + ------------------------------------
> + {{"hello world","happy birthday"}}
> + (1 row)
> + </programlisting>
> + or it also could be written as:
> + <programlisting>
> + SELECT ARRAY[ARRAY['hello world', 'happy birthday']];
> + array
> + ------------------------------------
> + {{"hello world","happy birthday"}}
> + (1 row)
> + </programlisting>
> + </para>
> +
> + <para>
> + A final method to represent an array, is through an
> + <command>ARRAY</command> sub-select expression. For example:
> + <programlisting>
> + SELECT ARRAY(SELECT oid FROM pg_proc WHERE proname LIKE 'bytea%');
> + ?column?
> + -------------------------------------------------------------
> + {2011,1954,1948,1952,1951,1244,1950,2005,1949,1953,2006,31}
> + (1 row)
> + </programlisting>
> + The sub-select may <emphasis>only</emphasis> return a single column. The
> + resulting one-dimensional array will have an element for each row in the
> + sub-select result, with an element type matching that of the sub-select's
> + target column.
> + </para>
> +
> + <para>
> + Arrays may be cast from one type to another in similar fashion to other
> + data types:
> +
> + <programlisting>
> + SELECT ARRAY[1,2,3]::oid[];
> + array
> + ---------
> + {1,2,3}
> + (1 row)
> +
> + SELECT CAST(ARRAY[1,2,3] AS float8[]);
> + array
> + ---------
> + {1,2,3}
> + (1 row)
> + </programlisting>
> +
> + </para>
> +
> </sect2>
>
> <sect2>
> ***************
> *** 316,321 ****
> --- 575,588 ----
> Alternatively, you can use backslash-escaping to protect all data characters
> that would otherwise be taken as array syntax or ignorable white space.
> </para>
> +
> + <note>
> + <para>
> + The discussion in the preceding paragraph with respect to double quoting does
> + not pertain to the <command>ARRAY</command> expression syntax. In that case,
> + each element is quoted exactly as any other literal value of the element type.
> + </para>
> + </note>
>
> <para>
> The array output routine will put double quotes around element values
> Index: doc/src/sgml/func.sgml
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/doc/src/sgml/func.sgml,v
> retrieving revision 1.154
> diff -c -r1.154 func.sgml
> *** doc/src/sgml/func.sgml 5 May 2003 15:08:49 -0000 1.154
> --- doc/src/sgml/func.sgml 1 Jun 2003 15:41:15 -0000
> ***************
> *** 6962,6967 ****
> --- 6962,7209 ----
>
> </sect1>
>
> + <sect1 id="functions-array">
> + <title>Array Functions</title>
> +
> + <para>
> + <xref linkend="array-operators-table"> shows the operators
> + available for the <type>array</type> types.
> + </para>
> +
> + <table id="array-operators-table">
> + <title><type>array</type> Operators</title>
> + <tgroup cols="4">
> + <thead>
> + <row>
> + <entry>Operator</entry>
> + <entry>Description</entry>
> + <entry>Example</entry>
> + <entry>Result</entry>
> + </row>
> + </thead>
> + <tbody>
> + <row>
> + <entry> <literal>=</literal> </entry>
> + <entry>equals</entry>
> + <entry><literal>ARRAY[1.1,2.1,3.1]::int[] = ARRAY[1,2,3]</literal></entry>
> + <entry><literal>t</literal></entry>
> + </row>
> + <row>
> + <entry> <literal>||</literal> </entry>
> + <entry>array-to-array concatenation</entry>
> + <entry><literal>ARRAY[1,2,3] || ARRAY[4,5,6]</literal></entry>
> + <entry><literal>{{1,2,3},{4,5,6}}</literal></entry>
> + </row>
> + <row>
> + <entry> <literal>||</literal> </entry>
> + <entry>array-to-array concatenation</entry>
> + <entry><literal>ARRAY[1,2,3] || ARRAY[[4,5,6],[7,8,9]]</literal></entry>
> + <entry><literal>{{1,2,3},{4,5,6},{7,8,9}}</literal></entry>
> + </row>
> + <row>
> + <entry> <literal>||</literal> </entry>
> + <entry>element-to-array concatenation</entry>
> + <entry><literal>3 || ARRAY[4,5,6]</literal></entry>
> + <entry><literal>{3,4,5,6}</literal></entry>
> + </row>
> + <row>
> + <entry> <literal>||</literal> </entry>
> + <entry>array-to-element concatenation</entry>
> + <entry><literal>ARRAY[4,5,6] || 7</literal></entry>
> + <entry><literal>{4,5,6,7}</literal></entry>
> + </row>
> + </tbody>
> + </tgroup>
> + </table>
> +
> + <para>
> + <xref linkend="array-functions-table"> shows the functions
> + available for use with array types. See <xref linkend="arrays">
> + for more discussion and examples for the use of these functions.
> + </para>
> +
> + <table id="array-functions-table">
> + <title><type>array</type> Functions</title>
> + <tgroup cols="5">
> + <thead>
> + <row>
> + <entry>Function</entry>
> + <entry>Return Type</entry>
> + <entry>Description</entry>
> + <entry>Example</entry>
> + <entry>Result</entry>
> + </row>
> + </thead>
> + <tbody>
> + <row>
> + <entry>
> + <literal>
> + <function>array_append</function>
> + (<type>anyarray</type>, <type>anyelement</type>)
> + </literal>
> + </entry>
> + <entry><type>anyarray</type></entry>
> + <entry>
> + append an element to the end of an array, returning
> + <literal>NULL</literal> for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>array_append(ARRAY[1,2], 3)</literal></entry>
> + <entry><literal>{1,2,3}</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>array_assign</function>
> + (<type>anyarray</type>, <type>integer</type>, <type>anyelement</type>)
> + </literal>
> + </entry>
> + <entry><type>anyarray</type></entry>
> + <entry>
> + assign a value to a specific array element, returning
> + <literal>NULL</literal> for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>array_assign(ARRAY[1,2,3], 2, 99)</literal></entry>
> + <entry><literal>{1,99,3}</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>array_cat</function>
> + (<type>anyarray</type>, <type>anyarray</type>)
> + </literal>
> + </entry>
> + <entry><type>anyarray</type></entry>
> + <entry>
> + concatenate two arrays, returning <literal>NULL</literal>
> + for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>array_cat(ARRAY[1,2,3], ARRAY[4,5,6])</literal></entry>
> + <entry><literal>{{1,2,3},{4,5,6}}</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>array_dims</function>
> + (<type>anyarray</type>)
> + </literal>
> + </entry>
> + <entry><type>text</type></entry>
> + <entry>
> + returns a text representation of array dimension lower and upper bounds,
> + generating an ERROR for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>array_dims(array[[1,2,3],[4,5,6]])</literal></entry>
> + <entry><literal>[1:2][1:3]</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>array_lower</function>
> + (<type>anyarray</type>, <type>integer</type>)
> + </literal>
> + </entry>
> + <entry><type>integer</type></entry>
> + <entry>
> + returns lower bound of the requested array dimension, returning
> + <literal>NULL</literal> for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>array_lower(array_prepend(0, ARRAY[1,2,3]), 1)</literal></entry>
> + <entry><literal>0</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>array_prepend</function>
> + (<type>anyelement</type>, <type>anyarray</type>)
> + </literal>
> + </entry>
> + <entry><type>anyarray</type></entry>
> + <entry>
> + append an element to the beginning of an array, returning
> + <literal>NULL</literal> for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>array_prepend(1, ARRAY[2,3])</literal></entry>
> + <entry><literal>{1,2,3}</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>array_subscript</function>
> + (<type>anyarray</type>, <type>integer</type>)
> + </literal>
> + </entry>
> + <entry><type>anyelement</type></entry>
> + <entry>
> + returns requested array element, returning
> + <literal>NULL</literal> for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>array_subscript(ARRAY[1,2,3], 3)</literal></entry>
> + <entry><literal>3</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>array_to_str</function>
> + (<type>anyarray</type>, <type>text</type>)
> + </literal>
> + </entry>
> + <entry><type>text</type></entry>
> + <entry>
> + concatenates array elements using provided delimiter, returning
> + <literal>NULL</literal> for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>array_to_str(array[1.1,2.2,3.3]::numeric(4,2)[],'~^~')</literal></entry>
> + <entry><literal>1.10~^~2.20~^~3.30</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>array_upper</function>
> + (<type>anyarray</type>, <type>integer</type>)
> + </literal>
> + </entry>
> + <entry><type>integer</type></entry>
> + <entry>
> + returns upper bound of the requested array dimension, returning
> + <literal>NULL</literal> for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>array_upper(array_append(ARRAY[1,2,3], 4), 1)</literal></entry>
> + <entry><literal>4</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>singleton_array</function>
> + (<type>anyelement</type>)
> + </literal>
> + </entry>
> + <entry><type>anyarray</type></entry>
> + <entry>
> + create an array from the provided element, returning
> + <literal>NULL</literal> for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>singleton_array(1)</literal></entry>
> + <entry><literal>{1}</literal></entry>
> + </row>
> + <row>
> + <entry>
> + <literal>
> + <function>str_to_array</function>
> + (<type>text</type>, <type>text</type>)
> + </literal>
> + </entry>
> + <entry><type>text[]</type></entry>
> + <entry>
> + splits string into array elements using provided delimiter, returning
> + <literal>NULL</literal> for <literal>NULL</literal> inputs
> + </entry>
> + <entry><literal>str_to_array('1.10~^~2.20~^~3.30','~^~')::float8[]</literal></entry>
> + <entry><literal>{1.1,2.2,3.3}</literal></entry>
> + </row>
> + </tbody>
> + </tgroup>
> + </table>
> + </sect1>
>
> <sect1 id="functions-aggregate">
> <title>Aggregate Functions</title>
> Index: src/backend/catalog/pg_aggregate.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/catalog/pg_aggregate.c,v
> retrieving revision 1.56
> diff -c -r1.56 pg_aggregate.c
> *** src/backend/catalog/pg_aggregate.c 18 Sep 2002 21:35:20 -0000 1.56
> --- src/backend/catalog/pg_aggregate.c 1 Jun 2003 15:41:15 -0000
> ***************
> *** 50,59 ****
> Oid finalfn = InvalidOid; /* can be omitted */
> Oid finaltype;
> Oid fnArgs[FUNC_MAX_ARGS];
> ! int nargs;
> Oid procOid;
> TupleDesc tupDesc;
> int i;
> ObjectAddress myself,
> referenced;
>
> --- 50,65 ----
> Oid finalfn = InvalidOid; /* can be omitted */
> Oid finaltype;
> Oid fnArgs[FUNC_MAX_ARGS];
> ! int nargs_transfn;
> ! int nargs_finalfn;
> Oid procOid;
> TupleDesc tupDesc;
> int i;
> + Oid rettype;
> + Oid *true_oid_array_transfn;
> + Oid *true_oid_array_finalfn;
> + bool retset;
> + FuncDetailCode fdresult;
> ObjectAddress myself,
> referenced;
>
> ***************
> *** 68,91 ****
> MemSet(fnArgs, 0, FUNC_MAX_ARGS * sizeof(Oid));
> fnArgs[0] = aggTransType;
> if (aggBaseType == ANYOID)
> ! nargs = 1;
> else
> {
> fnArgs[1] = aggBaseType;
> ! nargs = 2;
> }
> ! transfn = LookupFuncName(aggtransfnName, nargs, fnArgs);
> if (!OidIsValid(transfn))
> ! func_error("AggregateCreate", aggtransfnName, nargs, fnArgs, NULL);
> tup = SearchSysCache(PROCOID,
> ObjectIdGetDatum(transfn),
> 0, 0, 0);
> if (!HeapTupleIsValid(tup))
> ! func_error("AggregateCreate", aggtransfnName, nargs, fnArgs, NULL);
> proc = (Form_pg_proc) GETSTRUCT(tup);
> - if (proc->prorettype != aggTransType)
> - elog(ERROR, "return type of transition function %s is not %s",
> - NameListToString(aggtransfnName), format_type_be(aggTransType));
>
> /*
> * If the transfn is strict and the initval is NULL, make sure input
> --- 74,122 ----
> MemSet(fnArgs, 0, FUNC_MAX_ARGS * sizeof(Oid));
> fnArgs[0] = aggTransType;
> if (aggBaseType == ANYOID)
> ! nargs_transfn = 1;
> else
> {
> fnArgs[1] = aggBaseType;
> ! nargs_transfn = 2;
> }
> !
> ! /*
> ! * func_get_detail looks up the function in the catalogs, does
> ! * disambiguation for polymorphic functions, handles inheritance, and
> ! * returns the funcid and type and set or singleton status of the
> ! * function's return value. it also returns the true argument types
> ! * to the function.
> ! */
> ! fdresult = func_get_detail(aggtransfnName, NIL, nargs_transfn, fnArgs,
> ! &transfn, &rettype, &retset,
> ! &true_oid_array_transfn);
> !
> ! /* only valid case is a normal function */
> ! if (fdresult != FUNCDETAIL_NORMAL)
> ! func_error("AggregateCreate", aggtransfnName, nargs_transfn, fnArgs, NULL);
> !
> if (!OidIsValid(transfn))
> ! func_error("AggregateCreate", aggtransfnName, nargs_transfn, fnArgs, NULL);
> !
> ! /*
> ! * enforce consistency with ANYARRAY and ANYELEMENT argument
> ! * and return types, possibly modifying return type along the way
> ! */
> ! rettype = enforce_generic_type_consistency(fnArgs, true_oid_array_transfn,
> ! nargs_transfn, rettype);
> !
> ! if (rettype != aggTransType)
> ! elog(ERROR, "return type of transition function %s is not %s",
> ! NameListToString(aggtransfnName), format_type_be(aggTransType));
> !
> tup = SearchSysCache(PROCOID,
> ObjectIdGetDatum(transfn),
> 0, 0, 0);
> if (!HeapTupleIsValid(tup))
> ! func_error("AggregateCreate", aggtransfnName,
> ! nargs_transfn, fnArgs, NULL);
> proc = (Form_pg_proc) GETSTRUCT(tup);
>
> /*
> * If the transfn is strict and the initval is NULL, make sure input
> ***************
> *** 105,121 ****
> {
> MemSet(fnArgs, 0, FUNC_MAX_ARGS * sizeof(Oid));
> fnArgs[0] = aggTransType;
> ! finalfn = LookupFuncName(aggfinalfnName, 1, fnArgs);
> ! if (!OidIsValid(finalfn))
> func_error("AggregateCreate", aggfinalfnName, 1, fnArgs, NULL);
> ! tup = SearchSysCache(PROCOID,
> ! ObjectIdGetDatum(finalfn),
> ! 0, 0, 0);
> ! if (!HeapTupleIsValid(tup))
> func_error("AggregateCreate", aggfinalfnName, 1, fnArgs, NULL);
> ! proc = (Form_pg_proc) GETSTRUCT(tup);
> ! finaltype = proc->prorettype;
> ! ReleaseSysCache(tup);
> }
> else
> {
> --- 136,161 ----
> {
> MemSet(fnArgs, 0, FUNC_MAX_ARGS * sizeof(Oid));
> fnArgs[0] = aggTransType;
> ! nargs_finalfn = 1;
> !
> ! fdresult = func_get_detail(aggfinalfnName, NIL, 1, fnArgs,
> ! &finalfn, &rettype, &retset,
> ! &true_oid_array_finalfn);
> !
> ! /* only valid case is a normal function */
> ! if (fdresult != FUNCDETAIL_NORMAL)
> func_error("AggregateCreate", aggfinalfnName, 1, fnArgs, NULL);
> !
> ! if (!OidIsValid(finalfn))
> func_error("AggregateCreate", aggfinalfnName, 1, fnArgs, NULL);
> !
> ! /*
> ! * enforce consistency with ANYARRAY and ANYELEMENT argument
> ! * and return types, possibly modifying return type along the way
> ! */
> ! finaltype = enforce_generic_type_consistency(fnArgs,
> ! true_oid_array_finalfn,
> ! nargs_finalfn, rettype);
> }
> else
> {
> ***************
> *** 125,130 ****
> --- 165,191 ----
> finaltype = aggTransType;
> }
> Assert(OidIsValid(finaltype));
> +
> + /*
> + * special disallowed cases:
> + * 1) if finaltype is polymorphic, basetype cannot be ANY
> + * 2) if finaltype is polymorphic, both args to transfn must be
> + * polymorphic
> + */
> + if (finaltype == ANYARRAYOID || finaltype == ANYELEMENTOID)
> + {
> + if (aggBaseType == ANYOID)
> + elog(ERROR, "aggregate with base type ANY must have a " \
> + "non-polymorphic return type");
> +
> + if (nargs_transfn > 1 && (
> + (true_oid_array_transfn[0] != ANYARRAYOID &&
> + true_oid_array_transfn[0] != ANYELEMENTOID) ||
> + (true_oid_array_transfn[1] != ANYARRAYOID &&
> + true_oid_array_transfn[1] != ANYELEMENTOID)))
> + elog(ERROR, "aggregate with polymorphic return type requires " \
> + "state function with both arguments polymorphic");
> + }
>
> /*
> * Everything looks okay. Try to create the pg_proc entry for the
> Index: src/backend/commands/aggregatecmds.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/commands/aggregatecmds.c,v
> retrieving revision 1.5
> diff -c -r1.5 aggregatecmds.c
> *** src/backend/commands/aggregatecmds.c 4 Sep 2002 20:31:14 -0000 1.5
> --- src/backend/commands/aggregatecmds.c 1 Jun 2003 15:41:15 -0000
> ***************
> *** 119,125 ****
> baseTypeId = typenameTypeId(baseType);
>
> transTypeId = typenameTypeId(transType);
> ! if (get_typtype(transTypeId) == 'p')
> elog(ERROR, "Aggregate transition datatype cannot be %s",
> format_type_be(transTypeId));
>
> --- 119,127 ----
> baseTypeId = typenameTypeId(baseType);
>
> transTypeId = typenameTypeId(transType);
> ! if (get_typtype(transTypeId) == 'p' &&
> ! transTypeId != ANYARRAYOID &&
> ! transTypeId != ANYELEMENTOID)
> elog(ERROR, "Aggregate transition datatype cannot be %s",
> format_type_be(transTypeId));
>
> Index: src/backend/commands/analyze.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/commands/analyze.c,v
> retrieving revision 1.54
> diff -c -r1.54 analyze.c
> *** src/backend/commands/analyze.c 27 May 2003 17:49:45 -0000 1.54
> --- src/backend/commands/analyze.c 2 Jun 2003 01:06:40 -0000
> ***************
> *** 392,397 ****
> --- 392,398 ----
> Oid eqopr = InvalidOid;
> Oid eqfunc = InvalidOid;
> Oid ltopr = InvalidOid;
> + Oid elem_type = get_element_type(attr->atttypid);
> VacAttrStats *stats;
>
> /* Don't analyze dropped columns */
> ***************
> *** 413,418 ****
> --- 414,433 ----
> eqfunc = oprfuncid(func_operator);
> }
> ReleaseSysCache(func_operator);
> +
> + /*
> + * If we have an array type, we also need a suitable
> + * element equality operator. Check to be sure it exists.
> + */
> + if (elem_type != InvalidOid)
> + {
> + Operator elem_func_operator;
> +
> + elem_func_operator = equality_oper(elem_type, true);
> + if (elem_func_operator != NULL)
> + eqfunc = InvalidOid;
> + ReleaseSysCache(elem_func_operator);
> + }
> }
> if (!OidIsValid(eqfunc))
> return NULL;
> ***************
> *** 448,453 ****
> --- 463,482 ----
> if (oprrest == F_SCALARLTSEL)
> ltopr = oprid(func_operator);
> ReleaseSysCache(func_operator);
> +
> + /*
> + * If we have an array type, we also need a suitable
> + * element ordering operator. Check to be sure it exists.
> + */
> + if (elem_type != InvalidOid)
> + {
> + Operator elem_func_operator;
> +
> + elem_func_operator = ordering_oper(elem_type, true);
> + if (elem_func_operator != NULL)
> + ltopr = InvalidOid;
> + ReleaseSysCache(elem_func_operator);
> + }
> }
> stats->ltopr = ltopr;
>
> Index: src/backend/executor/nodeAgg.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/executor/nodeAgg.c,v
> retrieving revision 1.105
> diff -c -r1.105 nodeAgg.c
> *** src/backend/executor/nodeAgg.c 30 May 2003 20:23:10 -0000 1.105
> --- src/backend/executor/nodeAgg.c 1 Jun 2003 15:41:15 -0000
> ***************
> *** 58,63 ****
> --- 58,64 ----
> #include "executor/executor.h"
> #include "executor/nodeAgg.h"
> #include "miscadmin.h"
> + #include "nodes/makefuncs.h"
> #include "optimizer/clauses.h"
> #include "parser/parse_coerce.h"
> #include "parser/parse_expr.h"
> ***************
> *** 212,218 ****
> static void agg_fill_hash_table(AggState *aggstate);
> static TupleTableSlot *agg_retrieve_hash_table(AggState *aggstate);
> static Datum GetAggInitVal(Datum textInitVal, Oid transtype);
> !
>
> /*
> * Initialize all aggregates for a new group of input values.
