Query parsing speedup patch v2

[Mikko Tiihonen <mokki(at)elisanet(dot)fi>]

Hi,

This is a second version of the patch. It changes the following in the
VisibleBufferedInputStream:
- IndexOutOfBounds in scanCString caused by typo
- better buffer management, now tries harder to avoid short reads from
the wrapped stream

diff -Nur postgresql-jdbc-8.2dev-503.src/org/postgresql/core/Encoding.java postgresql-jdbc-8.2dev-503.src.patched/org/postgresql/core/Encoding.java
--- postgresql-jdbc-8.2dev-503.src/org/postgresql/core/Encoding.java	2005-07-04 05:18:32.000000000 +0300
+++ postgresql-jdbc-8.2dev-503.src.patched/org/postgresql/core/Encoding.java	2006-07-25 22:03:38.000000000 +0300
@@ -16,7 +16,7 @@
 import java.io.OutputStreamWriter;
 import java.io.Writer;
 import java.io.IOException;
-import java.util.Hashtable;
+import java.util.HashMap;
 
 /**
  * Representation of a particular character encoding.
@@ -28,7 +28,7 @@
     /*
      * Preferred JVM encodings for backend encodings.
      */
-    private static final Hashtable encodings = new Hashtable();
+    private static final HashMap encodings = new HashMap();
 
     static {
         //Note: this list should match the set of supported server
@@ -75,10 +75,22 @@
     }
 
     private final String encoding;
+    private final boolean fastASCIINumbers;
 
     protected Encoding(String encoding)
     {
         this.encoding = encoding;
+        fastASCIINumbers = testAsciiNumbers();
+    }
+    
+    /**
+     * Returns true if this encoding has characters
+     * '-' and '0'..'9' in exactly same posision as ascii.
+     *  
+     * @return true if the bytes can be scanned directly for ascii numbers.
+     */
+    public boolean hasAsciiNumbers() {
+        return fastASCIINumbers;
     }
 
     /**
@@ -115,9 +127,9 @@
         // encoding in the JVM we use that. Otherwise we fall back
         // to the default encoding of the JVM.
 
-        if (encodings.containsKey(databaseEncoding))
+        String[] candidates = (String[]) encodings.get(databaseEncoding);
+        if (candidates != null)
         {
-            String[] candidates = (String[]) encodings.get(databaseEncoding);
             for (int i = 0; i < candidates.length; i++)
             {
                 if (isAvailable(candidates[i]))
@@ -254,4 +266,28 @@
     public String toString() {
         return (encoding == null ? "<default JVM encoding>" : encoding);
     }
+    
+    /**
+     * Checks weather this encoding is compatible with ASCII for the number
+     * characters '-' and '0'..'9'. Where compatible means that they are encoded
+     * with exactly same values. 
+     * 
+     * @return If faster ASCII number parsing can be used with this encoding.
+     */
+    private boolean testAsciiNumbers() {
+        // TODO: test all postgres supported encoding to see if there are
+        // any which do _not_ have ascii numbers in same location
+        // at least all the encoding listed in the encodings hashmap have
+        // working ascii numbers
+	try {
+	    String test = "-0123456789";
+	    byte[] bytes = encode(test);
+	    String res = new String(bytes, "US-ASCII");
+	    return test.equals(res);
+	} catch (java.io.UnsupportedEncodingException e) {
+	    return false;
+	} catch (IOException e) {
+	    return false;
+	}
+    }
 }
diff -Nur postgresql-jdbc-8.2dev-503.src/org/postgresql/core/PGStream.java postgresql-jdbc-8.2dev-503.src.patched/org/postgresql/core/PGStream.java
--- postgresql-jdbc-8.2dev-503.src/org/postgresql/core/PGStream.java	2006-04-29 16:30:23.000000000 +0300
+++ postgresql-jdbc-8.2dev-503.src.patched/org/postgresql/core/PGStream.java	2006-07-25 22:03:38.000000000 +0300
@@ -9,7 +9,6 @@
 */
 package org.postgresql.core;
 
