]> granicus.if.org Git - postgresql/commitdiff
Add:
authorBruce Momjian <bruce@momjian.us>
Thu, 22 Aug 2002 03:24:01 +0000 (03:24 +0000)
committerBruce Momjian <bruce@momjian.us>
Thu, 22 Aug 2002 03:24:01 +0000 (03:24 +0000)
replace(string, from, to)
   -- replaces all occurrences of "from" in "string" to "to"
split(string, fldsep, column)
   -- splits "string" on "fldsep" and returns "column" number piece
to_hex(int32_num) & to_hex(int64_num)
   -- takes integer number and returns as hex string

Joe Conway

src/backend/utils/adt/varlena.c
src/include/catalog/pg_proc.h
src/include/utils/builtins.h
src/test/regress/expected/strings.out
src/test/regress/sql/strings.sql

index bdcffa954e0fe17aae68363d0ab2e93752c9f2fa..9f8a4d563ee4dfe6a069808de958fb0f57360dc4 100644 (file)
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *       $Header: /cvsroot/pgsql/src/backend/utils/adt/varlena.c,v 1.87 2002/08/04 06:44:47 thomas Exp $
+ *       $Header: /cvsroot/pgsql/src/backend/utils/adt/varlena.c,v 1.88 2002/08/22 03:24:00 momjian Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -18,6 +18,8 @@
 
 #include "mb/pg_wchar.h"
 #include "miscadmin.h"
+#include "access/tuptoaster.h"
+#include "lib/stringinfo.h"
 #include "utils/builtins.h"
 #include "utils/pg_locale.h"
 
@@ -27,8 +29,34 @@ typedef struct varlena unknown;
 #define DatumGetUnknownP(X)                    ((unknown *) PG_DETOAST_DATUM(X))
 #define PG_GETARG_UNKNOWN_P(n)         DatumGetUnknownP(PG_GETARG_DATUM(n))
 #define PG_RETURN_UNKNOWN_P(x)         PG_RETURN_POINTER(x)
+#define PG_TEXTARG_GET_STR(arg_) \
+    DatumGetCString(DirectFunctionCall1(textout, PG_GETARG_DATUM(arg_)))
+#define PG_TEXT_GET_STR(textp_) \
+    DatumGetCString(DirectFunctionCall1(textout, PointerGetDatum(textp_)))
+#define PG_STR_GET_TEXT(str_) \
+    DatumGetTextP(DirectFunctionCall1(textin, CStringGetDatum(str_)))
+#define TEXTLEN(textp) \
+       text_length(PointerGetDatum(textp))
+#define TEXTPOS(buf_text, from_sub_text) \
+       text_position(PointerGetDatum(buf_text), PointerGetDatum(from_sub_text), 1)
+#define TEXTDUP(textp) \
+       DatumGetTextPCopy(PointerGetDatum(textp))
+#define LEFT(buf_text, from_sub_text) \
+       text_substring(PointerGetDatum(buf_text), \
+                                       1, \
+                                       TEXTPOS(buf_text, from_sub_text) - 1, false)
+#define RIGHT(buf_text, from_sub_text, from_sub_text_len) \
+       text_substring(PointerGetDatum(buf_text), \
+                                       TEXTPOS(buf_text, from_sub_text) + from_sub_text_len, \
+                                       -1, true)
 
 static int     text_cmp(text *arg1, text *arg2);
+static int32 text_length(Datum str);
+static int32 text_position(Datum str, Datum search_str, int matchnum);
+static text *text_substring(Datum str,
+                                                       int32 start,
+                                                       int32 length,
+                                                       bool length_not_specified);
 
 
 /*****************************************************************************
@@ -285,19 +313,36 @@ unknownout(PG_FUNCTION_ARGS)
 Datum
 textlen(PG_FUNCTION_ARGS)
 {
-       text       *t = PG_GETARG_TEXT_P(0);
+       PG_RETURN_INT32(text_length(PG_GETARG_DATUM(0)));
+}
 
-#ifdef MULTIBYTE
-       /* optimization for single byte encoding */
-       if (pg_database_encoding_max_length() <= 1)
-               PG_RETURN_INT32(VARSIZE(t) - VARHDRSZ);
+/*
+ * text_length -
+ *     Does the real work for textlen()
+ *     This is broken out so it can be called directly by other string processing
+ *     functions.
+ */
+static int32
+text_length(Datum str)
+{
+       /* fastpath when max encoding length is one */
+       if (pg_database_encoding_max_length() == 1)
+               PG_RETURN_INT32(toast_raw_datum_size(str) - VARHDRSZ);
 