> --- 213,219 ----
> static void agg_fill_hash_table(AggState *aggstate);
> static TupleTableSlot *agg_retrieve_hash_table(AggState *aggstate);
> static Datum GetAggInitVal(Datum textInitVal, Oid transtype);
> ! static Oid resolve_type(Oid type_to_resolve, Oid context_type);
>
> /*
> * Initialize all aggregates for a new group of input values.
> ***************
> *** 351,364 ****
> fcinfo.context = NULL;
> fcinfo.resultinfo = NULL;
> fcinfo.isnull = false;
> -
> fcinfo.flinfo = &peraggstate->transfn;
> fcinfo.nargs = 2;
> fcinfo.arg[0] = pergroupstate->transValue;
> fcinfo.argnull[0] = pergroupstate->transValueIsNull;
> fcinfo.arg[1] = newVal;
> fcinfo.argnull[1] = isNull;
> -
> newVal = FunctionCallInvoke(&fcinfo);
>
> /*
> --- 352,363 ----
> ***************
> *** 1176,1182 ****
> --- 1175,1195 ----
> AclResult aclresult;
> Oid transfn_oid,
> finalfn_oid;
> + FuncExpr *transfnexpr,
> + *finalfnexpr;
> Datum textInitVal;
> + List *fargs;
> + Oid agg_rt_type;
> + Oid *transfn_arg_types;
> + List *transfn_args = NIL;
> + int transfn_nargs;
> + Oid transfn_ret_type;
> + Oid *finalfn_arg_types = NULL;
> + List *finalfn_args = NIL;
> + Oid finalfn_ret_type = InvalidOid;
> + int finalfn_nargs = 0;
> + Node *arg0;
> + Node *arg1;
> int i;
>
> /* Look for a previous duplicate aggregate */
> ***************
> *** 1224,1229 ****
> --- 1237,1402 ----
> &peraggstate->transtypeLen,
> &peraggstate->transtypeByVal);
>
> + peraggstate->transfn_oid = transfn_oid = aggform->aggtransfn;
> + peraggstate->finalfn_oid = finalfn_oid = aggform->aggfinalfn;
> +
> + /* get the runtime aggregate argument type */
> + fargs = aggref->args;
> + agg_rt_type = exprType((Node *) nth(0, fargs));
> +
> + /* get the transition function argument and return types */
> + transfn_ret_type = get_func_rettype(transfn_oid);
> + transfn_arg_types = get_func_argtypes(transfn_oid, &transfn_nargs);
> +
> + /* resolve any polymorphic types */
> + if (transfn_nargs == 2)
> + /* base type was not ANY */
> + {
> + if (transfn_arg_types[1] == ANYARRAYOID ||
> + transfn_arg_types[1] == ANYELEMENTOID)
> + transfn_arg_types[1] = agg_rt_type;
> +
> + transfn_arg_types[0] = resolve_type(transfn_arg_types[0],
> + agg_rt_type);
> +
> + /*
> + * Build arg list to use on the transfn FuncExpr node. We really
> + * only care that the node type is correct so that the transfn
> + * can discover the actual argument types at runtime using
> + * get_fn_expr_argtype()
> + */
> + arg0 = (Node *) makeRelabelType((Expr *) NULL, transfn_arg_types[0],
> + -1, COERCE_DONTCARE);
> + arg1 = (Node *) makeRelabelType((Expr *) NULL, transfn_arg_types[1],
> + -1, COERCE_DONTCARE);
> + transfn_args = makeList2(arg0, arg1);
> +
> + /*
> + * the state transition function always returns the same type
> + * as its first argument
> + */
> + if (transfn_ret_type == ANYARRAYOID ||
> + transfn_ret_type == ANYELEMENTOID)
> + transfn_ret_type = transfn_arg_types[0];
> + }
> + else if (transfn_nargs == 1)
> + /*
> + * base type was ANY, therefore the aggregate return type should
> + * be non-polymorphic
> + */
> + {
> + Oid finaltype = get_func_rettype(aggref->aggfnoid);
> +
> + /*
> + * this should have been prevented in AggregateCreate,
> + * but check anyway
> + */
> + if (finaltype == ANYARRAYOID || finaltype == ANYELEMENTOID)
> + elog(ERROR, "aggregate with base type ANY must have a " \
> + "non-polymorphic return type");
> +
> + /* see if we have a final function */
> + if (OidIsValid(finalfn_oid))
> + {
> + finalfn_arg_types = get_func_argtypes(finalfn_oid, &finalfn_nargs);
> + if (finalfn_nargs != 1)
> + elog(ERROR, "final function takes unexpected number " \
> + "of arguments: %d", finalfn_nargs);
> +
> + /*
> + * final function argument is always the same as the state
> + * function return type
> + */
> + if (finalfn_arg_types[0] != ANYARRAYOID &&
> + finalfn_arg_types[0] != ANYELEMENTOID)
> + {
> + /* if it is not ambiguous, use it */
> + transfn_ret_type = finalfn_arg_types[0];
> + }
> + else
> + {
> + /* if it is ambiguous, try to derive it */
> + finalfn_ret_type = finaltype;
> + finalfn_arg_types[0] = resolve_type(finalfn_arg_types[0],
> + finalfn_ret_type);
> + transfn_ret_type = finalfn_arg_types[0];
> + }
> + }
> + else
> + transfn_ret_type = finaltype;
> +
> + transfn_arg_types[0] = resolve_type(transfn_arg_types[0],
> + transfn_ret_type);
> +
> + /*
> + * Build arg list to use on the transfn FuncExpr node. We really
> + * only care that the node type is correct so that the transfn
> + * can discover the actual argument types at runtime using
> + * get_fn_expr_argtype()
> + */
> + arg0 = (Node *) makeRelabelType((Expr *) NULL, transfn_arg_types[0],
> + -1, COERCE_DONTCARE);
> + transfn_args = makeList1(arg0);
> + }
> + else
> + elog(ERROR, "state transition function takes unexpected number " \
> + "of arguments: %d", transfn_nargs);
> +
> + if (OidIsValid(finalfn_oid))
> + {
> + /* get the final function argument and return types */
> + if (finalfn_ret_type == InvalidOid)
> + finalfn_ret_type = get_func_rettype(finalfn_oid);
> +
> + if (!finalfn_arg_types)
> + {
> + finalfn_arg_types = get_func_argtypes(finalfn_oid, &finalfn_nargs);
> + if (finalfn_nargs != 1)
> + elog(ERROR, "final function takes unexpected number " \
> + "of arguments: %d", finalfn_nargs);
> + }
> +
> + /*
> + * final function argument is always the same as the state
> + * function return type, which by now should have been resolved
> + */
> + if (finalfn_arg_types[0] == ANYARRAYOID ||
> + finalfn_arg_types[0] == ANYELEMENTOID)
> + finalfn_arg_types[0] = transfn_ret_type;
> +
> + /*
> + * Build arg list to use on the finalfn FuncExpr node. We really
> + * only care that the node type is correct so that the finalfn
> + * can discover the actual argument type at runtime using
> + * get_fn_expr_argtype()
> + */
> + arg0 = (Node *) makeRelabelType((Expr *) NULL, finalfn_arg_types[0],
> + -1, COERCE_DONTCARE);
> + finalfn_args = makeList1(arg0);
> +
> + finalfn_ret_type = resolve_type(finalfn_ret_type,
> + finalfn_arg_types[0]);
> + }
> +
> + fmgr_info(transfn_oid, &peraggstate->transfn);
> + transfnexpr = (FuncExpr *) make_funcclause(transfn_oid,
> + transfn_ret_type,
> + false, /* cannot be a set */
> + COERCE_DONTCARE, /* to match any user expr */
> + transfn_args);
> + peraggstate->transfn.fn_expr = (Node *) transfnexpr;
> +
> + if (OidIsValid(finalfn_oid))
> + {
> + fmgr_info(finalfn_oid, &peraggstate->finalfn);
> + finalfnexpr = (FuncExpr *) make_funcclause(finalfn_oid,
> + finalfn_ret_type,
> + false, /* cannot be a set */
> + COERCE_DONTCARE, /* to match any user expr */
> + finalfn_args);
> + peraggstate->finalfn.fn_expr = (Node *) finalfnexpr;
> + }
> +
> /*
> * initval is potentially null, so don't try to access it as a
> * struct field. Must do it the hard way with SysCacheGetAttr.
> ***************
> *** 1236,1249 ****
> peraggstate->initValue = (Datum) 0;
> else
> peraggstate->initValue = GetAggInitVal(textInitVal,
> ! aggform->aggtranstype);
> !
> ! peraggstate->transfn_oid = transfn_oid = aggform->aggtransfn;
> ! peraggstate->finalfn_oid = finalfn_oid = aggform->aggfinalfn;
> !
> ! fmgr_info(transfn_oid, &peraggstate->transfn);
> ! if (OidIsValid(finalfn_oid))
> ! fmgr_info(finalfn_oid, &peraggstate->finalfn);
>
> /*
> * If the transfn is strict and the initval is NULL, make sure
> --- 1409,1415 ----
> peraggstate->initValue = (Datum) 0;
> else
> peraggstate->initValue = GetAggInitVal(textInitVal,
> ! transfn_arg_types[0]);
>
> /*
> * If the transfn is strict and the initval is NULL, make sure
> ***************
> *** 1454,1457 ****
> --- 1620,1656 ----
> elog(ERROR, "Aggregate function %u called as normal function",
> fcinfo->flinfo->fn_oid);
> return (Datum) 0; /* keep compiler quiet */
> + }
> +
> + static Oid
> + resolve_type(Oid type_to_resolve, Oid context_type)
> + {
> + Oid resolved_type;
> +
> + if (context_type == ANYARRAYOID || context_type == ANYELEMENTOID)
> + resolved_type = type_to_resolve;
> + else if (type_to_resolve == ANYARRAYOID)
> + /* any array */
> + {
> + Oid context_type_arraytype = get_array_type(context_type);
> +
> + if (context_type_arraytype != InvalidOid)
> + resolved_type = context_type_arraytype;
> + else
> + resolved_type = context_type;
> + }
> + else if (type_to_resolve == ANYELEMENTOID)
> + /* any element */
> + {
> + Oid context_type_elemtype = get_element_type(context_type);
> +
> + if (context_type_elemtype != InvalidOid)
> + resolved_type = context_type_elemtype;
> + else
> + resolved_type = context_type;
> + }
> + else
> + resolved_type = type_to_resolve;
> +
> + return resolved_type;
> }
> Index: src/backend/executor/nodeSubplan.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/executor/nodeSubplan.c,v
> retrieving revision 1.45
> diff -c -r1.45 nodeSubplan.c
> *** src/backend/executor/nodeSubplan.c 8 Apr 2003 23:20:01 -0000 1.45
> --- src/backend/executor/nodeSubplan.c 1 Jun 2003 15:41:47 -0000
> ***************
> *** 28,50 ****
> #include "utils/datum.h"
> #include "utils/lsyscache.h"
>
> -
> - typedef struct ArrayBuildState
> - {
> - MemoryContext mcontext; /* where all the temp stuff is kept */
> - Datum *dvalues; /* array of accumulated Datums */
> - /*
> - * The allocated size of dvalues[] is always a multiple of
> - * ARRAY_ELEMS_CHUNKSIZE
> - */
> - #define ARRAY_ELEMS_CHUNKSIZE 64
> - int nelems; /* number of valid Datums in dvalues[] */
> - Oid element_type; /* data type of the Datums */
> - int16 typlen; /* needed info about datatype */
> - bool typbyval;
> - char typalign;
> - } ArrayBuildState;
> -
> static Datum ExecHashSubPlan(SubPlanState *node,
> ExprContext *econtext,
> bool *isNull);
> --- 28,33 ----
> ***************
> *** 54,66 ****
> static void buildSubPlanHash(SubPlanState *node);
> static bool findPartialMatch(TupleHashTable hashtable, TupleTableSlot *slot);
> static bool tupleAllNulls(HeapTuple tuple);
> - static ArrayBuildState *accumArrayResult(ArrayBuildState *astate,
> - Datum dvalue, bool disnull,
> - Oid element_type,
> - MemoryContext rcontext);
> - static Datum makeArrayResult(ArrayBuildState *astate,
> - MemoryContext rcontext);
> -
>
> /* ----------------------------------------------------------------
> * ExecSubPlan
> --- 37,42 ----
> ***************
> *** 224,229 ****
> --- 200,206 ----
> PlanState *planstate = node->planstate;
> SubLinkType subLinkType = subplan->subLinkType;
> bool useOr = subplan->useOr;
> + bool isExpr = subplan->isExpr;
> MemoryContext oldcontext;
> TupleTableSlot *slot;
> Datum result;
> ***************
> *** 292,297 ****
> --- 269,279 ----
> bool rownull = false;
> int col = 1;
> List *plst;
> + int numelems;
> + int elemnum;
> + Datum dvalue;
> + Datum *dvalues = NULL;
> + bool disnull;
>
> if (subLinkType == EXISTS_SUBLINK)
> {
> ***************
> *** 329,337 ****
>
> if (subLinkType == ARRAY_SUBLINK)
> {
> - Datum dvalue;
> - bool disnull;
> -
> found = true;
> /* stash away current value */
> dvalue = heap_getattr(tup, 1, tdesc, &disnull);
> --- 311,316 ----
> ***************
> *** 349,446 ****
> found = true;
>
> /*
> ! * For ALL, ANY, and MULTIEXPR sublinks, iterate over combining
> ! * operators for columns of tuple.
> */
> ! plst = subplan->paramIds;
> ! foreach(lst, node->exprs)
> {
> ! ExprState *exprstate = (ExprState *) lfirst(lst);
> ! int paramid = lfirsti(plst);
> ! ParamExecData *prmdata;
> ! Datum expresult;
> ! bool expnull;
> !
> ! /*
> ! * Load up the Param representing this column of the sub-select.
> ! */
> ! prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
> ! Assert(prmdata->execPlan == NULL);
> ! prmdata->value = heap_getattr(tup, col, tdesc,
> ! &(prmdata->isnull));
>
> ! /*
> ! * Now we can eval the combining operator for this column.
> ! */
> ! expresult = ExecEvalExprSwitchContext(exprstate, econtext,
> ! &expnull, NULL);
> !
> ! /*
> ! * Combine the result into the row result as appropriate.
> ! */
> ! if (col == 1)
> {
> ! rowresult = expresult;
> ! rownull = expnull;
> }
> ! else if (useOr)
> {
> ! /* combine within row per OR semantics */
> ! if (expnull)
> ! rownull = true;
> ! else if (DatumGetBool(expresult))
> {
> ! rowresult = BoolGetDatum(true);
> ! rownull = false;
> ! break; /* needn't look at any more columns */
> }
> }
> else
> {
> ! /* combine within row per AND semantics */
> ! if (expnull)
> ! rownull = true;
> ! else if (!DatumGetBool(expresult))
> ! {
> ! rowresult = BoolGetDatum(false);
> ! rownull = false;
> ! break; /* needn't look at any more columns */
> ! }
> }
>
> - plst = lnext(plst);
> - col++;
> }
>
> ! if (subLinkType == ANY_SUBLINK)
> {
> ! /* combine across rows per OR semantics */
> ! if (rownull)
> ! *isNull = true;
> ! else if (DatumGetBool(rowresult))
> {
> ! result = BoolGetDatum(true);
> ! *isNull = false;
> ! break; /* needn't look at any more rows */
> }
> ! }
> ! else if (subLinkType == ALL_SUBLINK)
> ! {
> ! /* combine across rows per AND semantics */
> ! if (rownull)
> ! *isNull = true;
> ! else if (!DatumGetBool(rowresult))
> {
> ! result = BoolGetDatum(false);
> ! *isNull = false;
> ! break; /* needn't look at any more rows */
> }
> - }
> - else
> - {
> - /* must be MULTIEXPR_SUBLINK */
> - result = rowresult;
> - *isNull = rownull;
> }
> }
>
> --- 328,490 ----
> found = true;
>
> /*
> ! * When isExpr is true, we have either a scalar expression or an
> ! * array. In the former case, this is no different than the !isExpr
> ! * case. In the latter case, iterate over the elements as if they
> ! * were from multiple input tuples.
> */
> ! if (!isExpr)
> ! numelems = 1;
> ! else
> {
> ! Oid expr_typeid = tdesc->attrs[0]->atttypid;
>
> ! if (expr_typeid != subplan->exprtype)
> {
> ! subplan->exprtype = expr_typeid;
> ! subplan->elemtype = get_element_type(expr_typeid);
> !
> ! if (subplan->elemtype != InvalidOid)
> ! get_typlenbyvalalign(subplan->elemtype,
> ! &subplan->elmlen,
> ! &subplan->elmbyval,
> ! &subplan->elmalign);
> }
> !
> ! /* get current value */
> ! dvalue = heap_getattr(tup, 1, tdesc, &disnull);
> !
> ! /* XXX this will need work if/when arrays support NULL elements */
> ! if (!disnull)
> {
> ! if (subplan->elemtype != InvalidOid)
> ! {
> ! ArrayType *v = DatumGetArrayTypeP(dvalue);
> !
> ! deconstruct_array(v, subplan->elemtype, subplan->elmlen,
> ! subplan->elmbyval, subplan->elmalign,
> ! &dvalues, &numelems);
> ! }
> ! else
> {
> ! numelems = 1;
> ! dvalues = (Datum *) palloc(numelems * sizeof(Datum));
> ! dvalues[0] = dvalue;
> }
> }
> else
> {
> ! numelems = 1;
> ! dvalues = (Datum *) palloc(numelems * sizeof(Datum));
> ! dvalues[0] = (Datum) 0;
> }
>
> }
>
> ! for (elemnum = 0; elemnum < numelems; elemnum++)
> {
> ! /*
> ! * For ALL, ANY, and MULTIEXPR sublinks, iterate over combining
> ! * operators for columns of tuple.
> ! */
> ! col = 1;
> ! plst = subplan->paramIds;
> ! foreach(lst, node->exprs)
> {
> ! ExprState *exprstate = (ExprState *) lfirst(lst);
> ! int paramid = lfirsti(plst);
> ! ParamExecData *prmdata;
> ! Datum expresult;
> ! bool expnull;
> !
> ! /*
> ! * Load up the Param representing this column of the sub-select.
> ! */
> ! prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
> ! Assert(prmdata->execPlan == NULL);
> !
> ! if (!isExpr)
> ! prmdata->value = heap_getattr(tup, col, tdesc,
> ! &(prmdata->isnull));
> ! else
> ! {
> ! prmdata->value = dvalues[elemnum];
> ! prmdata->isnull = disnull;
> ! }
> !
> ! /*
> ! * Now we can eval the combining operator for this column.
> ! */
> ! expresult = ExecEvalExprSwitchContext(exprstate, econtext,
> ! &expnull, NULL);
> !
> ! /*
> ! * Combine the result into the row result as appropriate.
> ! */
> ! if (col == 1)
> ! {
> ! rowresult = expresult;
> ! rownull = expnull;
> ! }
> ! else if (useOr)
> ! {
> ! /* combine within row per OR semantics */
> ! if (expnull)
> ! rownull = true;
> ! else if (DatumGetBool(expresult))
> ! {
> ! rowresult = BoolGetDatum(true);
> ! rownull = false;
> ! break; /* needn't look at any more columns */
> ! }
> ! }
> ! else
> ! {
> ! /* combine within row per AND semantics */
> ! if (expnull)
> ! rownull = true;
> ! else if (!DatumGetBool(expresult))
> ! {
> ! rowresult = BoolGetDatum(false);
> ! rownull = false;
> ! break; /* needn't look at any more columns */
> ! }
> ! }
> !
> ! plst = lnext(plst);
> ! col++;
> }
> !
> ! if (subLinkType == ANY_SUBLINK)
> {
> ! /* combine across rows per OR semantics */
> ! if (rownull)
> ! *isNull = true;
> ! else if (DatumGetBool(rowresult))
> ! {
> ! result = BoolGetDatum(true);
> ! *isNull = false;
> ! break; /* needn't look at any more rows */
> ! }
> ! }
> ! else if (subLinkType == ALL_SUBLINK)
> ! {
> ! /* combine across rows per AND semantics */
> ! if (rownull)
> ! *isNull = true;
> ! else if (!DatumGetBool(rowresult))
> ! {
> ! result = BoolGetDatum(false);
> ! *isNull = false;
> ! break; /* needn't look at any more rows */
> ! }
> ! }
> ! else
> ! {
> ! /* must be MULTIEXPR_SUBLINK */
> ! result = rowresult;
> ! *isNull = rownull;
> }
> }
> }
>
> ***************
> *** 478,483 ****
> --- 522,528 ----
> buildSubPlanHash(SubPlanState *node)
> {
> SubPlan *subplan = (SubPlan *) node->xprstate.expr;
> + bool isExpr = subplan->isExpr;
> PlanState *planstate = node->planstate;
> int ncols = length(node->exprs);
> ExprContext *innerecontext = node->innerecontext;
> ***************
> *** 485,490 ****
> --- 530,536 ----
> MemoryContext oldcontext;
> int nbuckets;
> TupleTableSlot *slot;
> + TupleTableSlot *arrslot = NULL;
>
> Assert(subplan->subLinkType == ANY_SUBLINK);
> Assert(!subplan->useOr);
> ***************
> *** 562,604 ****
> {
> HeapTuple tup = slot->val;
> TupleDesc tdesc = slot->ttc_tupleDescriptor;
> ! int col = 1;
> List *plst;
> bool isnew;
>
> /*
> ! * Load up the Params representing the raw sub-select outputs,
> ! * then form the projection tuple to store in the hashtable.