-import java.io.BufferedInputStream;
 import java.io.BufferedOutputStream;
 import java.io.InputStream;
 import java.io.OutputStream;
@@ -37,7 +36,7 @@
     private final int port;
 
     private Socket connection;
-    private InputStream pg_input;
+    private VisibleBufferedInputStream pg_input;
     private OutputStream pg_output;
     private byte[] streamBuffer;
 
@@ -103,7 +102,7 @@
         connection.setTcpNoDelay(true);
 
         // Buffer sizes submitted by Sverre H Huseby <sverrehu(at)online(dot)no>
-        pg_input = new BufferedInputStream(connection.getInputStream(), 8192);
+        pg_input = new VisibleBufferedInputStream(connection.getInputStream(), 8192);
         pg_output = new BufferedOutputStream(connection.getOutputStream(), 8192);
 
         if (encoding != null)
@@ -268,14 +267,15 @@
      */
     public int ReceiveIntegerR(int siz) throws IOException
     {
+        if (!pg_input.ensureBytes(siz)) {
+            throw new EOFException();
+        }
         int n = 0;
 
         for (int i = 0 ; i < siz ; i++)
         {
-            int b = pg_input.read();
+            int b = pg_input.readRaw() & 0xFF;
 
-            if (b < 0)
-                throw new EOFException();
             n = b | (n << 8);
         }
 
@@ -315,36 +315,11 @@
      */
     public String ReceiveString() throws IOException
     {
-        int i = 0;
-        byte[] rst = byte_buf;
-        int buflen = rst.length;
-
-        while (true)
-        {
-            int c = pg_input.read();
-
-            if (c < 0)
-                throw new EOFException();
-
-            if (c == 0)
-                break;
-
-            if (i == buflen)
-            {
-                // Grow the buffer.
-                buflen *= 2;     // 100% bigger
-                if (buflen <= 0) // Watch for overflow
-                    throw new IOException("Impossibly long string");
-
-                byte[] newrst = new byte[buflen];
-                System.arraycopy(rst, 0, newrst, 0, i);
-                rst = newrst;
-            }
-
-            rst[i++] = (byte)c;
-        }
-
-        return encoding.decode(rst, 0, i);
+        int len = pg_input.scanCStringLength();
+        String res = encoding.decode(pg_input.getBuffer(), pg_input.getIndex(),
+                                     len - 1);
+        pg_input.skip(len);
+        return res;
     }
 