-       PG_RETURN_INT32(
-               pg_mbstrlen_with_len(VARDATA(t), VARSIZE(t) - VARHDRSZ)
-               );
-#else
-       PG_RETURN_INT32(VARSIZE(t) - VARHDRSZ);
-#endif
+       if (pg_database_encoding_max_length() > 1)
+       {
+               text       *t = DatumGetTextP(str);
+
+               PG_RETURN_INT32(pg_mbstrlen_with_len(VARDATA(t),
+                                                                        VARSIZE(t) - VARHDRSZ));
+       }
+
+       /* should never get here */
+       elog(ERROR, "Invalid backend encoding; encoding max length "
+                               "is less than one.");
+
+       /* not reached: suppress compiler warning */
+       return 0;
 }
 
 /*
@@ -308,9 +353,7 @@ textlen(PG_FUNCTION_ARGS)
 Datum
 textoctetlen(PG_FUNCTION_ARGS)
 {
-       text    *arg = PG_GETARG_TEXT_P(0);
-
-       PG_RETURN_INT32(VARSIZE(arg) - VARHDRSZ);
+       PG_RETURN_INT32(toast_raw_datum_size(PG_GETARG_DATUM(0)) - VARHDRSZ);
 }
 
 /*
@@ -382,90 +425,201 @@ textcat(PG_FUNCTION_ARGS)
  * - Thomas Lockhart 1998-12-10
  * Now uses faster TOAST-slicing interface
  * - John Gray 2002-02-22
+ * Remove "#ifdef MULTIBYTE" and test for encoding_max_length instead. Change
+ * behaviors conflicting with SQL92 to meet SQL92 (if E = S + L < S throw
+ * error; if E < 1, return '', not entire string). Fixed MB related bug when
+ * S > LC and < LC + 4 sometimes garbage characters are returned.
+ * - Joe Conway 2002-08-10 
  */
 Datum
 text_substr(PG_FUNCTION_ARGS)
 {
-       text       *string;
-       int32           m = PG_GETARG_INT32(1);
-       int32           n = PG_GETARG_INT32(2);
-       int32       sm;
-       int32       sn;
-       int         eml = 1;
-#ifdef MULTIBYTE
-       int                     i;
-       int                     len;
-       text       *ret;
-       char       *p;
-#endif 
+       PG_RETURN_TEXT_P(text_substring(PG_GETARG_DATUM(0),
+                                                                       PG_GETARG_INT32(1),
+                                                                       PG_GETARG_INT32(2),
+                                                                       false));
+}
 
-       /*
-        * starting position before the start of the string? then offset into
-        * the string per SQL92 spec...
-        */
-       if (m < 1)
+/*
+ * text_substr_no_len -
+ *       Wrapper to avoid opr_sanity failure due to
+ *       one function accepting a different number of args.
+ */
+Datum
+text_substr_no_len(PG_FUNCTION_ARGS)
+{
+       PG_RETURN_TEXT_P(text_substring(PG_GETARG_DATUM(0),
+                                                                       PG_GETARG_INT32(1),
+                                                                       -1, true));
+}
+
+/*
+ * text_substring -
+ *     Does the real work for text_substr() and text_substr_no_len()
+ *     This is broken out so it can be called directly by other string processing
+ *     functions.
+ */
+static text*
+text_substring(Datum str, int32 start, int32 length, bool length_not_specified)
+{
+       int32           eml = pg_database_encoding_max_length();
+       int32           S = start;                              /* start position */
+       int32           S1;                                             /* adjusted start position */
+       int32           L1;                                             /* adjusted substring length */
+
+       /* life is easy if the encoding max length is 1 */
+       if (eml == 1)
        {
-               n += (m - 1);
-               m = 1;
-       }
-       /* Check for m > octet length is made in TOAST access routine */
+               S1 = Max(S, 1);
 
-       /* m will now become a zero-based starting position */
-       sm = m - 1;
-       sn = n;
+               if (length_not_specified)       /* special case - get length to end of string */
+                       L1 = -1;
+               else
+               {
+                       /* end position */
+                       int     E = S + length;
 