> */
> ! foreach(plst, subplan->paramIds)
> {
> ! int paramid = lfirsti(plst);
> ! ParamExecData *prmdata;
>
> ! prmdata = &(innerecontext->ecxt_param_exec_vals[paramid]);
> ! Assert(prmdata->execPlan == NULL);
> ! prmdata->value = heap_getattr(tup, col, tdesc,
> ! &(prmdata->isnull));
> ! col++;
> ! }
> ! slot = ExecProject(node->projRight, NULL);
> ! tup = slot->val;
>
> - /*
> - * If result contains any nulls, store separately or not at all.
> - * (Since we know the projection tuple has no junk columns, we
> - * can just look at the overall hasnull info bit, instead of
> - * groveling through the columns.)
> - */
> - if (HeapTupleNoNulls(tup))
> - {
> - (void) LookupTupleHashEntry(node->hashtable, slot, &isnew);
> - node->havehashrows = true;
> }
> ! else if (node->hashnulls)
> {
> ! (void) LookupTupleHashEntry(node->hashnulls, slot, &isnew);
> ! node->havenullrows = true;
> }
>
> /*
> --- 608,746 ----
> {
> HeapTuple tup = slot->val;
> TupleDesc tdesc = slot->ttc_tupleDescriptor;
> ! TupleDesc arrtdesc = NULL;
> List *plst;
> bool isnew;
> + int numelems;
> + int elemnum;
> + Datum dvalue;
> + Datum *dvalues = NULL;
> + bool disnull;
>
> /*
> ! * When isExpr is true, we have either a scalar expression or an
> ! * array. In the former case, this is no different than the !isExpr
> ! * case. In the latter case, iterate over the elements as if they
> ! * were from multiple input tuples.
> */
> ! if (!isExpr)
> ! numelems = 1;
> ! else
> {
> ! Oid expr_typeid = tdesc->attrs[0]->atttypid;
>
> ! if (expr_typeid != subplan->exprtype)
> ! {
> ! subplan->exprtype = expr_typeid;
> ! subplan->elemtype = get_element_type(expr_typeid);
> !
> ! if (subplan->elemtype != InvalidOid)
> ! get_typlenbyvalalign(subplan->elemtype,
> ! &subplan->elmlen,
> ! &subplan->elmbyval,
> ! &subplan->elmalign);
> ! }
> !
> ! /* get current value */
> ! dvalue = heap_getattr(tup, 1, tdesc, &disnull);
> !
> ! if (subplan->elemtype != InvalidOid)
> ! {
> ! TupleTable tupleTable;
> ! ArrayType *v = DatumGetArrayTypeP(dvalue);
> !
> ! arrtdesc = CreateTemplateTupleDesc(1, false);
> ! TupleDescInitEntry(arrtdesc, 1, "elem", subplan->elemtype,
> ! -1, 0, false);
> !
> ! tupleTable = ExecCreateTupleTable(1);
> ! arrslot = ExecAllocTableSlot(tupleTable);
> ! ExecSetSlotDescriptor(arrslot, arrtdesc, true);
> !
> ! /* XXX this will need work if/when arrays support NULL elements */
> ! if (!disnull)
> ! {
> ! deconstruct_array(v, subplan->elemtype, subplan->elmlen,
> ! subplan->elmbyval, subplan->elmalign,
> ! &dvalues, &numelems);
> ! }
> ! else
> ! {
> ! numelems = 1;
> ! dvalues = (Datum *) palloc(numelems * sizeof(Datum));
> ! dvalues[0] = (Datum) 0;
> ! }
> ! }
> ! else
> ! {
> ! numelems = 1;
> ! dvalues = (Datum *) palloc(numelems * sizeof(Datum));
> ! dvalues[0] = dvalue;
> ! }
>
> }
> !
> ! for (elemnum = 0; elemnum < numelems; elemnum++)
> {
> ! int col = 1;
> !
> ! if (!isExpr || subplan->elemtype == InvalidOid)
> ! {
> ! /*
> ! * Load up the Params representing the raw sub-select outputs,
> ! * then form the projection tuple to store in the hashtable.
> ! */
> ! foreach(plst, subplan->paramIds)
> ! {
> ! int paramid = lfirsti(plst);
> ! ParamExecData *prmdata;
> !
> ! prmdata = &(innerecontext->ecxt_param_exec_vals[paramid]);
> ! Assert(prmdata->execPlan == NULL);
> !
> ! prmdata->value = heap_getattr(tup, col, tdesc,
> ! &(prmdata->isnull));
> !
> ! col++;
> ! }
> ! slot = ExecProject(node->projRight, NULL);
> ! tup = slot->val;
> ! }
> ! else
> ! {
> ! /*
> ! * For array type expressions, we need to build up our own
> ! * tuple and slot
> ! */
> ! char nullflag;
> !
> ! nullflag = disnull ? 'n' : ' ';
> ! tup = heap_formtuple(arrtdesc, &dvalues[elemnum], &nullflag);
> ! arrslot = ExecStoreTuple(tup, arrslot, InvalidBuffer, true);
> ! }
> !
> ! /*
> ! * If result contains any nulls, store separately or not at all.
> ! * (Since we know the projection tuple has no junk columns, we
> ! * can just look at the overall hasnull info bit, instead of
> ! * groveling through the columns.)
> ! */
> ! if (HeapTupleNoNulls(tup))
> ! {
> ! if (!isExpr)
> ! (void) LookupTupleHashEntry(node->hashtable, slot, &isnew);
> ! else
> ! (void) LookupTupleHashEntry(node->hashtable, arrslot, &isnew);
> ! node->havehashrows = true;
> ! }
> ! else if (node->hashnulls)
> ! {
> ! if (!isExpr)
> ! (void) LookupTupleHashEntry(node->hashnulls, slot, &isnew);
> ! else
> ! (void) LookupTupleHashEntry(node->hashnulls, arrslot, &isnew);
> ! node->havenullrows = true;
> ! }
> }
>
> /*
> ***************
> *** 615,620 ****
> --- 757,764 ----
> * have the potential for a double free attempt.
> */
> ExecClearTuple(node->projRight->pi_slot);
> + if (arrslot)
> + ExecClearTuple(arrslot);
>
> MemoryContextSwitchTo(oldcontext);
> }
> ***************
> *** 1084,1185 ****
> prm->execPlan = node;
> parent->chgParam = bms_add_member(parent->chgParam, paramid);
> }
> - }
> -
> - /*
> - * accumArrayResult - accumulate one (more) Datum for an ARRAY_SUBLINK
> - *
> - * astate is working state (NULL on first call)
> - * rcontext is where to keep working state
> - */
> - static ArrayBuildState *
> - accumArrayResult(ArrayBuildState *astate,
> - Datum dvalue, bool disnull,
> - Oid element_type,
> - MemoryContext rcontext)
> - {
> - MemoryContext arr_context,
> - oldcontext;
> -
> - if (astate == NULL)
> - {
> - /* First time through --- initialize */
> -
> - /* Make a temporary context to hold all the junk */
> - arr_context = AllocSetContextCreate(rcontext,
> - "ARRAY_SUBLINK Result",
> - ALLOCSET_DEFAULT_MINSIZE,
> - ALLOCSET_DEFAULT_INITSIZE,
> - ALLOCSET_DEFAULT_MAXSIZE);
> - oldcontext = MemoryContextSwitchTo(arr_context);
> - astate = (ArrayBuildState *) palloc(sizeof(ArrayBuildState));
> - astate->mcontext = arr_context;
> - astate->dvalues = (Datum *)
> - palloc(ARRAY_ELEMS_CHUNKSIZE * sizeof(Datum));
> - astate->nelems = 0;
> - astate->element_type = element_type;
> - get_typlenbyvalalign(element_type,
> - &astate->typlen,
> - &astate->typbyval,
> - &astate->typalign);
> - }
> - else
> - {
> - oldcontext = MemoryContextSwitchTo(astate->mcontext);
> - Assert(astate->element_type == element_type);
> - /* enlarge dvalues[] if needed */
> - if ((astate->nelems % ARRAY_ELEMS_CHUNKSIZE) == 0)
> - astate->dvalues = (Datum *)
> - repalloc(astate->dvalues,
> - (astate->nelems + ARRAY_ELEMS_CHUNKSIZE) * sizeof(Datum));
> - }
> -
> - if (disnull)
> - elog(ERROR, "NULL elements not allowed in Arrays");
> -
> - /* Use datumCopy to ensure pass-by-ref stuff is copied into mcontext */
> - astate->dvalues[astate->nelems++] =
> - datumCopy(dvalue, astate->typbyval, astate->typlen);
> -
> - MemoryContextSwitchTo(oldcontext);
> -
> - return astate;
> - }
> -
> - /*
> - * makeArrayResult - produce final result of ARRAY_SUBLINK
> - *
> - * astate is working state (not NULL)
> - * rcontext is where to construct result
> - */
> - static Datum
> - makeArrayResult(ArrayBuildState *astate,
> - MemoryContext rcontext)
> - {
> - ArrayType *result;
> - int dims[1];
> - int lbs[1];
> - MemoryContext oldcontext;
> -
> - /* Build the final array result in rcontext */
> - oldcontext = MemoryContextSwitchTo(rcontext);
> -
> - dims[0] = astate->nelems;
> - lbs[0] = 1;
> -
> - result = construct_md_array(astate->dvalues,
> - 1,
> - dims,
> - lbs,
> - astate->element_type,
> - astate->typlen,
> - astate->typbyval,
> - astate->typalign);
> -
> - MemoryContextSwitchTo(oldcontext);
> -
> - /* Clean up all the junk */
> - MemoryContextDelete(astate->mcontext);
> -
> - return PointerGetDatum(result);
> }
> --- 1228,1231 ----
> Index: src/backend/nodes/copyfuncs.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/nodes/copyfuncs.c,v
> retrieving revision 1.251
> diff -c -r1.251 copyfuncs.c
> *** src/backend/nodes/copyfuncs.c 28 May 2003 16:03:56 -0000 1.251
> --- src/backend/nodes/copyfuncs.c 1 Jun 2003 15:41:47 -0000
> ***************
> *** 727,732 ****
> --- 727,733 ----
> COPY_NODE_FIELD(target);
> COPY_SCALAR_FIELD(aggstar);
> COPY_SCALAR_FIELD(aggdistinct);
> + COPY_NODE_FIELD(args);
>
> return newnode;
> }
> ***************
> *** 825,830 ****
> --- 826,832 ----
>
> COPY_SCALAR_FIELD(subLinkType);
> COPY_SCALAR_FIELD(useOr);
> + COPY_SCALAR_FIELD(isExpr);
> COPY_NODE_FIELD(lefthand);
> COPY_NODE_FIELD(operName);
> COPY_OIDLIST_FIELD(operOids);
> ***************
> *** 843,848 ****
> --- 845,856 ----
>
> COPY_SCALAR_FIELD(subLinkType);
> COPY_SCALAR_FIELD(useOr);
> + COPY_SCALAR_FIELD(isExpr);
> + COPY_SCALAR_FIELD(exprtype);
> + COPY_SCALAR_FIELD(elemtype);
> + COPY_SCALAR_FIELD(elmlen);
> + COPY_SCALAR_FIELD(elmbyval);
> + COPY_SCALAR_FIELD(elmalign);
> COPY_NODE_FIELD(exprs);
> COPY_INTLIST_FIELD(paramIds);
> COPY_NODE_FIELD(plan);
> Index: src/backend/nodes/equalfuncs.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/nodes/equalfuncs.c,v
> retrieving revision 1.194
> diff -c -r1.194 equalfuncs.c
> *** src/backend/nodes/equalfuncs.c 28 May 2003 16:03:56 -0000 1.194
> --- src/backend/nodes/equalfuncs.c 1 Jun 2003 15:41:47 -0000
> ***************
> *** 204,209 ****
> --- 204,210 ----
> COMPARE_NODE_FIELD(target);
> COMPARE_SCALAR_FIELD(aggstar);
> COMPARE_SCALAR_FIELD(aggdistinct);
> + COMPARE_NODE_FIELD(args);
>
> return true;
> }
> ***************
> *** 300,305 ****
> --- 301,307 ----
> {
> COMPARE_SCALAR_FIELD(subLinkType);
> COMPARE_SCALAR_FIELD(useOr);
> + COMPARE_SCALAR_FIELD(isExpr);
> COMPARE_NODE_FIELD(lefthand);
> COMPARE_NODE_FIELD(operName);
> COMPARE_OIDLIST_FIELD(operOids);
> ***************
> *** 313,318 ****
> --- 315,326 ----
> {
> COMPARE_SCALAR_FIELD(subLinkType);
> COMPARE_SCALAR_FIELD(useOr);
> + COMPARE_SCALAR_FIELD(isExpr);
> + COMPARE_SCALAR_FIELD(exprtype);
> + COMPARE_SCALAR_FIELD(elemtype);
> + COMPARE_SCALAR_FIELD(elmlen);
> + COMPARE_SCALAR_FIELD(elmbyval);
> + COMPARE_SCALAR_FIELD(elmalign);
> COMPARE_NODE_FIELD(exprs);
> COMPARE_INTLIST_FIELD(paramIds);
> /* should compare plans, but have to settle for comparing plan IDs */
> Index: src/backend/nodes/outfuncs.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/nodes/outfuncs.c,v
> retrieving revision 1.206
> diff -c -r1.206 outfuncs.c
> *** src/backend/nodes/outfuncs.c 28 May 2003 16:03:56 -0000 1.206
> --- src/backend/nodes/outfuncs.c 1 Jun 2003 15:41:47 -0000
> ***************
> *** 615,620 ****
> --- 615,621 ----
> WRITE_NODE_FIELD(target);
> WRITE_BOOL_FIELD(aggstar);
> WRITE_BOOL_FIELD(aggdistinct);
> + WRITE_NODE_FIELD(args);
> }
>
> static void
> ***************
> *** 700,705 ****
> --- 701,707 ----
>
> WRITE_ENUM_FIELD(subLinkType, SubLinkType);
> WRITE_BOOL_FIELD(useOr);
> + WRITE_BOOL_FIELD(isExpr);
> WRITE_NODE_FIELD(lefthand);
> WRITE_NODE_FIELD(operName);
> WRITE_OIDLIST_FIELD(operOids);
> ***************
> *** 713,718 ****
> --- 715,726 ----
>
> WRITE_ENUM_FIELD(subLinkType, SubLinkType);
> WRITE_BOOL_FIELD(useOr);
> + WRITE_BOOL_FIELD(isExpr);
> + WRITE_OID_FIELD(exprtype);
> + WRITE_OID_FIELD(elemtype);
> + WRITE_INT_FIELD(elmlen);
> + WRITE_BOOL_FIELD(elmbyval);
> + WRITE_CHAR_FIELD(elmalign);
> WRITE_NODE_FIELD(exprs);
> WRITE_INTLIST_FIELD(paramIds);
> WRITE_NODE_FIELD(plan);
> Index: src/backend/nodes/readfuncs.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/nodes/readfuncs.c,v
> retrieving revision 1.153
> diff -c -r1.153 readfuncs.c
> *** src/backend/nodes/readfuncs.c 6 May 2003 00:20:32 -0000 1.153
> --- src/backend/nodes/readfuncs.c 1 Jun 2003 15:41:47 -0000
> ***************
> *** 415,420 ****
> --- 415,421 ----
> READ_NODE_FIELD(target);
> READ_BOOL_FIELD(aggstar);
> READ_BOOL_FIELD(aggdistinct);
> + READ_NODE_FIELD(args);
>
> READ_DONE();
> }
> ***************
> *** 544,549 ****
> --- 545,551 ----
>
> READ_ENUM_FIELD(subLinkType, SubLinkType);
> READ_BOOL_FIELD(useOr);
> + READ_BOOL_FIELD(isExpr);
> READ_NODE_FIELD(lefthand);
> READ_NODE_FIELD(operName);
> READ_OIDLIST_FIELD(operOids);
> Index: src/backend/optimizer/plan/subselect.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/optimizer/plan/subselect.c,v
> retrieving revision 1.75
> diff -c -r1.75 subselect.c
> *** src/backend/optimizer/plan/subselect.c 29 Apr 2003 22:13:09 -0000 1.75
> --- src/backend/optimizer/plan/subselect.c 1 Jun 2003 15:41:47 -0000
> ***************
> *** 67,73 ****
>
> static List *convert_sublink_opers(List *lefthand, List *operOids,
> List *targetlist, int rtindex,
> ! List **righthandIds);
> static bool subplan_is_hashable(SubLink *slink, SubPlan *node);
> static Node *replace_correlation_vars_mutator(Node *node, void *context);
> static Node *process_sublinks_mutator(Node *node, bool *isTopQual);
> --- 67,73 ----
>
> static List *convert_sublink_opers(List *lefthand, List *operOids,
> List *targetlist, int rtindex,
> ! bool isExpr, List **righthandIds);
> static bool subplan_is_hashable(SubLink *slink, SubPlan *node);
> static Node *replace_correlation_vars_mutator(Node *node, void *context);
> static Node *process_sublinks_mutator(Node *node, bool *isTopQual);
> ***************
> *** 240,245 ****
> --- 240,251 ----
> */
> node->subLinkType = slink->subLinkType;
> node->useOr = slink->useOr;
> + node->isExpr = slink->isExpr;
> + node->exprtype = InvalidOid;
> + node->elemtype = InvalidOid;
> + node->elmlen = 0;
> + node->elmbyval = false;
> + node->elmalign = '\0';
> node->exprs = NIL;
> node->paramIds = NIL;
> node->useHashTable = false;
> ***************
> *** 316,322 ****
> exprs = convert_sublink_opers(lefthand,
> slink->operOids,
> plan->targetlist,
> ! 0,
> &node->paramIds);
> node->setParam = listCopy(node->paramIds);
> PlannerInitPlan = lappend(PlannerInitPlan, node);
> --- 322,328 ----
> exprs = convert_sublink_opers(lefthand,
> slink->operOids,
> plan->targetlist,
> ! 0, node->isExpr,
> &node->paramIds);
> node->setParam = listCopy(node->paramIds);
> PlannerInitPlan = lappend(PlannerInitPlan, node);
> ***************
> *** 399,405 ****
> node->exprs = convert_sublink_opers(lefthand,
> slink->operOids,
> plan->targetlist,
> ! 0,
> &node->paramIds);
>
> /*
> --- 405,411 ----
> node->exprs = convert_sublink_opers(lefthand,
> slink->operOids,
> plan->targetlist,
> ! 0, node->isExpr,
> &node->paramIds);
>
> /*
> ***************
> *** 444,450 ****
> static List *
> convert_sublink_opers(List *lefthand, List *operOids,
> List *targetlist, int rtindex,
> ! List **righthandIds)
> {
> List *result = NIL;
> List *lst;
> --- 450,456 ----
> static List *
> convert_sublink_opers(List *lefthand, List *operOids,
> List *targetlist, int rtindex,
> ! bool isExpr, List **righthandIds)
> {
> List *result = NIL;
> List *lst;
> ***************
> *** 499,511 ****
> * are not expecting to have to resolve unknown Params, so
> * it's okay to pass a null pstate.)
> */
> ! result = lappend(result,
> ! make_op_expr(NULL,
> ! tup,
> ! leftop,
> ! rightop,
> ! exprType(leftop),
> ! te->resdom->restype));
>
> ReleaseSysCache(tup);
>
> --- 505,542 ----
> * are not expecting to have to resolve unknown Params, so
> * it's okay to pass a null pstate.)
> */
> ! if (!isExpr)
> ! {
> ! result = lappend(result,
> ! make_op_expr(NULL,
> ! tup,
> ! leftop,
> ! rightop,
> ! exprType(leftop),
> ! te->resdom->restype));
> ! }
> ! else
> ! {
> ! Oid exprtype = te->resdom->restype;
> ! Oid elemtype = get_element_type(exprtype);
> !
> ! if (elemtype != InvalidOid)
> ! result = lappend(result,
> ! make_op_expr(NULL,
> ! tup,
> ! leftop,
> ! rightop,
> ! exprType(leftop),
> ! elemtype));
> ! else
> ! result = lappend(result,
> ! make_op_expr(NULL,
> ! tup,
> ! leftop,
> ! rightop,
> ! exprType(leftop),
> ! exprtype));
> ! }
>
> ReleaseSysCache(tup);
>
> ***************
> *** 616,628 ****
> /*
> * The sublink type must be "= ANY" --- that is, an IN operator.
> * (We require the operator name to be unqualified, which may be
> ! * overly paranoid, or may not be.)
> */
> if (sublink->subLinkType != ANY_SUBLINK)
> return NULL;
> if (length(sublink->operName) != 1 ||
> strcmp(strVal(lfirst(sublink->operName)), "=") != 0)
> return NULL;
> /*
> * The sub-select must not refer to any Vars of the parent query.