     /**
diff -Nur postgresql-jdbc-8.2dev-503.src/org/postgresql/core/VisibleBufferedInputStream.java postgresql-jdbc-8.2dev-503.src.patched/org/postgresql/core/VisibleBufferedInputStream.java
--- postgresql-jdbc-8.2dev-503.src/org/postgresql/core/VisibleBufferedInputStream.java	1970-01-01 02:00:00.000000000 +0200
+++ postgresql-jdbc-8.2dev-503.src.patched/org/postgresql/core/VisibleBufferedInputStream.java	2006-07-26 14:32:53.000000000 +0300
@@ -0,0 +1,293 @@
+/*-------------------------------------------------------------------------
+ *
+ * Copyright (c) 2006, PostgreSQL Global Development Group
+ *
+ * IDENTIFICATION
+ *   $PostgreSQL: pgjdbc/org/postgresql/core/PGStream.java,v 1.19 2006/04/29 13:30:23 jurka Exp $
+ *
+ *-------------------------------------------------------------------------
+ */
+package org.postgresql.core;
+
+import java.io.EOFException;
+import java.io.IOException;
+import java.io.InputStream;
+
+/**
+ * A faster version of BufferedInputStream. Does no synchronisation and
+ * allows direct access to the used byte[] buffer. 
+ * 
+ * @author Mikko Tiihonen
+ */
+public class VisibleBufferedInputStream extends InputStream {
+
+    /**
+     * If a direct read to byte array is called that would require
+     * a smaller read from the wrapped stream that MINIMUM_READ then
+     * first fill the buffer and serve the bytes from there. Larger
+     * reads are directly done to the provided byte array.
+     */
+    private static final int MINIMUM_READ = 1024;
+    
+    /**
+     * In how large spans is the C string zero-byte scanned.
+     */
+    private static final int STRING_SCAN_SPAN = 1024;
+
+    /**
+     * The wrapped input stream.
+     */
+    private final InputStream wrapped;
+
+    /**
+     * The buffer.
+     */
+    private byte[] buffer;
+
+    /**
+     * Current read position in the buffer.
+     */
+    private int index;
+
+    /**
+     * How far is the buffer filled with valid data.
+     */
+    private int endIndex;
+
+    /**
+     * Creates a new buffer around the given stream.
+     * 
+     * @param in The stream to buffer.
+     * @param bufferSize The initial size of the buffer. 
+     */
+    public VisibleBufferedInputStream(InputStream in, int bufferSize) {
+        wrapped = in;
+        buffer = new byte[bufferSize < MINIMUM_READ ?
+                          MINIMUM_READ : bufferSize];
+    }
+
+    /**
+     * {(at)inheritDoc}
+     */
+    public int read() throws IOException {
+        if (ensureBytes(1)) {
+            return buffer[index++] & 0xFF;
+        }
+        return -1;
+    }
+
+    /**
+     * Reads byte from the buffer without any checks. This method never
+     * reads from the underlaying stream.
+     * Before calling this method the {(at)link #ensureBytes} method must
+     * have been called.
+     * 
+     * @return The next byte from the buffer.
+     * @throws ArrayIndexOutOfBoundsException If ensureBytes was not called
+     * to make sure the buffer contains the byte.  
+     */
+    public byte readRaw() {
+        return buffer[index++];
+    }
+
+    /**
+     * Ensures that the buffer contains at least n bytes.
+     * This method invalidates the buffer and index fields.
+     * 
+     * @param n The amount of bytes to ensure exists in buffer
+     * @return true if required bytes are available and false if EOF
+     * @throws IOException If reading of the wrapped stream failed.
+     */
+    public boolean ensureBytes(int n) throws IOException {
+        int required = n - endIndex + index;
+        while (required > 0) {
+            if (!readMore(required)) {
+                return false;
+            }
+            required = n - endIndex + index;
+        }
+        return true;
+    }
+
+    /**
+     * Reads more bytes into the buffer.
+     * 
+     * @param wanted How much should be at least read.
+     * @return True if at least some bytes were read.
+     * @throws IOException If reading of the wrapped stream failed.
+     */
+    private boolean readMore(int wanted) throws IOException {
+        if (endIndex == index) {
+            index = 0;
+            endIndex = 0;
+        }
+        int canFit = buffer.length - endIndex;
+        if (canFit < wanted) {
+            // would the wanted bytes fit if we compacted the buffer
+            // and still leave some slack
+            if (index + canFit > wanted + MINIMUM_READ) {
+                compact();
+            } else {
+                doubleBuffer();
+            }
+            canFit = buffer.length - endIndex;
+        }
+        int read = wrapped.read(buffer, endIndex, canFit);
+        if (read < 0) {
+            return false;