-#ifdef MULTIBYTE
-       eml = pg_database_encoding_max_length ();
+                       /*
+                        * A negative value for L is the only way for the end position
+                        * to be before the start. SQL99 says to throw an error.
+                        */
+                       if (E < S)
+                               elog(ERROR, "negative substring length not allowed");
+
+                       /* 
+                        * A zero or negative value for the end position can happen if the start
+                        * was negative or one. SQL99 says to return a zero-length string.
+                        */
+                       if (E < 1)
+                               return PG_STR_GET_TEXT("");
 
-       if (eml > 1)
+                       L1 = E - S1;
+               }
+
+               /* 
+                * If the start position is past the end of the string,
+                * SQL99 says to return a zero-length string -- 
+                * PG_GETARG_TEXT_P_SLICE() will do that for us.
+                * Convert to zero-based starting position
+                */
+               return DatumGetTextPSlice(str, S1 - 1, L1);
+       }
+       else if (eml > 1)
        {
-               sm = 0;
-               if (n > -1)
-                       sn = (m + n) * eml + 3; /* +3 to avoid mb characters overhanging slice end */
+               /*
+                * When encoding max length is > 1, we can't get LC without
+                * detoasting, so we'll grab a conservatively large slice
+                * now and go back later to do the right thing
+                */
+               int32           slice_start;
+               int32           slice_size;
+               int32           slice_strlen;
+               text            *slice;
+               int32           E1;
+               int32           i;
+               char       *p;
+               char       *s;
+               text       *ret;
+
+               /*
+                * if S is past the end of the string, the tuple toaster
+                * will return a zero-length string to us
+                */
+               S1 = Max(S, 1);
+
+               /*
+                * We need to start at position zero because there is no
+                * way to know in advance which byte offset corresponds to 
+                * the supplied start position.
+                */
+               slice_start = 0;
+
+               if (length_not_specified)       /* special case - get length to end of string */
+                       slice_size = L1 = -1;
                else
-                       sn = n;         /* n < 0 is special-cased by heap_tuple_untoast_attr_slice */
-       }
-#endif 
+               {
+                       int     E = S + length;
+
+                       /*
+                        * A negative value for L is the only way for the end position
+                        * to be before the start. SQL99 says to throw an error.
+                        */
+                       if (E < S)
+                               elog(ERROR, "negative substring length not allowed");
 
-       string = PG_GETARG_TEXT_P_SLICE (0, sm, sn);
+                       /* 
+                        * A zero or negative value for the end position can happen if the start
+                        * was negative or one. SQL99 says to return a zero-length string.
+                        */
+                       if (E < 1)
+                               return PG_STR_GET_TEXT("");
 
-       if (eml == 1) 
-       {
-               PG_RETURN_TEXT_P (string);
-       }
-#ifndef MULTIBYTE
-       PG_RETURN_NULL();   /* notreached: suppress compiler warning */
-#endif
-#ifdef MULTIBYTE
-       if (n > -1)
-               len = pg_mbstrlen_with_len (VARDATA (string), sn - 3);
-       else    /* n < 0 is special-cased; need full string length */
-               len = pg_mbstrlen_with_len (VARDATA (string), VARSIZE(string)-VARHDRSZ);
+                       /*
+                        * if E is past the end of the string, the tuple toaster
+                        * will truncate the length for us
+                        */
+                       L1 = E - S1;
 
-       if (m > len)
-       {
-               m = 1;
-               n = 0;
-       }
-       m--;
-       if (((m + n) > len) || (n < 0))
-               n = (len - m);
+                       /*
+                        * Total slice size in bytes can't be any longer than the start
+                        * position plus substring length times the encoding max length.
+                        */
+                       slice_size = (S1 + L1) * eml;
+               }
+               slice = DatumGetTextPSlice(str, slice_start, slice_size);
 
-       p = VARDATA(string);
-       for (i = 0; i < m; i++)
-               p += pg_mblen(p);
-       m = p - VARDATA(string);
-       for (i = 0; i < n; i++)
-               p += pg_mblen(p);
-       n = p - (VARDATA(string) + m);
+               /* see if we got back an empty string */
+               if ((VARSIZE(slice) - VARHDRSZ) == 0)
+                       return PG_STR_GET_TEXT("");
 
-       ret = (text *) palloc(VARHDRSZ + n);
-       VARATT_SIZEP(ret) = VARHDRSZ + n;
+               /* Now we can get the actual length of the slice in MB characters */
+               slice_strlen = pg_mbstrlen_with_len (VARDATA(slice), VARSIZE(slice) - VARHDRSZ);
 