> * (Vars of higher levels should be okay, though.)
> --- 647,663 ----
> /*
> * The sublink type must be "= ANY" --- that is, an IN operator.
> * (We require the operator name to be unqualified, which may be
> ! * overly paranoid, or may not be.) It must not be an Expression
> ! * sublink.
> */
> if (sublink->subLinkType != ANY_SUBLINK)
> return NULL;
> if (length(sublink->operName) != 1 ||
> strcmp(strVal(lfirst(sublink->operName)), "=") != 0)
> return NULL;
> + if (sublink->isExpr)
> + return NULL;
> +
> /*
> * The sub-select must not refer to any Vars of the parent query.
> * (Vars of higher levels should be okay, though.)
> ***************
> *** 675,681 ****
> exprs = convert_sublink_opers(sublink->lefthand,
> sublink->operOids,
> subselect->targetList,
> ! rtindex,
> &ininfo->sub_targetlist);
> return (Node *) make_ands_explicit(exprs);
> }
> --- 710,716 ----
> exprs = convert_sublink_opers(sublink->lefthand,
> sublink->operOids,
> subselect->targetList,
> ! rtindex, sublink->isExpr,
> &ininfo->sub_targetlist);
> return (Node *) make_ands_explicit(exprs);
> }
> Index: src/backend/optimizer/util/clauses.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/optimizer/util/clauses.c,v
> retrieving revision 1.138
> diff -c -r1.138 clauses.c
> *** src/backend/optimizer/util/clauses.c 28 May 2003 22:32:49 -0000 1.138
> --- src/backend/optimizer/util/clauses.c 1 Jun 2003 15:41:15 -0000
> ***************
> *** 133,138 ****
> --- 133,160 ----
> }
>
> /*****************************************************************************
> + * FUNCTION clause functions
> + *****************************************************************************/
> +
> + /*
> + * make_funcclause
> + * Creates a function clause given its function info and argument list.
> + */
> + Expr *
> + make_funcclause(Oid funcid, Oid funcresulttype, bool funcretset,
> + CoercionForm funcformat, List *funcargs)
> + {
> + FuncExpr *expr = makeNode(FuncExpr);
> +
> + expr->funcid = funcid;
> + expr->funcresulttype = funcresulttype;
> + expr->funcretset = funcretset;
> + expr->funcformat = funcformat;
> + expr->args = funcargs;
> + return (Expr *) expr;
> + }
> +
> + /*****************************************************************************
> * NOT clause functions
> *****************************************************************************/
>
> Index: src/backend/parser/gram.y
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/parser/gram.y,v
> retrieving revision 2.416
> diff -c -r2.416 gram.y
> *** src/backend/parser/gram.y 29 May 2003 20:40:36 -0000 2.416
> --- src/backend/parser/gram.y 1 Jun 2003 15:41:47 -0000
> ***************
> *** 5490,5495 ****
> --- 5490,5496 ----
> {
> SubLink *n = makeNode(SubLink);
> n->subLinkType = ANY_SUBLINK;
> + n->isExpr = false;
> n->lefthand = $1;
> n->operName = makeList1(makeString("="));
> n->subselect = $3;
> ***************
> *** 5500,5505 ****
> --- 5501,5507 ----
> /* Make an IN node */
> SubLink *n = makeNode(SubLink);
> n->subLinkType = ANY_SUBLINK;
> + n->isExpr = false;
> n->lefthand = $1;
> n->operName = makeList1(makeString("="));
> n->subselect = $4;
> ***************
> *** 5511,5516 ****
> --- 5513,5519 ----
> {
> SubLink *n = makeNode(SubLink);
> n->subLinkType = $3;
> + n->isExpr = false;
> n->lefthand = $1;
> n->operName = $2;
> n->subselect = $4;
> ***************
> *** 5521,5526 ****
> --- 5524,5530 ----
> {
> SubLink *n = makeNode(SubLink);
> n->subLinkType = MULTIEXPR_SUBLINK;
> + n->isExpr = false;
> n->lefthand = $1;
> n->operName = $2;
> n->subselect = $3;
> ***************
> *** 5904,5909 ****
> --- 5908,5914 ----
> {
> SubLink *n = (SubLink *)$3;
> n->subLinkType = ANY_SUBLINK;
> + n->isExpr = false;
> n->lefthand = makeList1($1);
> n->operName = makeList1(makeString("="));
> $$ = (Node *)n;
> ***************
> *** 5931,5936 ****
> --- 5936,5942 ----
> {
> /* Make an IN node */
> SubLink *n = (SubLink *)$4;
> + n->isExpr = false;
> n->subLinkType = ANY_SUBLINK;
> n->lefthand = makeList1($1);
> n->operName = makeList1(makeString("="));
> ***************
> *** 5957,5967 ****
> --- 5963,6000 ----
> {
> SubLink *n = makeNode(SubLink);
> n->subLinkType = $3;
> + n->isExpr = false;
> n->lefthand = makeList1($1);
> n->operName = $2;
> n->subselect = $4;
> $$ = (Node *)n;
> }
> + | a_expr qual_all_Op sub_type '(' a_expr ')' %prec Op
> + {
> + SubLink *n = makeNode(SubLink);
> + SelectStmt *s = makeNode(SelectStmt);
> + ResTarget *r = makeNode(ResTarget);
> +
> + r->name = NULL;
> + r->indirection = NIL;
> + r->val = (Node *)$5;
> +
> + s->distinctClause = NIL;
> + s->targetList = makeList1(r);
> + s->into = NULL;
> + s->intoColNames = NIL;
> + s->fromClause = NIL;
> + s->whereClause = NULL;
> + s->groupClause = NIL;
> + s->havingClause = NULL;
> +
> + n->subLinkType = $3;
> + n->isExpr = true;
> + n->lefthand = makeList1($1);
> + n->operName = $2;
> + n->subselect = (Node *) s;
> + $$ = (Node *)n;
> + }
> | UNIQUE select_with_parens %prec Op
> {
> /* Not sure how to get rid of the parentheses
> ***************
> *** 6538,6543 ****
> --- 6571,6577 ----
> {
> SubLink *n = makeNode(SubLink);
> n->subLinkType = EXPR_SUBLINK;
> + n->isExpr = false;
> n->lefthand = NIL;
> n->operName = NIL;
> n->subselect = $1;
> ***************
> *** 6547,6552 ****
> --- 6581,6587 ----
> {
> SubLink *n = makeNode(SubLink);
> n->subLinkType = EXISTS_SUBLINK;
> + n->isExpr = false;
> n->lefthand = NIL;
> n->operName = NIL;
> n->subselect = $2;
> ***************
> *** 6556,6561 ****
> --- 6591,6597 ----
> {
> SubLink *n = makeNode(SubLink);
> n->subLinkType = ARRAY_SUBLINK;
> + n->isExpr = false;
> n->lefthand = NIL;
> n->operName = NIL;
> n->subselect = $2;
> ***************
> *** 6730,6735 ****
> --- 6766,6772 ----
> in_expr: select_with_parens
> {
> SubLink *n = makeNode(SubLink);
> + n->isExpr = false;
> n->subselect = $1;
> /* other fields will be filled later */
> $$ = (Node *)n;
> Index: src/backend/parser/parse_coerce.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/parser/parse_coerce.c,v
> retrieving revision 2.97
> diff -c -r2.97 parse_coerce.c
> *** src/backend/parser/parse_coerce.c 26 May 2003 00:11:27 -0000 2.97
> --- src/backend/parser/parse_coerce.c 1 Jun 2003 16:15:40 -0000
> ***************
> *** 859,865 ****
> /* Get the element type based on the array type, if we have one */
> if (OidIsValid(array_typeid))
> {
> ! array_typelem = get_element_type(array_typeid);
> if (!OidIsValid(array_typelem))
> elog(ERROR, "Argument declared ANYARRAY is not an array: %s",
> format_type_be(array_typeid));
> --- 859,869 ----
> /* Get the element type based on the array type, if we have one */
> if (OidIsValid(array_typeid))
> {
> ! if (array_typeid != ANYARRAYOID)
> ! array_typelem = get_element_type(array_typeid);
> ! else
> ! array_typelem = ANYELEMENTOID;
> !
> if (!OidIsValid(array_typelem))
> elog(ERROR, "Argument declared ANYARRAY is not an array: %s",
> format_type_be(array_typeid));
> ***************
> *** 919,925 ****
> {
> if (!OidIsValid(array_typeid))
> {
> ! array_typeid = get_array_type(elem_typeid);
> if (!OidIsValid(array_typeid))
> elog(ERROR, "Cannot find array type for datatype %s",
> format_type_be(elem_typeid));
> --- 923,933 ----
> {
> if (!OidIsValid(array_typeid))
> {
> ! if (elem_typeid != ANYELEMENTOID)
> ! array_typeid = get_array_type(elem_typeid);
> ! else
> ! array_typeid = ANYARRAYOID;
> !
> if (!OidIsValid(array_typeid))
> elog(ERROR, "Cannot find array type for datatype %s",
> format_type_be(elem_typeid));
> ***************
> *** 1169,1174 ****
> --- 1177,1187 ----
> /* Somewhat-fast path for domain -> base type case */
> if (srctype == targettype)
> return true;
> +
> + /* Last of the fast-paths: check for matching polymorphic arrays */
> + if (targettype == ANYARRAYOID)
> + if (get_element_type(srctype) != InvalidOid)
> + return true;
>
> /* Else look in pg_cast */
> tuple = SearchSysCache(CASTSOURCETARGET,
> Index: src/backend/parser/parse_expr.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/parser/parse_expr.c,v
> retrieving revision 1.148
> diff -c -r1.148 parse_expr.c
> *** src/backend/parser/parse_expr.c 29 Apr 2003 22:13:10 -0000 1.148
> --- src/backend/parser/parse_expr.c 1 Jun 2003 15:41:47 -0000
> ***************
> *** 436,441 ****
> --- 436,442 ----
> sublink->operName = NIL;
> sublink->operOids = NIL;
> sublink->useOr = FALSE;
> + sublink->isExpr = FALSE;
> }
> else if (sublink->subLinkType == EXPR_SUBLINK ||
> sublink->subLinkType == ARRAY_SUBLINK)
> ***************
> *** 463,468 ****
> --- 464,470 ----
> sublink->operName = NIL;
> sublink->operOids = NIL;
> sublink->useOr = FALSE;
> + sublink->isExpr = FALSE;
> }
> else
> {
> ***************
> *** 538,547 ****
> * here, because make_subplan() will insert type
> * coercion calls if needed.
> */
> ! optup = oper(op,
> ! exprType(lexpr),
> ! exprType((Node *) tent->expr),
> ! false);
> opform = (Form_pg_operator) GETSTRUCT(optup);
>
> if (opform->oprresult != BOOLOID)
> --- 540,569 ----
> * here, because make_subplan() will insert type
> * coercion calls if needed.
> */
> ! if (!sublink->isExpr)
> ! {
> ! optup = oper(op,
> ! exprType(lexpr),
> ! exprType((Node *) tent->expr),
> ! false);
> ! }
> ! else
> ! {
> ! Oid exprtype = exprType((Node *) tent->expr);
> ! Oid elemtype = get_element_type(exprtype);
> !
> ! if (elemtype != InvalidOid)
> ! optup = oper(op,
> ! exprType(lexpr),
> ! elemtype,
> ! false);
> ! else
> ! optup = oper(op,
> ! exprType(lexpr),
> ! exprtype,
> ! false);
> ! }
> !
> opform = (Form_pg_operator) GETSTRUCT(optup);
>
> if (opform->oprresult != BOOLOID)
> Index: src/backend/parser/parse_func.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/parser/parse_func.c,v
> retrieving revision 1.148
> diff -c -r1.148 parse_func.c
> *** src/backend/parser/parse_func.c 26 May 2003 00:11:27 -0000 1.148
> --- src/backend/parser/parse_func.c 1 Jun 2003 15:41:15 -0000
> ***************
> *** 335,340 ****
> --- 335,341 ----
> aggref->target = lfirst(fargs);
> aggref->aggstar = agg_star;
> aggref->aggdistinct = agg_distinct;
> + aggref->args = fargs;
>
> retval = (Node *) aggref;
>
> Index: src/backend/parser/parse_oper.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/parser/parse_oper.c,v
> retrieving revision 1.64
> diff -c -r1.64 parse_oper.c
> *** src/backend/parser/parse_oper.c 26 May 2003 00:11:27 -0000 1.64
> --- src/backend/parser/parse_oper.c 1 Jun 2003 23:21:10 -0000
> ***************
> *** 221,226 ****
> --- 221,241 ----
> }
>
> /*
> + * ordering_oper_funcid - convenience routine for oprfuncid(ordering_oper())
> + */
> + Oid
> + ordering_oper_funcid(Oid argtype)
> + {
> + Operator optup;
> + Oid result;
> +
> + optup = ordering_oper(argtype, false);
> + result = oprfuncid(optup);
> + ReleaseSysCache(optup);
> + return result;
> + }
> +
> + /*
> * ordering_oper_opid - convenience routine for oprid(ordering_oper())
> *
> * This was formerly called any_ordering_op()
> Index: src/backend/utils/adt/acl.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/utils/adt/acl.c,v
> retrieving revision 1.86
> diff -c -r1.86 acl.c
> *** src/backend/utils/adt/acl.c 24 Jan 2003 21:53:29 -0000 1.86
> --- src/backend/utils/adt/acl.c 1 Jun 2003 19:07:00 -0000
> ***************
> *** 720,732 ****
> aidat = ACL_DAT(acl);
> for (i = 0; i < num; ++i)
> {
> ! if (aip->ai_grantee == aidat[i].ai_grantee &&
> ! aip->ai_privs == aidat[i].ai_privs)
> PG_RETURN_BOOL(true);
> }
> PG_RETURN_BOOL(false);
> }
>
>
> /*
> * has_table_privilege variants
> --- 720,740 ----
> aidat = ACL_DAT(acl);
> for (i = 0; i < num; ++i)
> {
> ! if (aclitemeq(aip, &aidat[i]))
> PG_RETURN_BOOL(true);
> }
> PG_RETURN_BOOL(false);
> }
>
> + /*
> + * user-facing version of aclitemeq() for use as the
> + * aclitem equality operator
> + */
> + Datum
> + aclitem_eq(PG_FUNCTION_ARGS)
> + {
> + PG_RETURN_BOOL(aclitemeq(PG_GETARG_ACLITEM_P(0), PG_GETARG_ACLITEM_P(1)));
> + }
>
> /*
> * has_table_privilege variants
> Index: src/backend/utils/adt/array_userfuncs.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/utils/adt/array_userfuncs.c,v
> retrieving revision 1.1
> diff -c -r1.1 array_userfuncs.c
> *** src/backend/utils/adt/array_userfuncs.c 8 Apr 2003 23:20:02 -0000 1.1
> --- src/backend/utils/adt/array_userfuncs.c 1 Jun 2003 15:41:15 -0000
> ***************
> *** 42,48 ****
> else
> ndims = 1;
>
> ! PG_RETURN_ARRAYTYPE_P(create_singleton_array(elem_type,
> PG_GETARG_DATUM(0),
> ndims));
> }
> --- 42,48 ----
> else
> ndims = 1;
>
> ! PG_RETURN_ARRAYTYPE_P(create_singleton_array(fcinfo, elem_type,
> PG_GETARG_DATUM(0),
> ndims));
> }
> ***************
> *** 70,75 ****
> --- 70,76 ----
> Oid arg1_typeid = get_fn_expr_argtype(fcinfo, 1);
> Oid arg0_elemid;
> Oid arg1_elemid;
> + ArrayMetaState *my_extra;
>
> if (arg0_typeid == InvalidOid || arg1_typeid == InvalidOid)
> elog(ERROR, "array_push: cannot determine input data types");
> ***************
> *** 95,122 ****
> PG_RETURN_NULL(); /* keep compiler quiet */
> }
>
> ! /* Sanity check: do we have a one-dimensional array */
> ! if (ARR_NDIM(v) != 1)
> ! elog(ERROR, "Arrays greater than one-dimension are not supported");
> !
> ! lb = ARR_LBOUND(v);
> ! dimv = ARR_DIMS(v);
> ! if (arg0_elemid != InvalidOid)
> {
> ! /* append newelem */
> ! int ub = dimv[0] + lb[0] - 1;
> ! indx = ub + 1;
> }
> else
> {
> ! /* prepend newelem */
> ! indx = lb[0] - 1;
> }
>
> ! get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign);
>
> ! result = array_set(v, 1, &indx, newelem, -1,
> ! typlen, typbyval, typalign, &isNull);
>
> PG_RETURN_ARRAYTYPE_P(result);
> }
> --- 96,156 ----
> PG_RETURN_NULL(); /* keep compiler quiet */
> }
>
> ! if (ARR_NDIM(v) == 1)
> {
> ! lb = ARR_LBOUND(v);
> ! dimv = ARR_DIMS(v);
> !
> ! if (arg0_elemid != InvalidOid)
> ! {
> ! /* append newelem */
> ! int ub = dimv[0] + lb[0] - 1;
> ! indx = ub + 1;
> ! }
> ! else
> ! {
> ! /* prepend newelem */
> ! indx = lb[0] - 1;
> ! }
> }
> + else if (ARR_NDIM(v) == 0)
> + indx = 1;
> else
> + elog(ERROR, "only empty and one-dimensional arrays are supported");
> +
> +
> + /*
> + * We arrange to look up info about element type only once per series
> + * of calls, assuming the element type doesn't change underneath us.
> + */
> + my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> + if (my_extra == NULL)
> {
> ! fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! my_extra->element_type = InvalidOid;
> }
>
> ! if (my_extra->element_type != element_type)
> ! {
> ! /* Get info about element type */
> ! get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign);
>
> ! my_extra->element_type = element_type;
> ! my_extra->typlen = typlen;
> ! my_extra->typbyval = typbyval;
> ! my_extra->typalign = typalign;
> ! }
> ! else
> ! {
> ! typlen = my_extra->typlen;
> ! typbyval = my_extra->typbyval;
> ! typalign = my_extra->typalign;
> ! }
> !
> ! result = array_set(v, 1, &indx, newelem, -1, typlen, typbyval,
> ! typalign, &isNull);
>
> PG_RETURN_ARRAYTYPE_P(result);
> }
> ***************
> *** 145,157 ****
>
> /*
> * We must have one of the following combinations of inputs:
> ! * 1) two arrays with ndims1 == ndims2
> ! * 2) ndims1 == ndims2 - 1
> ! * 3) ndims1 == ndims2 + 1
> */
> ndims1 = ARR_NDIM(v1);
> ndims2 = ARR_NDIM(v2);
>
> if (ndims1 != ndims2 && ndims1 != ndims2 - 1 && ndims1 != ndims2 + 1)
> elog(ERROR, "Cannot concatenate incompatible arrays of %d and "
> "%d dimensions", ndims1, ndims2);
> --- 179,206 ----
>
> /*
> * We must have one of the following combinations of inputs:
> ! * 1) one empty array, and one non-empty array
> ! * 2) both arrays empty
> ! * 3) two arrays with ndims1 == ndims2
> ! * 4) ndims1 == ndims2 - 1
> ! * 5) ndims1 == ndims2 + 1
> */
> ndims1 = ARR_NDIM(v1);
> ndims2 = ARR_NDIM(v2);
>
> + /*
> + * short circuit - if one input array is empty, and the other is not,
> + * we return the non-empty one as the result
> + *
> + * if both are empty, return the first one
> + */
> + if (ndims1 == 0 && ndims2 > 0)
> + PG_RETURN_ARRAYTYPE_P(v2);
> +
> + if (ndims2 == 0)
> + PG_RETURN_ARRAYTYPE_P(v1);
> +
> + /* the rest fall into combo 2, 3, or 4 */
> if (ndims1 != ndims2 && ndims1 != ndims2 - 1 && ndims1 != ndims2 + 1)
> elog(ERROR, "Cannot concatenate incompatible arrays of %d and "
> "%d dimensions", ndims1, ndims2);
> ***************
> *** 266,314 ****
> PG_RETURN_ARRAYTYPE_P(result);
> }
>
> - /*----------------------------------------------------------------------------
> - * array_accum :
> - * accumulator to build a 1-D array from input values -- this can be used
> - * to create custom aggregates.
> - *
> - * This function is not marked strict, so we have to be careful about nulls.
> - *----------------------------------------------------------------------------
> - */
> - Datum
> - array_accum(PG_FUNCTION_ARGS)
> - {
> - /* return NULL if both arguments are NULL */
> - if (PG_ARGISNULL(0) && PG_ARGISNULL(1))
> - PG_RETURN_NULL();
> -
> - /* create a new 1-D array from the new element if the array is NULL */
> - if (PG_ARGISNULL(0))
> - {
> - Oid tgt_type = get_fn_expr_rettype(fcinfo);
> - Oid tgt_elem_type;
> -
> - if (tgt_type == InvalidOid)
> - elog(ERROR, "Cannot determine target array type");
> - tgt_elem_type = get_element_type(tgt_type);
> - if (tgt_elem_type == InvalidOid)
> - elog(ERROR, "Target type is not an array");
> -
> - PG_RETURN_ARRAYTYPE_P(create_singleton_array(tgt_elem_type,
> - PG_GETARG_DATUM(1),
> - 1));
> - }
> -
> - /* return the array if the new element is NULL */
> - if (PG_ARGISNULL(1))
> - PG_RETURN_ARRAYTYPE_P(PG_GETARG_ARRAYTYPE_P_COPY(0));
> -
> - /*
> - * Otherwise this is equivalent to array_push. We hack the call a little
> - * so that array_push can see the fn_expr information.