+        }
+        endIndex += read;
+        return true;
+    }
+
+    /**
+     * Doubles the size of the buffer.
+     */
+    private void doubleBuffer() {
+        byte[] buf = new byte[buffer.length * 2];
+        moveBufferTo(buf);
+        buffer = buf;
+    }
+
+    /**
+     * Compacts the unread bytes of the buffer to the beginning of the buffer.
+     */
+    private void compact() {
+        moveBufferTo(buffer);
+    }
+
+    /**
+     * Moves bytes from the buffer to the begining of the destination buffer.
+     * Also sets the index and endIndex variables.
+     * 
+     * @param dest The destination buffer.
+     */
+    private void moveBufferTo(byte[] dest) {
+        int size = endIndex - index;
+        System.arraycopy(buffer, index, dest, 0, size);
+        index = 0;
+        endIndex = size;
+    }
+
+    /**
+     * {(at)inheritDoc}
+     */
+    public int read(byte to[], int off, int len) throws IOException {
+        if ((off | len | (off + len) | (to.length - (off + len))) < 0) {
+            throw new IndexOutOfBoundsException();
+        } else if (len == 0) {
+            return 0;
+        }
+
+        // if the read would go to wrapped stream, but would result
+        // in a small read then try read to the buffer instead
+        int avail = endIndex - index;
+        if (len - avail < MINIMUM_READ) {
+            ensureBytes(len);
+            avail = endIndex - index;
+        }
+
+        // first copy from buffer
+        if (avail > 0) {
+            if (len <= avail) {
+                System.arraycopy(buffer, index, to, off, len);
+                index += len;
+                return len;
+            }
+            System.arraycopy(buffer, index, to, off, avail);
+            len -= avail;
+            off += avail;
+        }
+        int read = avail;
+
+        // good place to reset index because the buffer is fully drained
+        index = 0;
+        endIndex = 0;
+        
+        // then directly from wrapped stream
+        do {
+            int r = wrapped.read(to, off, len);
+            if (r <= 0) {
+                return (read == 0) ? r : read;
+            }
+            read += r;
+            off += r;
+            len -= r;
+        } while (len > 0);
+
+        return read;
+    }
+
+    /**
+     * {(at)inheritDoc}
+     */
+    public long skip(long n) throws IOException {
+        int avail = endIndex - index;
+        if (avail >= n) {
+            index += n;
+            return n;
+        }
+        n -= avail;
+        index = 0;
+        endIndex = 0;
+        return wrapped.skip(n);
+    }
+
+    /**
+     * {(at)inheritDoc}
+     */
+    public int available() throws IOException {
+        int avail = endIndex - index;
+        return avail > 0 ? avail : wrapped.available();
+    }
+
+    /**
+     * {(at)inheritDoc}
+     */
+    public void close() throws IOException {
+        wrapped.close();
+    }
+
+    /**
+     * Returns direct handle to the used buffer. Use the {(at)link #ensureBytes}
+     * to prefill required bytes the buffer and {(at)link #getIndex} to fetch
+     * the current position of the buffer.
+     *  
+     * @return The underlaying buffer. 
+     */
+    public byte[] getBuffer() {
+        return buffer;
+    }
+
+    /**
+     * Returns the current read position in the buffer.
+     * 
+     * @return the current read position in the buffer.
+     */
+    public int getIndex() {
+        return index;
+    }
+
+    /**
+     * Scans the length of the next null terminated string (C-style string) from
+     * the stream.
+     * 
+     * @return The length of the next null terminated string.
+     * @throws IOException If reading of stream fails.
+     * @throws EOFxception If the stream did not contain any null terminators.
+     */
+    public int scanCStringLength() throws IOException {
+        int pos = index;
+        for (;;) {
+            while (pos < endIndex) {
+                if (buffer[pos++] == '\0') {
+                    return pos - index;
+                }
+            }
+            if (!readMore(STRING_SCAN_SPAN)) {
+                throw new EOFException();
+            }
+            pos = index;
+        }
+    }
+}
diff -Nur postgresql-jdbc-8.2dev-503.src/org/postgresql/jdbc2/AbstractJdbc2ResultSet.java postgresql-jdbc-8.2dev-503.src.patched/org/postgresql/jdbc2/AbstractJdbc2ResultSet.java
--- postgresql-jdbc-8.2dev-503.src/org/postgresql/jdbc2/AbstractJdbc2ResultSet.java	2005-12-04 23:40:33.000000000 +0200
+++ postgresql-jdbc-8.2dev-503.src.patched/org/postgresql/jdbc2/AbstractJdbc2ResultSet.java	2006-07-25 22:03:38.000000000 +0300
@@ -1957,13 +1957,148 @@
 