-       memcpy(VARDATA(ret), VARDATA(string) + m, n);
+               /* Check that the start position wasn't > slice_strlen. If so,
+                * SQL99 says to return a zero-length string.
+                */
+               if (S1 > slice_strlen)
+                       return PG_STR_GET_TEXT("");
 
-       PG_RETURN_TEXT_P(ret);
-#endif
+               /*
+                * Adjust L1 and E1 now that we know the slice string length.
+                * Again remember that S1 is one based, and slice_start is zero based.
+                */
+               if (L1 > -1)
+                       E1 = Min(S1 + L1 , slice_start + 1 + slice_strlen);
+               else
+                       E1 = slice_start + 1 + slice_strlen;
+
+               /*
+                * Find the start position in the slice;
+                * remember S1 is not zero based
+                */
+               p = VARDATA(slice);
+               for (i = 0; i < S1 - 1; i++)
+                       p += pg_mblen(p);
+
+               /* hang onto a pointer to our start position */
+               s = p;
+
+               /*
+                * Count the actual bytes used by the substring of 
+                * the requested length.
+                */
+               for (i = S1; i < E1; i++)
+                       p += pg_mblen(p);
+
+               ret = (text *) palloc(VARHDRSZ + (p - s));
+               VARATT_SIZEP(ret) = VARHDRSZ + (p - s);
+               memcpy(VARDATA(ret), s, (p - s));
+
+               return ret;
+       }
+       else
+               elog(ERROR, "Invalid backend encoding; encoding max length "
+                                       "is less than one.");
+
+       /* not reached: suppress compiler warning */
+       return PG_STR_GET_TEXT("");
 }
 
 /*
@@ -481,56 +635,95 @@ text_substr(PG_FUNCTION_ARGS)
 Datum
 textpos(PG_FUNCTION_ARGS)
 {
-       text       *t1 = PG_GETARG_TEXT_P(0);
-       text       *t2 = PG_GETARG_TEXT_P(1);
-       int                     pos;
-       int                     px,
-                               p;
-       int                     len1,
+       PG_RETURN_INT32(text_position(PG_GETARG_DATUM(0), PG_GETARG_DATUM(1), 1));
+}
+
+/*
+ * text_position -
+ *     Does the real work for textpos()
+ *     This is broken out so it can be called directly by other string processing
+ *     functions.
+ */
+static int32
+text_position(Datum str, Datum search_str, int matchnum)
+{
+       int                     eml = pg_database_encoding_max_length();
+       text       *t1 = DatumGetTextP(str);
+       text       *t2 = DatumGetTextP(search_str);
+       int                     match = 0,
+                               pos = 0,
+                               p = 0,
+                               px,
+                               len1,
                                len2;
-       pg_wchar   *p1,
-                          *p2;
 
-#ifdef MULTIBYTE
-       pg_wchar   *ps1,
-                          *ps2;
-#endif
+       if(matchnum == 0)
+               return 0;               /* result for 0th match */
 
        if (VARSIZE(t2) <= VARHDRSZ)
                PG_RETURN_INT32(1);             /* result for empty pattern */
 