> - */
> - return array_push(fcinfo);
> - }
> -
> /*-----------------------------------------------------------------------------
> * array_assign :
> * assign an element of an array to a new value and return the
> --- 315,320 ----
> ***************
> *** 329,334 ****
> --- 335,341 ----
> int16 typlen;
> bool typbyval;
> char typalign;
> + ArrayMetaState *my_extra;
>
> v = PG_GETARG_ARRAYTYPE_P(0);
> idx_to_chg = PG_GETARG_INT32(1);
> ***************
> *** 349,355 ****
> if (element_type == 0)
> elog(ERROR, "Invalid array element type: %u", element_type);
>
> ! get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign);
>
> result = array_set(v, 1, &idx_to_chg, newelem, -1,
> typlen, typbyval, typalign, &isNull);
> --- 356,390 ----
> if (element_type == 0)
> elog(ERROR, "Invalid array element type: %u", element_type);
>
> ! /*
> ! * We arrange to look up info about element type only once per series
> ! * of calls, assuming the element type doesn't change underneath us.
> ! */
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! if (my_extra == NULL)
> ! {
> ! fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! my_extra->element_type = InvalidOid;
> ! }
> !
> ! if (my_extra->element_type != element_type)
> ! {
> ! /* Get info about element type */
> ! get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign);
> !
> ! my_extra->element_type = element_type;
> ! my_extra->typlen = typlen;
> ! my_extra->typbyval = typbyval;
> ! my_extra->typalign = typalign;
> ! }
> ! else
> ! {
> ! typlen = my_extra->typlen;
> ! typbyval = my_extra->typbyval;
> ! typalign = my_extra->typalign;
> ! }
>
> result = array_set(v, 1, &idx_to_chg, newelem, -1,
> typlen, typbyval, typalign, &isNull);
> ***************
> *** 375,380 ****
> --- 410,416 ----
> int16 typlen;
> bool typbyval;
> char typalign;
> + ArrayMetaState *my_extra;
>
> v = PG_GETARG_ARRAYTYPE_P(0);
> idx = PG_GETARG_INT32(1);
> ***************
> *** 394,400 ****
> if (element_type == 0)
> elog(ERROR, "Invalid array element type: %u", element_type);
>
> ! get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign);
>
> result = array_ref(v, 1, &idx, -1, typlen, typbyval, typalign, &isNull);
>
> --- 430,464 ----
> if (element_type == 0)
> elog(ERROR, "Invalid array element type: %u", element_type);
>
> ! /*
> ! * We arrange to look up info about element type only once per series
> ! * of calls, assuming the element type doesn't change underneath us.
> ! */
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! if (my_extra == NULL)
> ! {
> ! fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! my_extra->element_type = InvalidOid;
> ! }
> !
> ! if (my_extra->element_type != element_type)
> ! {
> ! /* Get info about element type */
> ! get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign);
> !
> ! my_extra->element_type = element_type;
> ! my_extra->typlen = typlen;
> ! my_extra->typbyval = typbyval;
> ! my_extra->typalign = typalign;
> ! }
> ! else
> ! {
> ! typlen = my_extra->typlen;
> ! typbyval = my_extra->typbyval;
> ! typalign = my_extra->typalign;
> ! }
>
> result = array_ref(v, 1, &idx, -1, typlen, typbyval, typalign, &isNull);
>
> ***************
> *** 402,412 ****
> }
>
> /*
> ! * actually does the work for singleton_array(), and array_accum() if it is
> ! * given a null input array.
> */
> ArrayType *
> ! create_singleton_array(Oid element_type, Datum element, int ndims)
> {
> Datum dvalues[1];
> int16 typlen;
> --- 466,478 ----
> }
>
> /*
> ! * actually does the work for singleton_array()
> */
> ArrayType *
> ! create_singleton_array(FunctionCallInfo fcinfo,
> ! Oid element_type,
> ! Datum element,
> ! int ndims)
> {
> Datum dvalues[1];
> int16 typlen;
> ***************
> *** 415,420 ****
> --- 481,487 ----
> int dims[MAXDIM];
> int lbs[MAXDIM];
> int i;
> + ArrayMetaState *my_extra;
>
> if (element_type == 0)
> elog(ERROR, "Invalid array element type: %u", element_type);
> ***************
> *** 429,435 ****
> lbs[i] = 1;
> }
>
> ! get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign);
>
> return construct_md_array(dvalues, ndims, dims, lbs, element_type,
> typlen, typbyval, typalign);
> --- 496,530 ----
> lbs[i] = 1;
> }
>
> ! /*
> ! * We arrange to look up info about element type only once per series
> ! * of calls, assuming the element type doesn't change underneath us.
> ! */
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! if (my_extra == NULL)
> ! {
> ! fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! my_extra->element_type = InvalidOid;
> ! }
> !
> ! if (my_extra->element_type != element_type)
> ! {
> ! /* Get info about element type */
> ! get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign);
> !
> ! my_extra->element_type = element_type;
> ! my_extra->typlen = typlen;
> ! my_extra->typbyval = typbyval;
> ! my_extra->typalign = typalign;
> ! }
> ! else
> ! {
> ! typlen = my_extra->typlen;
> ! typbyval = my_extra->typbyval;
> ! typalign = my_extra->typalign;
> ! }
>
> return construct_md_array(dvalues, ndims, dims, lbs, element_type,
> typlen, typbyval, typalign);
> Index: src/backend/utils/adt/arrayfuncs.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/utils/adt/arrayfuncs.c,v
> retrieving revision 1.89
> diff -c -r1.89 arrayfuncs.c
> *** src/backend/utils/adt/arrayfuncs.c 9 May 2003 23:01:45 -0000 1.89
> --- src/backend/utils/adt/arrayfuncs.c 2 Jun 2003 02:24:39 -0000
> ***************
> *** 21,28 ****
> --- 21,30 ----
> #include "catalog/pg_type.h"
> #include "libpq/pqformat.h"
> #include "parser/parse_coerce.h"
> + #include "parser/parse_oper.h"
> #include "utils/array.h"
> #include "utils/builtins.h"
> + #include "utils/datum.h"
> #include "utils/memutils.h"
> #include "utils/lsyscache.h"
> #include "utils/syscache.h"
> ***************
> *** 70,85 ****
>
> #define RETURN_NULL(type) do { *isNull = true; return (type) 0; } while (0)
>
> - /* I/O function selector for system_cache_lookup */
> - typedef enum IOFuncSelector
> - {
> - IOFunc_input,
> - IOFunc_output,
> - IOFunc_receive,
> - IOFunc_send
> - } IOFuncSelector;
> -
> -
> static int ArrayCount(char *str, int *dim, char typdelim);
> static Datum *ReadArrayStr(char *arrayStr, int nitems, int ndim, int *dim,
> FmgrInfo *inputproc, Oid typelem, int32 typmod,
> --- 72,77 ----
> ***************
> *** 93,102 ****
> static void CopyArrayEls(char *p, Datum *values, int nitems,
> int typlen, bool typbyval, char typalign,
> bool freedata);
> - static void system_cache_lookup(Oid element_type, IOFuncSelector which_func,
> - int *typlen, bool *typbyval,
> - char *typdelim, Oid *typelem,
> - Oid *proc, char *typalign);
> static Datum ArrayCast(char *value, bool byval, int len);
> static int ArrayCastAndSet(Datum src,
> int typlen, bool typbyval, char typalign,
> --- 85,90 ----
> ***************
> *** 119,125 ****
> char *destPtr,
> int *st, int *endp, char *srcPtr,
> int typlen, bool typbyval, char typalign);
> !
>
> /*---------------------------------------------------------------------
> * array_in :
> --- 107,113 ----
> char *destPtr,
> int *st, int *endp, char *srcPtr,
> int typlen, bool typbyval, char typalign);
> ! static int array_cmp(FunctionCallInfo fcinfo);
>
> /*---------------------------------------------------------------------
> * array_in :
> ***************
> *** 154,165 ****
> dim[MAXDIM],
> lBound[MAXDIM];
> char typalign;
>
> ! /* Get info about element type, including its input conversion proc */
> ! system_cache_lookup(element_type, IOFunc_input,
> ! &typlen, &typbyval, &typdelim,
> ! &typelem, &typinput, &typalign);
> ! fmgr_info(typinput, &inputproc);
>
> /* Make a modifiable copy of the input */
> /* XXX why are we allocating an extra 2 bytes here? */
> --- 142,190 ----
> dim[MAXDIM],
> lBound[MAXDIM];
> char typalign;
> + ArrayMetaState *my_extra;
>
> ! /*
> ! * We arrange to look up info about element type, including its input
> ! * conversion proc only once per series of calls, assuming the element
> ! * type doesn't change underneath us.
> ! */
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! if (my_extra == NULL)
> ! {
> ! fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! my_extra->element_type = InvalidOid;
> ! }
> !
> ! if (my_extra->element_type != element_type)
> ! {
> ! /* Get info about element type, including its input conversion proc */
> ! get_type_metadata(element_type, IOFunc_input,
> ! &typlen, &typbyval, &typdelim,
> ! &typelem, &typinput, &typalign);
> ! fmgr_info(typinput, &inputproc);
> !
> ! my_extra->element_type = element_type;
> ! my_extra->typlen = typlen;
> ! my_extra->typbyval = typbyval;
> ! my_extra->typdelim = typdelim;
> ! my_extra->typelem = typelem;
> ! my_extra->typiofunc = typinput;
> ! my_extra->typalign = typalign;
> ! my_extra->proc = inputproc;
> ! }
> ! else
> ! {
> ! typlen = my_extra->typlen;
> ! typbyval = my_extra->typbyval;
> ! typdelim = my_extra->typdelim;
> ! typelem = my_extra->typelem;
> ! typinput = my_extra->typiofunc;
> ! typalign = my_extra->typalign;
> ! inputproc = my_extra->proc;
> ! }
>
> /* Make a modifiable copy of the input */
> /* XXX why are we allocating an extra 2 bytes here? */
> ***************
> *** 636,647 ****
> indx[MAXDIM];
> int ndim,
> *dim;
>
> element_type = ARR_ELEMTYPE(v);
> ! system_cache_lookup(element_type, IOFunc_output,
> ! &typlen, &typbyval, &typdelim,
> ! &typelem, &typoutput, &typalign);
> ! fmgr_info(typoutput, &outputproc);
>
> ndim = ARR_NDIM(v);
> dim = ARR_DIMS(v);
> --- 661,711 ----
> indx[MAXDIM];
> int ndim,
> *dim;
> + ArrayMetaState *my_extra;
>
> element_type = ARR_ELEMTYPE(v);
> !
> ! /*
> ! * We arrange to look up info about element type, including its input
> ! * conversion proc only once per series of calls, assuming the element
> ! * type doesn't change underneath us.
> ! */
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! if (my_extra == NULL)
> ! {
> ! fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! my_extra->element_type = InvalidOid;
> ! }
> !
> ! if (my_extra->element_type != element_type)
> ! {
> ! /* Get info about element type, including its output conversion proc */
> ! get_type_metadata(element_type, IOFunc_output,
> ! &typlen, &typbyval, &typdelim,
> ! &typelem, &typoutput, &typalign);
> ! fmgr_info(typoutput, &outputproc);
> !
> ! my_extra->element_type = element_type;
> ! my_extra->typlen = typlen;
> ! my_extra->typbyval = typbyval;
> ! my_extra->typdelim = typdelim;
> ! my_extra->typelem = typelem;
> ! my_extra->typiofunc = typoutput;
> ! my_extra->typalign = typalign;
> ! my_extra->proc = outputproc;
> ! }
> ! else
> ! {
> ! typlen = my_extra->typlen;
> ! typbyval = my_extra->typbyval;
> ! typdelim = my_extra->typdelim;
> ! typelem = my_extra->typelem;
> ! typoutput = my_extra->typiofunc;
> ! typalign = my_extra->typalign;
> ! outputproc = my_extra->proc;
> ! }
>
> ndim = ARR_NDIM(v);
> dim = ARR_DIMS(v);
> ***************
> *** 800,805 ****
> --- 864,870 ----
> dim[MAXDIM],
> lBound[MAXDIM];
> char typalign;
> + ArrayMetaState *my_extra;
>
> /* Get the array header information */
> ndim = pq_getmsgint(buf, 4);
> ***************
> *** 831,844 ****
> PG_RETURN_ARRAYTYPE_P(retval);
> }
>
> ! /* Get info about element type, including its receive conversion proc */
> ! system_cache_lookup(element_type, IOFunc_receive,
> ! &typlen, &typbyval, &typdelim,
> ! &typelem, &typreceive, &typalign);
> ! if (!OidIsValid(typreceive))
> ! elog(ERROR, "No binary input function available for type %s",
> ! format_type_be(element_type));
> ! fmgr_info(typreceive, &receiveproc);
>
> dataPtr = ReadArrayBinary(buf, nitems, &receiveproc, typelem,
> typlen, typbyval, typalign,
> --- 896,945 ----
> PG_RETURN_ARRAYTYPE_P(retval);
> }
>
> ! /*
> ! * We arrange to look up info about element type, including its receive
> ! * conversion proc only once per series of calls, assuming the element
> ! * type doesn't change underneath us.
> ! */
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! if (my_extra == NULL)
> ! {
> ! fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! my_extra->element_type = InvalidOid;
> ! }
> !
> ! if (my_extra->element_type != element_type)
> ! {
> ! /* Get info about element type, including its receive conversion proc */
> ! get_type_metadata(element_type, IOFunc_receive,
> ! &typlen, &typbyval, &typdelim,
> ! &typelem, &typreceive, &typalign);
> ! if (!OidIsValid(typreceive))
> ! elog(ERROR, "No binary input function available for type %s",
> ! format_type_be(element_type));
> ! fmgr_info(typreceive, &receiveproc);
> !
> ! my_extra->element_type = element_type;
> ! my_extra->typlen = typlen;
> ! my_extra->typbyval = typbyval;
> ! my_extra->typdelim = typdelim;
> ! my_extra->typelem = typelem;
> ! my_extra->typiofunc = typreceive;
> ! my_extra->typalign = typalign;
> ! my_extra->proc = receiveproc;
> ! }
> ! else
> ! {
> ! typlen = my_extra->typlen;
> ! typbyval = my_extra->typbyval;
> ! typdelim = my_extra->typdelim;
> ! typelem = my_extra->typelem;
> ! typreceive = my_extra->typiofunc;
> ! typalign = my_extra->typalign;
> ! receiveproc = my_extra->proc;
> ! }
>
> dataPtr = ReadArrayBinary(buf, nitems, &receiveproc, typelem,
> typlen, typbyval, typalign,
> ***************
> *** 976,990 ****
> int ndim,
> *dim;
> StringInfoData buf;
>
> /* Get information about the element type and the array dimensions */
> element_type = ARR_ELEMTYPE(v);
> ! system_cache_lookup(element_type, IOFunc_send, &typlen, &typbyval,
> ! &typdelim, &typelem, &typsend, &typalign);
> ! if (!OidIsValid(typsend))
> ! elog(ERROR, "No binary output function available for type %s",
> ! format_type_be(element_type));
> ! fmgr_info(typsend, &sendproc);
>
> ndim = ARR_NDIM(v);
> dim = ARR_DIMS(v);
> --- 1077,1130 ----
> int ndim,
> *dim;
> StringInfoData buf;
> + ArrayMetaState *my_extra;
>
> /* Get information about the element type and the array dimensions */
> element_type = ARR_ELEMTYPE(v);
> !
> ! /*
> ! * We arrange to look up info about element type, including its send
> ! * proc only once per series of calls, assuming the element
> ! * type doesn't change underneath us.
> ! */
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! if (my_extra == NULL)
> ! {
> ! fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> ! my_extra->element_type = InvalidOid;
> ! }
> !
> ! if (my_extra->element_type != element_type)
> ! {
> ! /* Get info about element type, including its send proc */
> ! get_type_metadata(element_type, IOFunc_send, &typlen, &typbyval,
> ! &typdelim, &typelem, &typsend, &typalign);
> ! if (!OidIsValid(typsend))
> ! elog(ERROR, "No binary output function available for type %s",
> ! format_type_be(element_type));
> ! fmgr_info(typsend, &sendproc);
> !
> ! my_extra->element_type = element_type;
> ! my_extra->typlen = typlen;
> ! my_extra->typbyval = typbyval;
> ! my_extra->typdelim = typdelim;
> ! my_extra->typelem = typelem;
> ! my_extra->typiofunc = typsend;
> ! my_extra->typalign = typalign;
> ! my_extra->proc = sendproc;
> ! }
> ! else
> ! {
> ! typlen = my_extra->typlen;
> ! typbyval = my_extra->typbyval;
> ! typdelim = my_extra->typdelim;
> ! typelem = my_extra->typelem;
> ! typsend = my_extra->typiofunc;
> ! typalign = my_extra->typalign;
> ! sendproc = my_extra->proc;
> ! }
>
> ndim = ARR_NDIM(v);
> dim = ARR_DIMS(v);
> ***************
> *** 1476,1481 ****
> --- 1616,1641 ----
> array = DatumGetArrayTypeP(PointerGetDatum(array));
>
> ndim = ARR_NDIM(array);
> +
> + /*
> + * if number of dims is zero, i.e. an empty array, create an array
> + * with nSubscripts dimensions, and set the lower bounds to the supplied
> + * subscripts
> + */
> + if (ndim == 0)
> + {
> + Oid elmtype = ARR_ELEMTYPE(array);
> +
> + for (i = 0; i < nSubscripts; i++)
> + {
> + dim[i] = 1;
> + lb[i] = indx[i];
> + }
> +
> + return construct_md_array(&dataValue, nSubscripts, dim, lb, elmtype,
> + elmlen, elmbyval, elmalign);
> + }
> +
> if (ndim != nSubscripts || ndim <= 0 || ndim > MAXDIM)
> elog(ERROR, "Invalid array subscripts");
>
> ***************
> *** 1632,1637 ****
> --- 1792,1822 ----
> /* note: we assume srcArray contains no toasted elements */
>
> ndim = ARR_NDIM(array);
> +
> + /*
> + * if number of dims is zero, i.e. an empty array, create an array
> + * with nSubscripts dimensions, and set the upper and lower bounds
> + * to the supplied subscripts
> + */
> + if (ndim == 0)
> + {
> + Datum *dvalues;
> + int nelems;
> + Oid elmtype = ARR_ELEMTYPE(array);
> +
> + deconstruct_array(srcArray, elmtype, elmlen, elmbyval, elmalign,
> + &dvalues, &nelems);
> +
> + for (i = 0; i < nSubscripts; i++)
> + {
> + dim[i] = 1 + upperIndx[i] - lowerIndx[i];
> + lb[i] = lowerIndx[i];
> + }
> +
> + return construct_md_array(dvalues, nSubscripts, dim, lb, elmtype,
> + elmlen, elmbyval, elmalign);
> + }
> +
> if (ndim < nSubscripts || ndim <= 0 || ndim > MAXDIM)
> elog(ERROR, "Invalid array subscripts");
>
> ***************
> *** 1811,1816 ****
> --- 1996,2008 ----
> Oid typelem;
> Oid proc;
> char *s;
> + typedef struct {
> + ArrayMetaState *inp_extra;
> + ArrayMetaState *ret_extra;
> + } am_extra;
> + am_extra *my_extra;
> + ArrayMetaState *inp_extra;
> + ArrayMetaState *ret_extra;
>
> /* Get input array */
> if (fcinfo->nargs < 1)
> ***************
> *** 1829,1839 ****
> if (nitems <= 0)
> PG_RETURN_ARRAYTYPE_P(v);
>
> ! /* Lookup source and result types. Unneeded variables are reused. */
> ! system_cache_lookup(inpType, IOFunc_input, &inp_typlen, &inp_typbyval,
> ! &typdelim, &typelem, &proc, &inp_typalign);
> ! system_cache_lookup(retType, IOFunc_input, &typlen, &typbyval,
> ! &typdelim, &typelem, &proc, &typalign);
>
> /* Allocate temporary array for new values */
> values = (Datum *) palloc(nitems * sizeof(Datum));
> --- 2021,2101 ----
> if (nitems <= 0)
> PG_RETURN_ARRAYTYPE_P(v);
>
> ! /*
> ! * We arrange to look up info about input and return element types only
> ! * once per series of calls, assuming the element type doesn't change
> ! * underneath us.
> ! */
> ! my_extra = (am_extra *) fcinfo->flinfo->fn_extra;
> ! if (my_extra == NULL)
> ! {
> ! fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(am_extra));
> ! my_extra = (am_extra *) fcinfo->flinfo->fn_extra;
> !
> ! my_extra->inp_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! inp_extra = my_extra->inp_extra;
> ! inp_extra->element_type = InvalidOid;
> !