     public int getInt(int columnIndex) throws SQLException
     {
+        Encoding encoding = connection.getEncoding();
+        if (encoding.hasAsciiNumbers()) {
+            try {
+        	return getFastInt(columnIndex);
+            } catch (NumberFormatException ex) {
+            }
+        }
         return toInt( getFixedString(columnIndex) );
     }
 
     public long getLong(int columnIndex) throws SQLException
     {
+        Encoding encoding = connection.getEncoding();
+        if (encoding.hasAsciiNumbers()) {
+            try {
+        	return getFastLong(columnIndex);
+            } catch (NumberFormatException ex) {
+            }
+        }
         return toLong( getFixedString(columnIndex) );
     }
+        
+    private static final NumberFormatException FAST_NUMBER_FAILED = 
+	new NumberFormatException();
+
+    /**
+     * Optimised byte[] to number parser.
+     * 
+     * @param columnIndex The column to parse.
+     * @return The parsed number.
+     * @throws SQLException If an error occurs while fetching column.
+     * @throws NumberFormatException If the number is invalid or the
+     * out of range for fast parsing. The value must then be parsed by
+     * {(at)link #toLong(String)}.
+     */
+    private long getFastLong(int columnIndex) throws SQLException,
+        NumberFormatException {
+	
+        checkResultSet( columnIndex );
+        
+        columnIndex--;
+        if (this_row[columnIndex] == null) {
+            return 0; // SQL NULL
+        }
+        
+        byte[] bytes = this_row[columnIndex];
+        
+        if (bytes.length == 0) {
+            throw FAST_NUMBER_FAILED;
+        }
+        
+        long val = 0;
+        int start;
+        boolean neg;
+        if (bytes[0] == '-') {
+            neg = true;
+            start = 1;
+            if (bytes.length > 19) {
+                throw FAST_NUMBER_FAILED;
+            }
+        } else {
+            start = 0;
+            neg = false;
+            if (bytes.length > 18) {
+                throw FAST_NUMBER_FAILED;
+            }
+        }
+        
+        while (start < bytes.length) {
+            byte b = bytes[start++];
+            if (b < '0' || b > '9') {
+                throw FAST_NUMBER_FAILED;
+            }
+            
+            val *= 10;
+            val += b - '0';
+        }
+        
+        if (neg) {
+            val = -val;
+        }        
+        
+        return val;
+    }
+
+    /**
+     * Optimised byte[] to number parser.
+     * 
+     * @param columnIndex The column to parse.
+     * @return The parsed number.
+     * @throws SQLException If an error occurs while fetching column.
+     * @throws NumberFormatException If the number is invalid or the
+     * out of range for fast parsing. The value must then be parsed by
+     * {(at)link #toLong(String)}.
+     */
+    private int getFastInt(int columnIndex) throws SQLException {
+        checkResultSet( columnIndex );
+        
+        columnIndex--;
+        if (this_row[columnIndex] == null) {
+            return 0; // SQL NULL
+        }
+        
+        byte[] bytes = this_row[columnIndex];
+        
+        if (bytes.length == 0) {
+            throw FAST_NUMBER_FAILED;
+        }
+        
+        int val = 0;
+        int start;
+        boolean neg;
+        if (bytes[0] == '-') {
+            neg = true;
+            start = 1;
+            if (bytes.length > 9) {
+                throw FAST_NUMBER_FAILED;
+            }
+        } else {
+            start = 0;
+            neg = false;
+            if (bytes.length > 8) {
+                throw FAST_NUMBER_FAILED;
+            }
+        }
+        
+        while (start < bytes.length) {
+            byte b = bytes[start++];
+            if (b < '0' || b > '9') {
+                throw FAST_NUMBER_FAILED;
+            }
+            
+            val *= 10;
+            val += b - '0';
+        }
+        
+        if (neg) {
+            val = -val;
+        }        
+        
+        return val;
+    }
 
     public float getFloat(int columnIndex) throws SQLException
     {
@@ -2358,24 +2493,31 @@
      */
     public String getFixedString(int col) throws SQLException
     {
-        String s = getString(col);
-
         // Handle SQL Null
         wasNullFlag = (this_row[col - 1] == null);
         if (wasNullFlag)
             return null;
 
+        String s = getString(col);
+
         // if we don't have at least 2 characters it can't be money.
         if (s.length() < 2)
             return s;
 
         // Handle Money
         char ch = s.charAt(0);
+        
+        // optimise for non-money type
+        // '(', '$' and '-' are all before numbers in ascii
+        if (ch >= '0') {
+            return s;
+        }
+        
         if (ch == '(')
         {
             s = "-" + PGtokenizer.removePara(s).substring(1);
         }
-        if (ch == '$')
+        else if (ch == '$')
         {
             s = s.substring(1);
         }