        len1 = (VARSIZE(t1) - VARHDRSZ);
        len2 = (VARSIZE(t2) - VARHDRSZ);
-#ifdef MULTIBYTE
-       ps1 = p1 = (pg_wchar *) palloc((len1 + 1) * sizeof(pg_wchar));
-       (void) pg_mb2wchar_with_len((unsigned char *) VARDATA(t1), p1, len1);
-       len1 = pg_wchar_strlen(p1);
-       ps2 = p2 = (pg_wchar *) palloc((len2 + 1) * sizeof(pg_wchar));
-       (void) pg_mb2wchar_with_len((unsigned char *) VARDATA(t2), p2, len2);
-       len2 = pg_wchar_strlen(p2);
-#else
-       p1 = VARDATA(t1);
-       p2 = VARDATA(t2);
-#endif
-       pos = 0;
+
+       /* no use in searching str past point where search_str will fit */
        px = (len1 - len2);
-       for (p = 0; p <= px; p++)
+
+       if (eml == 1)   /* simple case - single byte encoding */
        {
-#ifdef MULTIBYTE
-               if ((*p2 == *p1) && (pg_wchar_strncmp(p1, p2, len2) == 0))
-#else
-               if ((*p2 == *p1) && (strncmp(p1, p2, len2) == 0))
-#endif
+               char   *p1,
+                          *p2;
+
+               p1 = VARDATA(t1);
+               p2 = VARDATA(t2);
+
+               for (p = 0; p <= px; p++)
                {
-                       pos = p + 1;
-                       break;
-               };
-               p1++;
-       };
-#ifdef MULTIBYTE
-       pfree(ps1);
-       pfree(ps2);
-#endif
+                       if ((*p2 == *p1) && (strncmp(p1, p2, len2) == 0))
+                       {
+                               if (++match == matchnum)
+                               {
+                                       pos = p + 1;
+                                       break;
+                               }
+                       }
+                       p1++;
+               }
+       }
+       else if (eml > 1)       /* not as simple - multibyte encoding */
+       {
+               pg_wchar   *p1,
+                                  *p2,
+                                  *ps1,
+                                  *ps2;
+
+               ps1 = p1 = (pg_wchar *) palloc((len1 + 1) * sizeof(pg_wchar));
+               (void) pg_mb2wchar_with_len((unsigned char *) VARDATA(t1), p1, len1);
+               len1 = pg_wchar_strlen(p1);
+               ps2 = p2 = (pg_wchar *) palloc((len2 + 1) * sizeof(pg_wchar));
+               (void) pg_mb2wchar_with_len((unsigned char *) VARDATA(t2), p2, len2);
+               len2 = pg_wchar_strlen(p2);
+
+               for (p = 0; p <= px; p++)
+               {
+                       if ((*p2 == *p1) && (pg_wchar_strncmp(p1, p2, len2) == 0))
+                       {
+                               if (++match == matchnum)
+                               {
+                                       pos = p + 1;
+                                       break;
+                               }
+                       }
+                       p1++;
+               }
+
+               pfree(ps1);
+               pfree(ps2);
+       }
+       else
+               elog(ERROR, "Invalid backend encoding; encoding max length "
+                                       "is less than one.");
+
        PG_RETURN_INT32(pos);
 }
 
@@ -758,9 +951,7 @@ text_smaller(PG_FUNCTION_ARGS)
 Datum
 byteaoctetlen(PG_FUNCTION_ARGS)
 {
-       bytea      *v = PG_GETARG_BYTEA_P(0);
-
-       PG_RETURN_INT32(VARSIZE(v) - VARHDRSZ);
+       PG_RETURN_INT32(toast_raw_datum_size(PG_GETARG_DATUM(0)) - VARHDRSZ);
 }
 
 /*
@@ -805,6 +996,8 @@ byteacat(PG_FUNCTION_ARGS)
        PG_RETURN_BYTEA_P(result);
 }
 
+#define PG_STR_GET_BYTEA(str_) \
+    DatumGetByteaP(DirectFunctionCall1(byteain, CStringGetDatum(str_)))
 /*
  * bytea_substr()
  * Return a substring starting at the specified position.
@@ -813,33 +1006,71 @@ byteacat(PG_FUNCTION_ARGS)
  * Input:
  *     - string
  *     - starting position (is one-based)
- *     - string length
+ *     - string length (optional)
  *
  * If the starting position is zero or less, then return from the start of the string
  * adjusting the length to be consistent with the "negative start" per SQL92.
- * If the length is less than zero, return the remaining string.
- *
+ * If the length is less than zero, an ERROR is thrown. If no third argument
+ * (length) is provided, the length to the end of the string is assumed.
  */
 Datum
 bytea_substr(PG_FUNCTION_ARGS)
 {
-       int32           m = PG_GETARG_INT32(1);
-       int32           n = PG_GETARG_INT32(2);
+       int             S = PG_GETARG_INT32(1); /* start position */
+       int             S1;                                             /* adjusted start position */
+       int             L1;                                             /* adjusted substring length */
 
-       /*
-        * starting position before the start of the string? then offset into
-        * the string per SQL92 spec...
-        */
-       if (m < 1)
+       S1 = Max(S, 1);
+
+       if (fcinfo->nargs == 2)
        {
-               n += (m - 1);
-               m = 1;
+               /*
+                * Not passed a length - PG_GETARG_BYTEA_P_SLICE()
+                * grabs everything to the end of the string if we pass it
+                * a negative value for length.
+                */
+               L1 = -1;
        }
+       else
+       {
+               /* end position */
+               int     E = S + PG_GETARG_INT32(2);
 