> ! my_extra->ret_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> ! sizeof(ArrayMetaState));
> ! ret_extra = my_extra->ret_extra;
> ! ret_extra->element_type = InvalidOid;
> ! }
> ! else
> ! {
> ! inp_extra = my_extra->inp_extra;
> ! ret_extra = my_extra->ret_extra;
> ! }
> !
> ! if (inp_extra->element_type != inpType)
> ! {
> ! /* Lookup source and result types. Unneeded variables are reused. */
> ! get_type_metadata(inpType, IOFunc_input, &inp_typlen, &inp_typbyval,
> ! &typdelim, &typelem, &proc, &inp_typalign);
> !
> ! inp_extra->element_type = inpType;
> ! inp_extra->typlen = inp_typlen;
> ! inp_extra->typbyval = inp_typbyval;
> ! inp_extra->typdelim = typdelim;
> ! inp_extra->typelem = typelem;
> ! inp_extra->typiofunc = proc;
> ! inp_extra->typalign = inp_typalign;
> ! }
> ! else
> ! {
> ! inp_typlen = inp_extra->typlen;
> ! inp_typbyval = inp_extra->typbyval;
> ! typdelim = inp_extra->typdelim;
> ! typelem = inp_extra->typelem;
> ! proc = inp_extra->typiofunc;
> ! inp_typalign = inp_extra->typalign;
> ! }
> !
> ! if (ret_extra->element_type != retType)
> ! {
> ! /* Lookup source and result types. Unneeded variables are reused. */
> ! get_type_metadata(retType, IOFunc_input, &typlen, &typbyval,
> ! &typdelim, &typelem, &proc, &typalign);
> !
> ! ret_extra->element_type = retType;
> ! ret_extra->typlen = typlen;
> ! ret_extra->typbyval = typbyval;
> ! ret_extra->typdelim = typdelim;
> ! ret_extra->typelem = typelem;
> ! ret_extra->typiofunc = proc;
> ! ret_extra->typalign = typalign;
> ! }
> ! else
> ! {
> ! typlen = ret_extra->typlen;
> ! typbyval = ret_extra->typbyval;
> ! typdelim = ret_extra->typdelim;
> ! typelem = ret_extra->typelem;
> ! proc = ret_extra->typiofunc;
> ! typalign = ret_extra->typalign;
> ! }
>
> /* Allocate temporary array for new values */
> values = (Datum *) palloc(nitems * sizeof(Datum));
> ***************
> *** 2049,2056 ****
> * compares two arrays for equality
> * result :
> * returns true if the arrays are equal, false otherwise.
> - *
> - * XXX bitwise equality is pretty bogus ...
> *-----------------------------------------------------------------------------
> */
> Datum
> --- 2311,2316 ----
> ***************
> *** 2058,2069 ****
> {
> ArrayType *array1 = PG_GETARG_ARRAYTYPE_P(0);
> ArrayType *array2 = PG_GETARG_ARRAYTYPE_P(1);
> bool result = true;
>
> ! if (ARR_SIZE(array1) != ARR_SIZE(array2))
> ! result = false;
> ! else if (memcmp(array1, array2, ARR_SIZE(array1)) != 0)
> result = false;
>
> /* Avoid leaking memory when handed toasted input. */
> PG_FREE_IF_COPY(array1, 0);
> --- 2318,2435 ----
> {
> ArrayType *array1 = PG_GETARG_ARRAYTYPE_P(0);
> ArrayType *array2 = PG_GETARG_ARRAYTYPE_P(1);
> + char *p1 = (char *) ARR_DATA_PTR(array1);
> + char *p2 = (char *) ARR_DATA_PTR(array2);
> + int ndims1 = ARR_NDIM(array1);
> + int ndims2 = ARR_NDIM(array2);
> + int *dims1 = ARR_DIMS(array1);
> + int *dims2 = ARR_DIMS(array2);
> + int nitems1 = ArrayGetNItems(ndims1, dims1);
> + int nitems2 = ArrayGetNItems(ndims2, dims2);
> + Oid element_type = ARR_ELEMTYPE(array1);
> + FmgrInfo *ae_fmgr_info = fcinfo->flinfo;
> bool result = true;
> + int typlen;
> + bool typbyval;
> + char typdelim;
> + Oid typelem;
> + char typalign;
> + Oid typiofunc;
> + int i;
> + ArrayMetaState *my_extra;
> + FunctionCallInfoData locfcinfo;
>
> ! /* fast path if the arrays do not have the same number of elements */
> ! if (nitems1 != nitems2)
> result = false;
> + else
> + {
> + /*
> + * We arrange to look up the equality function only once per series of
> + * calls, assuming the element type doesn't change underneath us.
> + */
> + my_extra = (ArrayMetaState *) ae_fmgr_info->fn_extra;
> + if (my_extra == NULL)
> + {
> + ae_fmgr_info->fn_extra = MemoryContextAlloc(ae_fmgr_info->fn_mcxt,
> + sizeof(ArrayMetaState));
> + my_extra = (ArrayMetaState *) ae_fmgr_info->fn_extra;
> + my_extra->element_type = InvalidOid;
> + }
> +
> + if (my_extra->element_type != element_type)
> + {
> + Oid opfuncid = equality_oper_funcid(element_type);
> +
> + if (OidIsValid(opfuncid))
> + fmgr_info_cxt(opfuncid, &my_extra->proc, ae_fmgr_info->fn_mcxt);
> + else
> + elog(ERROR,
> + "array_eq: cannot find equality operator for type: %u",
> + element_type);
> +
> + get_type_metadata(element_type, IOFunc_output,
> + &typlen, &typbyval, &typdelim,
> + &typelem, &typiofunc, &typalign);
> +
> + my_extra->element_type = element_type;
> + my_extra->typlen = typlen;
> + my_extra->typbyval = typbyval;
> + my_extra->typdelim = typdelim;
> + my_extra->typelem = typelem;
> + my_extra->typiofunc = typiofunc;
> + my_extra->typalign = typalign;
> + }
> + else
> + {
> + typlen = my_extra->typlen;
> + typbyval = my_extra->typbyval;
> + typdelim = my_extra->typdelim;
> + typelem = my_extra->typelem;
> + typiofunc = my_extra->typiofunc;
> + typalign = my_extra->typalign;
> + }
> +
> + /*
> + * apply the operator to each pair of array elements.
> + */
> + MemSet(&locfcinfo, 0, sizeof(locfcinfo));
> + locfcinfo.flinfo = &my_extra->proc;
> + locfcinfo.nargs = 2;
> +
> + /* Loop over source data */
> + for (i = 0; i < nitems1; i++)
> + {
> + Datum elt1;
> + Datum elt2;
> + bool oprresult;
> +
> + /* Get element pair */
> + elt1 = fetch_att(p1, typbyval, typlen);
> + elt2 = fetch_att(p2, typbyval, typlen);
> +
> + p1 = att_addlength(p1, typlen, PointerGetDatum(p1));
> + p1 = (char *) att_align(p1, typalign);
> +
> + p2 = att_addlength(p2, typlen, PointerGetDatum(p2));
> + p2 = (char *) att_align(p2, typalign);
> +
> + /*
> + * Apply the operator to the element pair
> + */
> + locfcinfo.arg[0] = elt1;
> + locfcinfo.arg[1] = elt2;
> + locfcinfo.argnull[0] = false;
> + locfcinfo.argnull[1] = false;
> + locfcinfo.isnull = false;
> + oprresult = DatumGetBool(FunctionCallInvoke(&locfcinfo));
> + if (!oprresult)
> + {
> + result = false;
> + break;
> + }
> + }
> + }
>
> /* Avoid leaking memory when handed toasted input. */
> PG_FREE_IF_COPY(array1, 0);
> ***************
> *** 2073,2125 ****
> }
>
>
> ! /***************************************************************************/
> ! /******************| Support Routines |*****************/
> ! /***************************************************************************/
>
> ! static void
> ! system_cache_lookup(Oid element_type,
> ! IOFuncSelector which_func,
> ! int *typlen,
> ! bool *typbyval,
> ! char *typdelim,
> ! Oid *typelem,
> ! Oid *proc,
> ! char *typalign)
> ! {
> ! HeapTuple typeTuple;
> ! Form_pg_type typeStruct;
> !
> ! typeTuple = SearchSysCache(TYPEOID,
> ! ObjectIdGetDatum(element_type),
> ! 0, 0, 0);
> ! if (!HeapTupleIsValid(typeTuple))
> ! elog(ERROR, "cache lookup failed for type %u", element_type);
> ! typeStruct = (Form_pg_type) GETSTRUCT(typeTuple);
> !
> ! *typlen = typeStruct->typlen;
> ! *typbyval = typeStruct->typbyval;
> ! *typdelim = typeStruct->typdelim;
> ! *typelem = typeStruct->typelem;
> ! *typalign = typeStruct->typalign;
> ! switch (which_func)
> ! {
> ! case IOFunc_input:
> ! *proc = typeStruct->typinput;
> ! break;
> ! case IOFunc_output:
> ! *proc = typeStruct->typoutput;
> ! break;
> ! case IOFunc_receive:
> ! *proc = typeStruct->typreceive;
> ! break;
> ! case IOFunc_send:
> ! *proc = typeStruct->typsend;
> ! break;
> }
> ! ReleaseSysCache(typeTuple);
> }
>
> /*
> * Fetch array element at pointer, converted correctly to a Datum
> */
> --- 2439,2628 ----
> }
>
>
> ! /*-----------------------------------------------------------------------------
> ! * array-array bool operators:
> ! * Given two arrays, iterate comparison operators
> ! * over the array. Uses logic similar to text comparison
> ! * functions, except element-by-element instead of
> ! * character-by-character.
> ! *----------------------------------------------------------------------------
> ! */
> ! Datum
> ! array_ne(PG_FUNCTION_ARGS)
> ! {
> ! PG_RETURN_BOOL(!DatumGetBool(array_eq(fcinfo)));
> ! }
>
> ! Datum
> ! array_lt(PG_FUNCTION_ARGS)
> ! {
> ! PG_RETURN_BOOL(array_cmp(fcinfo) < 0);
> ! }
> !
> ! Datum
> ! array_gt(PG_FUNCTION_ARGS)
> ! {
> ! PG_RETURN_BOOL(array_cmp(fcinfo) > 0);
> ! }
> !
> ! Datum
> ! array_le(PG_FUNCTION_ARGS)
> ! {
> ! PG_RETURN_BOOL(array_cmp(fcinfo) <= 0);
> ! }
> !
> ! Datum
> ! array_ge(PG_FUNCTION_ARGS)
> ! {
> ! PG_RETURN_BOOL(array_cmp(fcinfo) >= 0);
> ! }
> !
> ! Datum
> ! btarraycmp(PG_FUNCTION_ARGS)
> ! {
> ! PG_RETURN_INT32(array_cmp(fcinfo));
> ! }
> !
> ! /*
> ! * array_cmp()
> ! * Internal comparison function for arrays.
> ! *
> ! * Returns -1, 0 or 1
> ! */
> ! static int
> ! array_cmp(FunctionCallInfo fcinfo)
> ! {
> ! ArrayType *array1 = PG_GETARG_ARRAYTYPE_P(0);
> ! ArrayType *array2 = PG_GETARG_ARRAYTYPE_P(1);
> ! FmgrInfo *ac_fmgr_info = fcinfo->flinfo;
> ! Datum opresult;
> ! int result = 0;
> ! Oid element_type = InvalidOid;
> ! int typlen;
> ! bool typbyval;
> ! char typdelim;
> ! Oid typelem;
> ! char typalign;
> ! Oid typiofunc;
> ! Datum *dvalues1;
> ! int nelems1;
> ! Datum *dvalues2;
> ! int nelems2;
> ! int min_nelems;
> ! int i;
> ! typedef struct
> ! {
> ! Oid element_type;
> ! int typlen;
> ! bool typbyval;
> ! char typdelim;
> ! Oid typelem;
> ! Oid typiofunc;
> ! char typalign;
> ! FmgrInfo eqproc;
> ! FmgrInfo ordproc;
> ! } ac_extra;
> ! ac_extra *my_extra;
> !
> ! element_type = ARR_ELEMTYPE(array1);
> !
> ! /*
> ! * We arrange to look up the element type operator function only once
> ! * per series of calls, assuming the element type and opname don't
> ! * change underneath us.
> ! */
> ! my_extra = (ac_extra *) ac_fmgr_info->fn_extra;
> ! if (my_extra == NULL)
> ! {
> ! ac_fmgr_info->fn_extra = MemoryContextAlloc(ac_fmgr_info->fn_mcxt,
> ! sizeof(ac_extra));
> ! my_extra = (ac_extra *) ac_fmgr_info->fn_extra;
> ! my_extra->element_type = InvalidOid;
> ! }
> !
> ! if (my_extra->element_type != element_type)
> ! {
> ! Oid eqfuncid = equality_oper_funcid(element_type);
> ! Oid ordfuncid = ordering_oper_funcid(element_type);
> !
> ! fmgr_info_cxt(eqfuncid, &my_extra->eqproc, ac_fmgr_info->fn_mcxt);
> ! fmgr_info_cxt(ordfuncid, &my_extra->ordproc, ac_fmgr_info->fn_mcxt);
> !
> ! if (my_extra->eqproc.fn_nargs != 2)
> ! elog(ERROR, "Equality operator does not take 2 arguments: %u",
> ! eqfuncid);
> ! if (my_extra->ordproc.fn_nargs != 2)
> ! elog(ERROR, "Ordering operator does not take 2 arguments: %u",
> ! ordfuncid);
> !
> ! get_type_metadata(element_type, IOFunc_output,
> ! &typlen, &typbyval, &typdelim,
> ! &typelem, &typiofunc, &typalign);
> !
> ! my_extra->element_type = element_type;
> ! my_extra->typlen = typlen;
> ! my_extra->typbyval = typbyval;
> ! my_extra->typdelim = typdelim;
> ! my_extra->typelem = typelem;
> ! my_extra->typiofunc = InvalidOid;
> ! my_extra->typalign = typalign;
> ! }
> ! else
> ! {
> ! typlen = my_extra->typlen;
> ! typbyval = my_extra->typbyval;
> ! typalign = my_extra->typalign;
> ! }
> !
> ! /* extract a C array of arg array datums */
> ! deconstruct_array(array1, element_type, typlen, typbyval, typalign,
> ! &dvalues1, &nelems1);
> !
> ! deconstruct_array(array2, element_type, typlen, typbyval, typalign,
> ! &dvalues2, &nelems2);
> !
> ! min_nelems = Min(nelems1, nelems2);
> ! for (i = 0; i < min_nelems; i++)
> ! {
> ! /* are they equal */
> ! opresult = FunctionCall2(&my_extra->eqproc,
> ! dvalues1[i], dvalues2[i]);
> !
> ! if (!DatumGetBool(opresult))
> ! {
> ! /* nope, see if arg1 is less than arg2 */
> ! opresult = FunctionCall2(&my_extra->ordproc,
> ! dvalues1[i], dvalues2[i]);
> ! if (DatumGetBool(opresult))
> ! {
> ! /* arg1 is less than arg2 */
> ! result = -1;
> ! break;
> ! }
> ! else
> ! {
> ! /* arg1 is greater than arg2 */
> ! result = 1;
> ! break;
> ! }
> ! }
> }
> !
> ! if ((result == 0) && (nelems1 != nelems2))
> ! result = (nelems1 < nelems2) ? -1 : 1;
> !
> ! /* Avoid leaking memory when handed toasted input. */
> ! PG_FREE_IF_COPY(array1, 0);
> ! PG_FREE_IF_COPY(array2, 1);
> !
> ! return result;
> }
>
> +
> + /***************************************************************************/
> + /******************| Support Routines |*****************/
> + /***************************************************************************/
> +
> /*
> * Fetch array element at pointer, converted correctly to a Datum
> */
> ***************
> *** 2423,2428 ****
> --- 2926,2943 ----
> if (tgt_elem_type == InvalidOid)
> elog(ERROR, "Target type is not an array");
>
> + /*
> + * We don't deal with domain constraints yet, so bail out.
> + * This isn't currently a problem, because we also don't
> + * support arrays of domain type elements either. But in the
> + * future we might. At that point consideration should be given
> + * to removing the check below and adding a domain constraints
> + * check to the coercion.
> + */
> + if (getBaseType(tgt_elem_type) != tgt_elem_type)
> + elog(ERROR, "array coercion to domain type elements not " \
> + "currently supported");
> +
> if (!find_coercion_pathway(tgt_elem_type, src_elem_type,
> COERCION_EXPLICIT, &funcId))
> {
> ***************
> *** 2439,2448 ****
> }
>
> /*
> ! * If it's binary-compatible, return the array unmodified.
> */
> if (my_extra->coerce_finfo.fn_oid == InvalidOid)
> ! PG_RETURN_ARRAYTYPE_P(src);
>
> /*
> * Use array_map to apply the function to each array element.
> --- 2954,2969 ----
> }
>
> /*
> ! * If it's binary-compatible, modify the element type in the array header,
> ! * but otherwise leave the array as we received it.
> */
> if (my_extra->coerce_finfo.fn_oid == InvalidOid)
> ! {
> ! ArrayType *result = DatumGetArrayTypePCopy(PG_GETARG_DATUM(0));
> !
> ! ARR_ELEMTYPE(result) = my_extra->desttype;
> ! PG_RETURN_ARRAYTYPE_P(result);
> ! }
>
> /*
> * Use array_map to apply the function to each array element.
> ***************
> *** 2453,2456 ****
> --- 2974,3092 ----
> locfcinfo.arg[0] = PointerGetDatum(src);
>
> return array_map(&locfcinfo, my_extra->srctype, my_extra->desttype);
> + }
> +
> + /*
> + * accumArrayResult - accumulate one (more) Datum for an ARRAY_SUBLINK
> + *
> + * astate is working state (NULL on first call)
> + * rcontext is where to keep working state
> + */
> + ArrayBuildState *
> + accumArrayResult(ArrayBuildState *astate,
> + Datum dvalue, bool disnull,
> + Oid element_type,
> + MemoryContext rcontext)
> + {
> + MemoryContext arr_context,
> + oldcontext;
> +
> + if (astate == NULL)
> + {
> + /* First time through --- initialize */
> +
> + /* Make a temporary context to hold all the junk */
> + arr_context = AllocSetContextCreate(rcontext,
> + "accumArrayResult",
> + ALLOCSET_DEFAULT_MINSIZE,
> + ALLOCSET_DEFAULT_INITSIZE,
> + ALLOCSET_DEFAULT_MAXSIZE);
> + oldcontext = MemoryContextSwitchTo(arr_context);
> + astate = (ArrayBuildState *) palloc(sizeof(ArrayBuildState));
> + astate->mcontext = arr_context;
> + astate->dvalues = (Datum *)
> + palloc(ARRAY_ELEMS_CHUNKSIZE * sizeof(Datum));
> + astate->nelems = 0;
> + astate->element_type = element_type;
> + get_typlenbyvalalign(element_type,
> + &astate->typlen,
> + &astate->typbyval,
> + &astate->typalign);
> + }
> + else
> + {
> + oldcontext = MemoryContextSwitchTo(astate->mcontext);
> + Assert(astate->element_type == element_type);
> + /* enlarge dvalues[] if needed */
> + if ((astate->nelems % ARRAY_ELEMS_CHUNKSIZE) == 0)
> + astate->dvalues = (Datum *)
> + repalloc(astate->dvalues,
> + (astate->nelems + ARRAY_ELEMS_CHUNKSIZE) * sizeof(Datum));
> + }
> +
> + if (disnull)
> + elog(ERROR, "NULL elements not allowed in Arrays");
> +
> + /* Use datumCopy to ensure pass-by-ref stuff is copied into mcontext */
> + astate->dvalues[astate->nelems++] =
> + datumCopy(dvalue, astate->typbyval, astate->typlen);
> +
> + MemoryContextSwitchTo(oldcontext);
> +
> + return astate;
> + }
> +
> + /*
> + * makeArrayResult - produce final result of accumArrayResult
> + *
> + * astate is working state (not NULL)
> + * rcontext is where to construct result
> + */
> + Datum
> + makeArrayResult(ArrayBuildState *astate,
> + MemoryContext rcontext)
> + {
> + int dims[1];
> + int lbs[1];
> +
> + dims[0] = astate->nelems;
> + lbs[0] = 1;
> +
> + return makeMdArrayResult(astate, 1, dims, lbs, rcontext);
> + }
> +
> + /*
> + * makeMdArrayResult - produce md final result of accumArrayResult
> + *
> + * astate is working state (not NULL)
> + * rcontext is where to construct result
> + */
> + Datum
> + makeMdArrayResult(ArrayBuildState *astate,
> + int ndims,
> + int *dims,
> + int *lbs,
> + MemoryContext rcontext)
> + {
> + ArrayType *result;
> + MemoryContext oldcontext;
> +
> + /* Build the final array result in rcontext */
> + oldcontext = MemoryContextSwitchTo(rcontext);
> +
> + result = construct_md_array(astate->dvalues,
> + ndims,
> + dims,
> + lbs,
> + astate->element_type,
> + astate->typlen,
> + astate->typbyval,
> + astate->typalign);
> +
> + MemoryContextSwitchTo(oldcontext);
> +
> + /* Clean up all the junk */
> + MemoryContextDelete(astate->mcontext);
> +
> + return PointerGetDatum(result);
> }
> Index: src/backend/utils/adt/varlena.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/utils/adt/varlena.c,v
> retrieving revision 1.98
> diff -c -r1.98 varlena.c
> *** src/backend/utils/adt/varlena.c 15 May 2003 15:50:19 -0000 1.98
> --- src/backend/utils/adt/varlena.c 1 Jun 2003 15:41:15 -0000
> ***************
> *** 19,29 ****
> --- 19,32 ----
> #include "mb/pg_wchar.h"
> #include "miscadmin.h"
> #include "access/tuptoaster.h"
> + #include "catalog/pg_type.h"
> #include "lib/stringinfo.h"
> #include "libpq/crypt.h"
> #include "libpq/pqformat.h"
> + #include "utils/array.h"
> #include "utils/builtins.h"
> #include "utils/pg_locale.h"
> + #include "utils/lsyscache.h"
>
>
> typedef struct varlena unknown;
> ***************
> *** 1983,1990 ****
> if (fldnum == 1) /* first field - just return the input
> * string */
> PG_RETURN_TEXT_P(inputstring);
> ! else
> ! /* otherwise return an empty string */
> PG_RETURN_TEXT_P(PG_STR_GET_TEXT(""));
> }
>
> --- 1986,1992 ----
> if (fldnum == 1) /* first field - just return the input
> * string */
> PG_RETURN_TEXT_P(inputstring);
> ! else /* otherwise return an empty string */
> PG_RETURN_TEXT_P(PG_STR_GET_TEXT(""));
> }
>
> ***************
> *** 2004,2011 ****
> if (fldnum == 1) /* first field - just return the input
> * string */
> PG_RETURN_TEXT_P(inputstring);
> ! else
> ! /* otherwise return an empty string */
> PG_RETURN_TEXT_P(PG_STR_GET_TEXT(""));
> }
> else if ((start_posn != 0) && (end_posn == 0))
> --- 2006,2012 ----
> if (fldnum == 1) /* first field - just return the input
> * string */
> PG_RETURN_TEXT_P(inputstring);
> ! else /* otherwise return an empty string */
> PG_RETURN_TEXT_P(PG_STR_GET_TEXT(""));
> }
> else if ((start_posn != 0) && (end_posn == 0))
> ***************
> *** 2026,2031 ****
> --- 2027,2217 ----
> result_text = text_substring(PointerGetDatum(inputstring), start_posn + fldsep_len, end_posn - start_posn - fldsep_len, false);
> PG_RETURN_TEXT_P(result_text);
> }
> + }
> +
> + /*
> + * text_to_array
> + * parse input string
> + * return text array of elements
> + * based on provided field separator
> + */
> + Datum
> + text_to_array(PG_FUNCTION_ARGS)
> + {
> + text *inputstring = PG_GETARG_TEXT_P(0);
> + int inputstring_len = TEXTLEN(inputstring);
> + text *fldsep = PG_GETARG_TEXT_P(1);
> + int fldsep_len = TEXTLEN(fldsep);
> + int fldnum;
> + int start_posn = 0;
> + int end_posn = 0;
> + text *result_text = NULL;
> + ArrayBuildState *astate = NULL;
> + MemoryContext oldcontext = CurrentMemoryContext;
> +
> + /* return NULL for empty input string */
> + if (inputstring_len < 1)
> + PG_RETURN_NULL();
> +
> + /* empty field separator
> + * return one element, 1D, array using the input string */
> + if (fldsep_len < 1)
> + PG_RETURN_ARRAYTYPE_P(create_singleton_array(fcinfo, TEXTOID,
> + CStringGetDatum(inputstring), 1));
> +
> + /* start with end position holding the initial start position */
> + end_posn = 0;
> + for (fldnum=1;;fldnum++) /* field number is 1 based */
> + {
> + Datum dvalue;
> + bool disnull = false;
> +
> + start_posn = end_posn;
> + end_posn = text_position(PointerGetDatum(inputstring),
> + PointerGetDatum(fldsep),
> + fldnum);
> +
> + if ((start_posn == 0) && (end_posn == 0)) /* fldsep not found */
> + {
> + if (fldnum == 1)
> + {
> + /* first element
> + * return one element, 1D, array using the input string */
> + PG_RETURN_ARRAYTYPE_P(create_singleton_array(fcinfo, TEXTOID,
> + CStringGetDatum(inputstring), 1));
> + }
> + else
> + {
> + /* otherwise create array and exit */
> + PG_RETURN_ARRAYTYPE_P(makeArrayResult(astate, oldcontext));
> + }
> + }
> + else if ((start_posn != 0) && (end_posn == 0))
> + {
> + /* last field requested */
> + result_text = text_substring(PointerGetDatum(inputstring), start_posn + fldsep_len, -1, true);
> + }
> + else if ((start_posn == 0) && (end_posn != 0))
> + {
> + /* first field requested */
> + result_text = LEFT(inputstring, fldsep);
> + }
> + else
> + {
> + /* prior to last field requested */
> + result_text = text_substring(PointerGetDatum(inputstring), start_posn + fldsep_len, end_posn - start_posn - fldsep_len, false);
> + }
> +
> + /* stash away current value */
> + dvalue = PointerGetDatum(result_text);
> + astate = accumArrayResult(astate, dvalue,
> + disnull, TEXTOID, oldcontext);
> +
> + }
> +
> + /* never reached -- keep compiler quiet */
> + PG_RETURN_NULL();
> + }
> +
> + /*
> + * array_to_text
> + * concatenate Cstring representation of input array elements
> + * using provided field separator
> + */
> + Datum
> + array_to_text(PG_FUNCTION_ARGS)
> + {
> + ArrayType *v = PG_GETARG_ARRAYTYPE_P(0);
> + char *fldsep = PG_TEXTARG_GET_STR(1);
> + int nitems, *dims, ndims;
> + char *p;
> + Oid element_type;
> + int typlen;
> + bool typbyval;
> + char typdelim;
> + Oid typoutput,
> + typelem;
> + FmgrInfo outputproc;
> + char typalign;
> + StringInfo result_str = makeStringInfo();
> + int i;
> + ArrayMetaState *my_extra;
> +
> + p = ARR_DATA_PTR(v);
> + ndims = ARR_NDIM(v);
> + dims = ARR_DIMS(v);
> + nitems = ArrayGetNItems(ndims, dims);
> +
> + /* if there are no elements, return an empty string */
> + if (nitems == 0)
> + PG_RETURN_TEXT_P(PG_STR_GET_TEXT(""));
> +
> + element_type = ARR_ELEMTYPE(v);
> +
> + /*
> + * We arrange to look up info about element type, including its output
> + * conversion proc only once per series of calls, assuming the element
> + * type doesn't change underneath us.
> + */
> + my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> + if (my_extra == NULL)
> + {
> + fcinfo->flinfo->fn_extra = MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
> + sizeof(ArrayMetaState));
> + my_extra = (ArrayMetaState *) fcinfo->flinfo->fn_extra;
> + my_extra->element_type = InvalidOid;
> + }
> +
> + if (my_extra->element_type != element_type)
> + {
> + /* Get info about element type, including its output conversion proc */
> + get_type_metadata(element_type, IOFunc_output,
> + &typlen, &typbyval, &typdelim,
> + &typelem, &typoutput, &typalign);
> + fmgr_info(typoutput, &outputproc);
> +
> + my_extra->element_type = element_type;
> + my_extra->typlen = typlen;
> + my_extra->typbyval = typbyval;
> + my_extra->typdelim = typdelim;
> + my_extra->typelem = typelem;
> + my_extra->typiofunc = typoutput;
> + my_extra->typalign = typalign;
> + my_extra->proc = outputproc;
> + }
> + else
> + {
> + typlen = my_extra->typlen;
> + typbyval = my_extra->typbyval;
> + typdelim = my_extra->typdelim;
> + typelem = my_extra->typelem;
> + typoutput = my_extra->typiofunc;
> + typalign = my_extra->typalign;
> + outputproc = my_extra->proc;
> + }
> +
> + for (i = 0; i < nitems; i++)
> + {
> + Datum itemvalue;
> + char *value;
> +
> + itemvalue = fetch_att(p, typbyval, typlen);
> +
> + value = DatumGetCString(FunctionCall3(&outputproc,
> + itemvalue,
> + ObjectIdGetDatum(typelem),
> + Int32GetDatum(-1)));
> +
> + if (i > 0)
> + appendStringInfo(result_str, "%s%s", fldsep, value);
> + else
> + appendStringInfo(result_str, "%s", value);
> +
> + p = att_addlength(p, typlen, PointerGetDatum(p));
> + p = (char *) att_align(p, typalign);
> + }
> +
> + PG_RETURN_TEXT_P(PG_STR_GET_TEXT(result_str->data));
> }
>
> #define HEXBASE 16
> Index: src/backend/utils/cache/lsyscache.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/utils/cache/lsyscache.c,v
> retrieving revision 1.95
> diff -c -r1.95 lsyscache.c
> *** src/backend/utils/cache/lsyscache.c 26 May 2003 00:11:27 -0000 1.95
> --- src/backend/utils/cache/lsyscache.c 1 Jun 2003 15:41:15 -0000
> ***************
> *** 625,630 ****
> --- 625,664 ----
> }
>
> /*
> + * get_func_argtypes
> + * Given procedure id, return the function's argument types.
> + * Also pass back the number of arguments.
> + */
> + Oid *
> + get_func_argtypes(Oid funcid, int *nargs)
> + {
> + HeapTuple tp;
> + Form_pg_proc procstruct;
> + Oid *result = NULL;
> + int i;
> +
> + tp = SearchSysCache(PROCOID,
> + ObjectIdGetDatum(funcid),
> + 0, 0, 0);
> + if (!HeapTupleIsValid(tp))
> + elog(ERROR, "Function OID %u does not exist", funcid);
> +
> + procstruct = (Form_pg_proc) GETSTRUCT(tp);
> + *nargs = (int) procstruct->pronargs;
> +
> + if (*nargs > 0)
> + {
> + result = (Oid *) palloc(*nargs * sizeof(Oid));
> +
> + for (i = 0; i < *nargs; i++)
> + result[i] = procstruct->proargtypes[i];
> + }
> +
> + ReleaseSysCache(tp);
> + return result;
> + }
> +
> + /*
> * get_func_retset
> * Given procedure id, return the function's proretset flag.
> */
> ***************
> *** 994,999 ****
> --- 1028,1083 ----
> *typbyval = typtup->typbyval;
> *typalign = typtup->typalign;
> ReleaseSysCache(tp);
> + }
> +
> + /*
> + * get_type_metadata
> + *
> + * A six-fer: given the type OID, return typlen, typbyval, typalign,
> + * typdelim, typelem, IO function Oid. The IO function
> + * returned is controlled by IOFuncSelector
> + */
> + void
> + get_type_metadata(Oid element_type,
> + IOFuncSelector which_func,
> + int *typlen,
> + bool *typbyval,
> + char *typdelim,
> + Oid *typelem,
> + Oid *proc,
> + char *typalign)
> + {
> + HeapTuple typeTuple;
> + Form_pg_type typeStruct;
> +
> + typeTuple = SearchSysCache(TYPEOID,
> + ObjectIdGetDatum(element_type),
> + 0, 0, 0);
> + if (!HeapTupleIsValid(typeTuple))
> + elog(ERROR, "cache lookup failed for type %u", element_type);
> + typeStruct = (Form_pg_type) GETSTRUCT(typeTuple);
> +
> + *typlen = typeStruct->typlen;
> + *typbyval = typeStruct->typbyval;
> + *typdelim = typeStruct->typdelim;
> + *typelem = typeStruct->typelem;
> + *typalign = typeStruct->typalign;
> + switch (which_func)
> + {
> + case IOFunc_input:
> + *proc = typeStruct->typinput;
> + break;
> + case IOFunc_output:
> + *proc = typeStruct->typoutput;
> + break;
> + case IOFunc_receive:
> + *proc = typeStruct->typreceive;
> + break;
> + case IOFunc_send:
> + *proc = typeStruct->typsend;
> + break;
> + }
> + ReleaseSysCache(typeTuple);
> }
>
> #ifdef NOT_USED
> Index: src/backend/utils/fmgr/fmgr.c
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/backend/utils/fmgr/fmgr.c,v
> retrieving revision 1.68
> diff -c -r1.68 fmgr.c
> *** src/backend/utils/fmgr/fmgr.c 8 Apr 2003 23:20:02 -0000 1.68
> --- src/backend/utils/fmgr/fmgr.c 1 Jun 2003 15:41:15 -0000
> ***************
> *** 1673,1675 ****
> --- 1673,1701 ----
>
> return exprType((Node *) nth(argnum, args));
> }
> +
> + /*
> + * Get the OID of the function or operator
> + *
> + * Returns InvalidOid if information is not available
> + */
> + Oid
> + get_fn_expr_functype(FunctionCallInfo fcinfo)
> + {
> + Node *expr;
> +
> + /*
> + * can't return anything useful if we have no FmgrInfo or if
> + * its fn_expr node has not been initialized
> + */
> + if (!fcinfo || !fcinfo->flinfo || !fcinfo->flinfo->fn_expr)
> + return InvalidOid;
> +
> + expr = fcinfo->flinfo->fn_expr;
> + if (IsA(expr, FuncExpr))
> + return ((FuncExpr *) expr)->funcid;
> + else if (IsA(expr, OpExpr))
> + return ((OpExpr *) expr)->opno;
> + else
> + return InvalidOid;
> + }
> Index: src/include/fmgr.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/fmgr.h,v
> retrieving revision 1.27
> diff -c -r1.27 fmgr.h
> *** src/include/fmgr.h 8 Apr 2003 23:20:04 -0000 1.27
> --- src/include/fmgr.h 1 Jun 2003 15:41:15 -0000
> ***************
> *** 18,23 ****
> --- 18,24 ----
> #ifndef FMGR_H
> #define FMGR_H
>
> + #include "nodes/nodes.h"
>
> /*
> * All functions that can be called directly by fmgr must have this signature.
> ***************
> *** 372,385 ****
> Datum arg6, Datum arg7, Datum arg8,
> Datum arg9);
>
> -
> /*
> * Routines in fmgr.c
> */
> extern Pg_finfo_record *fetch_finfo_record(void *filehandle, char *funcname);
> ! extern Oid fmgr_internal_function(const char *proname);
> ! extern Oid get_fn_expr_rettype(FunctionCallInfo fcinfo);
> ! extern Oid get_fn_expr_argtype(FunctionCallInfo fcinfo, int argnum);
>
> /*
> * Routines in dfmgr.c
> --- 373,386 ----
> Datum arg6, Datum arg7, Datum arg8,
> Datum arg9);
>
> /*
> * Routines in fmgr.c
> */
> extern Pg_finfo_record *fetch_finfo_record(void *filehandle, char *funcname);
> ! extern Oid fmgr_internal_function(const char *proname);
> ! extern Oid get_fn_expr_rettype(FunctionCallInfo fcinfo);
> ! extern Oid get_fn_expr_argtype(FunctionCallInfo fcinfo, int argnum);
> ! extern Oid get_fn_expr_functype(FunctionCallInfo fcinfo);
>
> /*
> * Routines in dfmgr.c
> Index: src/include/catalog/pg_amop.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/catalog/pg_amop.h,v
> retrieving revision 1.49
> diff -c -r1.49 pg_amop.h
> *** src/include/catalog/pg_amop.h 26 May 2003 00:11:27 -0000 1.49
> --- src/include/catalog/pg_amop.h 2 Jun 2003 02:00:05 -0000
> ***************
> *** 418,423 ****
> --- 418,432 ----
> DATA(insert ( 2098 4 f 2335 ));
> DATA(insert ( 2098 5 f 2336 ));
>
> + /*
> + * btree array_ops
> + */
> +
> + DATA(insert ( 397 1 f 1072 ));
> + DATA(insert ( 397 2 f 1074 ));
> + DATA(insert ( 397 3 f 1070 ));
> + DATA(insert ( 397 4 f 1075 ));
> + DATA(insert ( 397 5 f 1073 ));
>
> /*
> * hash index _ops
> Index: src/include/catalog/pg_amproc.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/catalog/pg_amproc.h,v
> retrieving revision 1.37
> diff -c -r1.37 pg_amproc.h
> *** src/include/catalog/pg_amproc.h 26 May 2003 00:11:27 -0000 1.37
> --- src/include/catalog/pg_amproc.h 2 Jun 2003 02:25:34 -0000
> ***************
> *** 78,83 ****
> --- 78,84 ----
>
>
> /* btree */
> + DATA(insert ( 397 1 398 ));
> DATA(insert ( 421 1 357 ));
> DATA(insert ( 423 1 1596 ));
> DATA(insert ( 424 1 1693 ));
> Index: src/include/catalog/pg_opclass.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/catalog/pg_opclass.h,v
> retrieving revision 1.50
> diff -c -r1.50 pg_opclass.h
> *** src/include/catalog/pg_opclass.h 28 May 2003 16:04:00 -0000 1.50
> --- src/include/catalog/pg_opclass.h 2 Jun 2003 01:47:07 -0000
> ***************
> *** 87,92 ****
> --- 87,93 ----
> */
>
> DATA(insert OID = 421 ( 403 abstime_ops PGNSP PGUID 702 t 0 ));
> + DATA(insert OID = 397 ( 403 array_ops PGNSP PGUID 2277 t 0 ));
> DATA(insert OID = 422 ( 402 bigbox_ops PGNSP PGUID 603 f 0 ));
> DATA(insert OID = 423 ( 403 bit_ops PGNSP PGUID 1560 t 0 ));
> DATA(insert OID = 424 ( 403 bool_ops PGNSP PGUID 16 t 0 ));
> Index: src/include/catalog/pg_operator.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/catalog/pg_operator.h,v
> retrieving revision 1.114
> diff -c -r1.114 pg_operator.h
> *** src/include/catalog/pg_operator.h 26 May 2003 00:11:27 -0000 1.114
> --- src/include/catalog/pg_operator.h 1 Jun 2003 23:10:35 -0000
> ***************
> *** 116,125 ****
> DATA(insert OID = 97 ( "<" PGNSP PGUID b f 23 23 16 521 525 0 0 0 0 int4lt scalarltsel scalarltjoinsel ));
> DATA(insert OID = 98 ( "=" PGNSP PGUID b t 25 25 16 98 531 664 664 664 666 texteq eqsel eqjoinsel ));
>
> ! DATA(insert OID = 329 ( "=" PGNSP PGUID b f 2277 2277 16 329 0 0 0 0 0 array_eq eqsel eqjoinsel ));
> ! DATA(insert OID = 349 ( "||" PGNSP PGUID b f 2277 2283 2277 0 0 0 0 0 0 array_append - - ));
> ! DATA(insert OID = 374 ( "||" PGNSP PGUID b f 2283 2277 2277 0 0 0 0 0 0 array_prepend - - ));
> ! DATA(insert OID = 375 ( "||" PGNSP PGUID b f 2277 2277 2277 0 0 0 0 0 0 array_cat - - ));
>
> DATA(insert OID = 352 ( "=" PGNSP PGUID b t 28 28 16 352 0 0 0 0 0 xideq eqsel eqjoinsel ));
> DATA(insert OID = 353 ( "=" PGNSP PGUID b t 28 23 16 0 0 0 0 0 0 xideqint4 eqsel eqjoinsel ));
> --- 116,130 ----
> DATA(insert OID = 97 ( "<" PGNSP PGUID b f 23 23 16 521 525 0 0 0 0 int4lt scalarltsel scalarltjoinsel ));
> DATA(insert OID = 98 ( "=" PGNSP PGUID b t 25 25 16 98 531 664 664 664 666 texteq eqsel eqjoinsel ));
>
> ! DATA(insert OID = 1070 ( "=" PGNSP PGUID b f 2277 2277 16 1070 1071 1072 1072 1072 1073 array_eq eqsel eqjoinsel ));
> ! DATA(insert OID = 1071 ( "<>" PGNSP PGUID b f 2277 2277 16 1071 1070 0 0 0 0 array_ne neqsel neqjoinsel ));
> ! DATA(insert OID = 1072 ( "<" PGNSP PGUID b f 2277 2277 16 1073 1075 0 0 0 0 array_lt scalarltsel scalarltjoinsel ));
> ! DATA(insert OID = 1073 ( ">" PGNSP PGUID b f 2277 2277 16 1072 1074 0 0 0 0 array_gt scalargtsel scalargtjoinsel ));
> ! DATA(insert OID = 1074 ( "<=" PGNSP PGUID b f 2277 2277 16 1075 1073 0 0 0 0 array_le scalarltsel scalarltjoinsel ));
> ! DATA(insert OID = 1075 ( ">=" PGNSP PGUID b f 2277 2277 16 1074 1072 0 0 0 0 array_ge scalargtsel scalargtjoinsel ));
> ! DATA(insert OID = 349 ( "||" PGNSP PGUID b f 2277 2283 2277 0 0 0 0 0 0 array_append - - ));
> ! DATA(insert OID = 374 ( "||" PGNSP PGUID b f 2283 2277 2277 0 0 0 0 0 0 array_prepend - - ));
> ! DATA(insert OID = 375 ( "||" PGNSP PGUID b f 2277 2277 2277 0 0 0 0 0 0 array_cat - - ));
>
> DATA(insert OID = 352 ( "=" PGNSP PGUID b t 28 28 16 352 0 0 0 0 0 xideq eqsel eqjoinsel ));
> DATA(insert OID = 353 ( "=" PGNSP PGUID b t 28 23 16 0 0 0 0 0 0 xideqint4 eqsel eqjoinsel ));
> ***************
> *** 425,430 ****
> --- 430,436 ----
> DATA(insert OID = 966 ( "+" PGNSP PGUID b f 1034 1033 1034 0 0 0 0 0 0 aclinsert - - ));
> DATA(insert OID = 967 ( "-" PGNSP PGUID b f 1034 1033 1034 0 0 0 0 0 0 aclremove - - ));
> DATA(insert OID = 968 ( "~" PGNSP PGUID b f 1034 1033 16 0 0 0 0 0 0 aclcontains - - ));
> + DATA(insert OID = 974 ( "=" PGNSP PGUID b f 1033 1033 16 0 0 0 0 0 0 aclitemeq eqsel eqjoinsel ));
>
> /* additional geometric operators - thomas 1997-07-09 */
> DATA(insert OID = 969 ( "@@" PGNSP PGUID l f 0 601 600 0 0 0 0 0 0 lseg_center - - ));
> Index: src/include/catalog/pg_proc.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/catalog/pg_proc.h,v
> retrieving revision 1.302
> diff -c -r1.302 pg_proc.h
> *** src/include/catalog/pg_proc.h 26 May 2003 00:11:27 -0000 1.302
> --- src/include/catalog/pg_proc.h 2 Jun 2003 02:21:05 -0000
> ***************
> *** 758,763 ****
> --- 758,765 ----
> DESCR("btree less-equal-greater");
> DATA(insert OID = 360 ( bttextcmp PGNSP PGUID 12 f f t f i 2 23 "25 25" bttextcmp - _null_ ));
> DESCR("btree less-equal-greater");
> + DATA(insert OID = 398 ( btarraycmp PGNSP PGUID 12 f f t f i 2 23 "2277 2277" btarraycmp - _null_ ));
> + DESCR("btree less-equal-greater");
>
> DATA(insert OID = 361 ( lseg_distance PGNSP PGUID 12 f f t f i 2 701 "601 601" lseg_distance - _null_ ));
> DESCR("distance between");
> ***************
> *** 984,997 ****
> DATA(insert OID = 743 ( text_ge PGNSP PGUID 12 f f t f i 2 16 "25 25" text_ge - _null_ ));
> DESCR("greater-than-or-equal");
>
> - DATA(insert OID = 744 ( array_eq PGNSP PGUID 12 f f t f i 2 16 "2277 2277" array_eq - _null_ ));
> - DESCR("array equal");
> -
> DATA(insert OID = 745 ( current_user PGNSP PGUID 12 f f t f s 0 19 "" current_user - _null_ ));
> DESCR("current user name");
> DATA(insert OID = 746 ( session_user PGNSP PGUID 12 f f t f s 0 19 "" session_user - _null_ ));
> DESCR("session user name");
>
> DATA(insert OID = 747 ( array_dims PGNSP PGUID 12 f f t f i 1 25 "2277" array_dims - _null_ ));
> DESCR("array dimensions");
> DATA(insert OID = 750 ( array_in PGNSP PGUID 12 f f t f s 3 2277 "2275 26 23" array_in - _null_ ));
> --- 986,1008 ----
> DATA(insert OID = 743 ( text_ge PGNSP PGUID 12 f f t f i 2 16 "25 25" text_ge - _null_ ));
> DESCR("greater-than-or-equal");
>
> DATA(insert OID = 745 ( current_user PGNSP PGUID 12 f f t f s 0 19 "" current_user - _null_ ));
> DESCR("current user name");
> DATA(insert OID = 746 ( session_user PGNSP PGUID 12 f f t f s 0 19 "" session_user - _null_ ));
> DESCR("session user name");
>
> + DATA(insert OID = 744 ( array_eq PGNSP PGUID 12 f f t f i 2 16 "2277 2277" array_eq - _null_ ));
> + DESCR("array equal");
> + DATA(insert OID = 390 ( array_ne PGNSP PGUID 12 f f t f i 2 16 "2277 2277" array_ne - _null_ ));
> + DESCR("array not equal");
> + DATA(insert OID = 391 ( array_lt PGNSP PGUID 12 f f t f i 2 16 "2277 2277" array_lt - _null_ ));
> + DESCR("array less than");
> + DATA(insert OID = 392 ( array_gt PGNSP PGUID 12 f f t f i 2 16 "2277 2277" array_gt - _null_ ));
> + DESCR("array greater than");
> + DATA(insert OID = 393 ( array_le PGNSP PGUID 12 f f t f i 2 16 "2277 2277" array_le - _null_ ));
> + DESCR("array less than or equal");
> + DATA(insert OID = 396 ( array_ge PGNSP PGUID 12 f f t f i 2 16 "2277 2277" array_ge - _null_ ));
> + DESCR("array greater than or equal");
> DATA(insert OID = 747 ( array_dims PGNSP PGUID 12 f f t f i 1 25 "2277" array_dims - _null_ ));
> DESCR("array dimensions");
> DATA(insert OID = 750 ( array_in PGNSP PGUID 12 f f t f s 3 2277 "2275 26 23" array_in - _null_ ));
> ***************
> *** 1008,1015 ****
> DESCR("append element onto end of array");
> DATA(insert OID = 379 ( array_prepend PGNSP PGUID 12 f f t f i 2 2277 "2283 2277" array_push - _null_ ));
> DESCR("prepend element onto front of array");
> - DATA(insert OID = 380 ( array_accum PGNSP PGUID 12 f f f f i 2 2277 "2277 2283" array_accum - _null_ ));
> - DESCR("push element onto end of array, creating array if needed");
> DATA(insert OID = 381 ( array_assign PGNSP PGUID 12 f f t f i 3 2277 "2277 23 2283" array_assign - _null_ ));
> DESCR("assign specific array element");
> DATA(insert OID = 382 ( array_subscript PGNSP PGUID 12 f f t f i 2 2283 "2277 23" array_subscript - _null_ ));
> --- 1019,1024 ----
> ***************
> *** 1018,1023 ****
> --- 1027,1036 ----
> DESCR("concatenate two arrays");
> DATA(insert OID = 384 ( array_coerce PGNSP PGUID 12 f f t f i 1 2277 "2277" array_type_coerce - _null_ ));
> DESCR("coerce array type to another array type");
> + DATA(insert OID = 394 ( str_to_array PGNSP PGUID 12 f f t f i 2 1009 "25 25" text_to_array - _null_ ));
> + DESCR("split delimited text into text[]");
> + DATA(insert OID = 395 ( array_to_str PGNSP PGUID 12 f f t f i 2 25 "2277 25" array_to_text - _null_ ));
> + DESCR("concatenate array elements, using delimiter, into text");
>
> DATA(insert OID = 760 ( smgrin PGNSP PGUID 12 f f t f s 1 210 "2275" smgrin - _null_ ));
> DESCR("I/O");
> ***************
> *** 1318,1323 ****
> --- 1331,1338 ----
> DESCR("remove ACL item");
> DATA(insert OID = 1037 ( aclcontains PGNSP PGUID 12 f f t f s 2 16 "1034 1033" aclcontains - _null_ ));
> DESCR("does ACL contain item?");
> + DATA(insert OID = 1062 ( aclitemeq PGNSP PGUID 12 f f t f s 2 16 "1033 1033" aclitem_eq - _null_ ));
> + DESCR("equality operator for ACL items");
> DATA(insert OID = 1038 ( seteval PGNSP PGUID 12 f f t t v 1 23 "26" seteval - _null_ ));
> DESCR("internal function supporting PostQuel-style sets");
> DATA(insert OID = 1044 ( bpcharin PGNSP PGUID 12 f f t f i 3 1042 "2275 26 23" bpcharin - _null_ ));
> Index: src/include/nodes/primnodes.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/nodes/primnodes.h,v
> retrieving revision 1.82
> diff -c -r1.82 primnodes.h
> *** src/include/nodes/primnodes.h 6 May 2003 00:20:33 -0000 1.82
> --- src/include/nodes/primnodes.h 1 Jun 2003 15:41:47 -0000
> ***************
> *** 225,230 ****
> --- 225,231 ----
> Expr *target; /* expression we are aggregating on */
> bool aggstar; /* TRUE if argument was really '*' */
> bool aggdistinct; /* TRUE if it's agg(DISTINCT ...) */
> + List *args; /* arguments to the aggregate */
> } Aggref;
>
> /* ----------------
> ***************
> *** 357,371 ****
> /* ----------------
> * SubLink
> *
> ! * A SubLink represents a subselect appearing in an expression, and in some
> ! * cases also the combining operator(s) just above it. The subLinkType
> ! * indicates the form of the expression represented:
> * EXISTS_SUBLINK EXISTS(SELECT ...)
> * ALL_SUBLINK (lefthand) op ALL (SELECT ...)
> * ANY_SUBLINK (lefthand) op ANY (SELECT ...)
> * MULTIEXPR_SUBLINK (lefthand) op (SELECT ...)
> * EXPR_SUBLINK (SELECT with single targetlist item ...)
> * ARRAY_SUBLINK ARRAY(SELECT with single targetlist item ...)
> * For ALL, ANY, and MULTIEXPR, the lefthand is a list of expressions of the
> * same length as the subselect's targetlist. MULTIEXPR will *always* have
> * a list with more than one entry; if the subselect has just one target
> --- 358,376 ----
> /* ----------------
> * SubLink
> *
> ! * A SubLink represents a subselect, or an expression, appearing in an
> ! * expression, and in some cases also the combining operator(s) just above
> ! * it. The subLinkType indicates the form of the expression represented:
> * EXISTS_SUBLINK EXISTS(SELECT ...)
> * ALL_SUBLINK (lefthand) op ALL (SELECT ...)
> * ANY_SUBLINK (lefthand) op ANY (SELECT ...)
> * MULTIEXPR_SUBLINK (lefthand) op (SELECT ...)
> * EXPR_SUBLINK (SELECT with single targetlist item ...)
> * ARRAY_SUBLINK ARRAY(SELECT with single targetlist item ...)
> + * If an expression is used in place of the subselect, it is transformed
> + * into a simple "(SELECT expr)" in gram.y. This is to allow arrays to be
> + * used as if they were the result of a single column subselect. If the
> + * expression is scalar, it is treated as a one element array.
> * For ALL, ANY, and MULTIEXPR, the lefthand is a list of expressions of the
> * same length as the subselect's targetlist. MULTIEXPR will *always* have
> * a list with more than one entry; if the subselect has just one target
> ***************
> *** 414,419 ****
> --- 419,426 ----
> SubLinkType subLinkType; /* EXISTS, ALL, ANY, MULTIEXPR, EXPR */
> bool useOr; /* TRUE to combine column results with
> * "OR" not "AND" */
> + bool isExpr; /* TRUE if the subselect is really derived
> + * from a single expression */
> List *lefthand; /* list of outer-query expressions on the
> * left */
> List *operName; /* originally specified operator name */
> ***************
> *** 455,460 ****
> --- 462,476 ----
> SubLinkType subLinkType; /* EXISTS, ALL, ANY, MULTIEXPR, EXPR */
> bool useOr; /* TRUE to combine column results with
> * "OR" not "AND" */
> + bool isExpr; /* TRUE if the subselect is really derived
> + * from a single expression */
> + /* runtime cache for single array expressions */
> + Oid exprtype; /* array and element type, and other info
> + * needed deconstruct the array */
> + Oid elemtype;
> + int16 elmlen;
> + bool elmbyval;
> + char elmalign;
> /* The combining operators, transformed to executable expressions: */
> List *exprs; /* list of OpExpr expression trees */
> List *paramIds; /* IDs of Params embedded in the above */
> Index: src/include/optimizer/clauses.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/optimizer/clauses.h,v
> retrieving revision 1.63
> diff -c -r1.63 clauses.h
> *** src/include/optimizer/clauses.h 28 May 2003 16:04:02 -0000 1.63
> --- src/include/optimizer/clauses.h 1 Jun 2003 15:41:15 -0000
> ***************
> *** 28,33 ****
> --- 28,36 ----
> extern Node *get_leftop(Expr *clause);
> extern Node *get_rightop(Expr *clause);
>
> + extern Expr *make_funcclause(Oid funcid, Oid funcresulttype, bool funcretset,
> + CoercionForm funcformat, List *funcargs);
> +
> extern bool not_clause(Node *clause);
> extern Expr *make_notclause(Expr *notclause);
> extern Expr *get_notclausearg(Expr *notclause);
> Index: src/include/parser/parse_oper.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/parser/parse_oper.h,v
> retrieving revision 1.25
> diff -c -r1.25 parse_oper.h
> *** src/include/parser/parse_oper.h 29 Apr 2003 22:13:11 -0000 1.25
> --- src/include/parser/parse_oper.h 1 Jun 2003 23:21:48 -0000
> ***************
> *** 44,49 ****
> --- 44,50 ----
> /* Convenience routines for common calls on the above */
> extern Oid compatible_oper_opid(List *op, Oid arg1, Oid arg2, bool noError);
> extern Oid equality_oper_funcid(Oid argtype);
> + extern Oid ordering_oper_funcid(Oid argtype);
> extern Oid ordering_oper_opid(Oid argtype);
>
> /* Extract operator OID or underlying-function OID from an Operator tuple */
> Index: src/include/utils/acl.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/utils/acl.h,v
> retrieving revision 1.51
> diff -c -r1.51 acl.h
> *** src/include/utils/acl.h 23 Jan 2003 23:39:07 -0000 1.51
> --- src/include/utils/acl.h 1 Jun 2003 19:03:25 -0000
> ***************
> *** 188,193 ****
> --- 188,194 ----
> extern Datum aclinsert(PG_FUNCTION_ARGS);
> extern Datum aclremove(PG_FUNCTION_ARGS);
> extern Datum aclcontains(PG_FUNCTION_ARGS);
> + extern Datum aclitem_eq(PG_FUNCTION_ARGS);
>
> /*
> * prototypes for functions in aclchk.c
> Index: src/include/utils/array.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/utils/array.h,v
> retrieving revision 1.38
> diff -c -r1.38 array.h
> *** src/include/utils/array.h 8 May 2003 22:19:57 -0000 1.38
> --- src/include/utils/array.h 2 Jun 2003 02:25:03 -0000
> ***************
> *** 32,37 ****
> --- 32,68 ----
> Oid elemtype; /* element type OID */
> } ArrayType;
>
> + typedef struct ArrayBuildState
> + {
> + MemoryContext mcontext; /* where all the temp stuff is kept */
> + Datum *dvalues; /* array of accumulated Datums */
> + /*
> + * The allocated size of dvalues[] is always a multiple of
> + * ARRAY_ELEMS_CHUNKSIZE
> + */
> + #define ARRAY_ELEMS_CHUNKSIZE 64
> + int nelems; /* number of valid Datums in dvalues[] */
> + Oid element_type; /* data type of the Datums */
> + int16 typlen; /* needed info about datatype */
> + bool typbyval;
> + char typalign;
> + } ArrayBuildState;
> +
> + /*
> + * structure to cache type metadata needed for array manipulation
> + */
> + typedef struct ArrayMetaState
> + {
> + Oid element_type;
> + int typlen;
> + bool typbyval;
> + char typdelim;
> + Oid typelem;
> + Oid typiofunc;
> + char typalign;
> + FmgrInfo proc;
> + } ArrayMetaState;
> +
> /*
> * fmgr macros for array objects
> */
> ***************
> *** 86,91 ****
> --- 117,128 ----
> extern Datum array_send(PG_FUNCTION_ARGS);
> extern Datum array_length_coerce(PG_FUNCTION_ARGS);
> extern Datum array_eq(PG_FUNCTION_ARGS);
> + extern Datum array_ne(PG_FUNCTION_ARGS);
> + extern Datum array_lt(PG_FUNCTION_ARGS);
> + extern Datum array_gt(PG_FUNCTION_ARGS);
> + extern Datum array_le(PG_FUNCTION_ARGS);
> + extern Datum array_ge(PG_FUNCTION_ARGS);
> + extern Datum btarraycmp(PG_FUNCTION_ARGS);
> extern Datum array_dims(PG_FUNCTION_ARGS);
> extern Datum array_lower(PG_FUNCTION_ARGS);
> extern Datum array_upper(PG_FUNCTION_ARGS);
> ***************
> *** 124,130 ****
> Oid elmtype,
> int elmlen, bool elmbyval, char elmalign,
> Datum **elemsp, int *nelemsp);
> !
>
> /*
> * prototypes for functions defined in arrayutils.c
> --- 161,174 ----
> Oid elmtype,
> int elmlen, bool elmbyval, char elmalign,
> Datum **elemsp, int *nelemsp);
> ! extern ArrayBuildState *accumArrayResult(ArrayBuildState *astate,
> ! Datum dvalue, bool disnull,
> ! Oid element_type,
> ! MemoryContext rcontext);
> ! extern Datum makeArrayResult(ArrayBuildState *astate,
> ! MemoryContext rcontext);
> ! extern Datum makeMdArrayResult(ArrayBuildState *astate, int ndims,
> ! int *dims, int *lbs, MemoryContext rcontext);
>
> /*
> * prototypes for functions defined in arrayutils.c
> ***************
> *** 143,152 ****
> */
> extern Datum singleton_array(PG_FUNCTION_ARGS);
> extern Datum array_push(PG_FUNCTION_ARGS);
> - extern Datum array_accum(PG_FUNCTION_ARGS);
> extern Datum array_cat(PG_FUNCTION_ARGS);
>
> ! extern ArrayType *create_singleton_array(Oid element_type,
> Datum element,
> int ndims);
>
> --- 187,196 ----
> */
> extern Datum singleton_array(PG_FUNCTION_ARGS);
> extern Datum array_push(PG_FUNCTION_ARGS);
> extern Datum array_cat(PG_FUNCTION_ARGS);
>
> ! extern ArrayType *create_singleton_array(FunctionCallInfo fcinfo,
> ! Oid element_type,
> Datum element,
> int ndims);
>
> Index: src/include/utils/builtins.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/utils/builtins.h,v
> retrieving revision 1.219
> diff -c -r1.219 builtins.h
> *** src/include/utils/builtins.h 26 May 2003 00:11:28 -0000 1.219
> --- src/include/utils/builtins.h 1 Jun 2003 15:41:15 -0000
> ***************
> *** 530,535 ****
> --- 530,537 ----
> List **namelist);
> extern Datum replace_text(PG_FUNCTION_ARGS);
> extern Datum split_text(PG_FUNCTION_ARGS);
> + extern Datum text_to_array(PG_FUNCTION_ARGS);
> + extern Datum array_to_text(PG_FUNCTION_ARGS);
> extern Datum to_hex32(PG_FUNCTION_ARGS);
> extern Datum to_hex64(PG_FUNCTION_ARGS);
> extern Datum md5_text(PG_FUNCTION_ARGS);
> Index: src/include/utils/lsyscache.h
> ===================================================================
> RCS file: /opt/src/cvs/pgsql-server/src/include/utils/lsyscache.h,v
> retrieving revision 1.70
> diff -c -r1.70 lsyscache.h
> *** src/include/utils/lsyscache.h 26 May 2003 00:11:28 -0000 1.70
> --- src/include/utils/lsyscache.h 1 Jun 2003 15:41:15 -0000
> ***************
> *** 15,20 ****
> --- 15,29 ----
>
> #include "access/htup.h"
>
> + /* I/O function selector for system_cache_lookup */
> + typedef enum IOFuncSelector
> + {
> + IOFunc_input,
> + IOFunc_output,
> + IOFunc_receive,
> + IOFunc_send
> + } IOFuncSelector;
> +
> extern bool op_in_opclass(Oid opno, Oid opclass);
> extern bool op_requires_recheck(Oid opno, Oid opclass);
> extern Oid get_opclass_member(Oid opclass, int16 strategy);
> ***************
> *** 39,44 ****
> --- 48,54 ----
> extern RegProcedure get_oprjoin(Oid opno);
> extern char *get_func_name(Oid funcid);
> extern Oid get_func_rettype(Oid funcid);
> + extern Oid *get_func_argtypes(Oid funcid, int *nargs);
> extern bool get_func_retset(Oid funcid);
> extern bool func_strict(Oid funcid);
> extern char func_volatile(Oid funcid);
> ***************
> *** 54,59 ****
> --- 64,77 ----
> extern void get_typlenbyval(Oid typid, int16 *typlen, bool *typbyval);
> extern void get_typlenbyvalalign(Oid typid, int16 *typlen, bool *typbyval,
> char *typalign);
> + extern void get_type_metadata(Oid element_type,
> + IOFuncSelector which_func,
> + int *typlen,
> + bool *typbyval,
> + char *typdelim,
> + Oid *typelem,
> + Oid *proc,
> + char *typalign);
> extern char get_typstorage(Oid typid);
> extern int32 get_typtypmod(Oid typid);
> extern Node *get_typdefault(Oid typid);