-       /* m will now become a zero-based starting position */
-       m--;
+               /*
+                * A negative value for L is the only way for the end position
+                * to be before the start. SQL99 says to throw an error.
+                */
+               if (E < S)
+                       elog(ERROR, "negative substring length not allowed");
 
-       PG_RETURN_BYTEA_P(PG_GETARG_BYTEA_P_SLICE (0, m, n));
+               /* 
+                * A zero or negative value for the end position can happen if the start
+                * was negative or one. SQL99 says to return a zero-length string.
+                */
+               if (E < 1)
+                       PG_RETURN_BYTEA_P(PG_STR_GET_BYTEA(""));
+
+               L1 = E - S1;
+       }
+
+       /* 
+        * If the start position is past the end of the string,
+        * SQL99 says to return a zero-length string -- 
+        * PG_GETARG_TEXT_P_SLICE() will do that for us.
+        * Convert to zero-based starting position
+        */
+       PG_RETURN_BYTEA_P(PG_GETARG_BYTEA_P_SLICE (0, S1 - 1, L1));
+}
+
+/*
+ * bytea_substr_no_len -
+ *       Wrapper to avoid opr_sanity failure due to
+ *       one function accepting a different number of args.
+ */
+Datum
+bytea_substr_no_len(PG_FUNCTION_ARGS)
+{
+       return bytea_substr(fcinfo);
 }
 
 /*
@@ -1422,3 +1653,182 @@ byteacmp(PG_FUNCTION_ARGS)
 
        PG_RETURN_INT32(cmp);
 }
+
+/*
+ * replace_text
+ * replace all occurences of 'old_sub_str' in 'orig_str'
+ * with 'new_sub_str' to form 'new_str'
+ * 
+ * returns 'orig_str' if 'old_sub_str' == '' or 'orig_str' == ''
+ * otherwise returns 'new_str' 
+ */
+Datum
+replace_text(PG_FUNCTION_ARGS)
+{
+       text            *left_text;
+       text            *right_text;
+       text            *buf_text;
+       text            *ret_text;
+       int                     curr_posn;
+       text            *src_text = PG_GETARG_TEXT_P(0);
+       int                     src_text_len = TEXTLEN(src_text);
+       text            *from_sub_text = PG_GETARG_TEXT_P(1);
+       int                     from_sub_text_len = TEXTLEN(from_sub_text);
+       text            *to_sub_text = PG_GETARG_TEXT_P(2);
+       char            *to_sub_str = PG_TEXT_GET_STR(to_sub_text);
+       StringInfo      str = makeStringInfo();
+
+       if (src_text_len == 0 || from_sub_text_len == 0)
+               PG_RETURN_TEXT_P(src_text);
+
+       buf_text = TEXTDUP(src_text);
+       curr_posn = TEXTPOS(buf_text, from_sub_text);
+
+       while (curr_posn > 0)
+       {
+               left_text = LEFT(buf_text, from_sub_text);
+               right_text = RIGHT(buf_text, from_sub_text, from_sub_text_len);
+
+               appendStringInfo(str, PG_TEXT_GET_STR(left_text));
+               appendStringInfo(str, to_sub_str);
+
+               pfree(buf_text);
+               pfree(left_text);
+               buf_text = right_text;
+               curr_posn = TEXTPOS(buf_text, from_sub_text);
+       }
+
+       appendStringInfo(str, PG_TEXT_GET_STR(buf_text));
+       pfree(buf_text);
+
+       ret_text = PG_STR_GET_TEXT(str->data);
+       pfree(str->data);
+       pfree(str);
+
+       PG_RETURN_TEXT_P(ret_text);
+}
+
+/*
+ * split_text
+ * parse input string
+ * return ord item (1 based)
+ * based on provided field separator
+ */
+Datum
+split_text(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 = PG_GETARG_INT32(2);
+       int                     start_posn = 0;
+       int                     end_posn = 0;
+       text            *result_text;
+
+       /* return empty string for empty input string */
+       if (inputstring_len < 1)
+               PG_RETURN_TEXT_P(PG_STR_GET_TEXT(""));
+
+       /* empty field separator */
+       if (fldsep_len < 1)
+       {
+               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(""));
+       }
+
+       /* field number is 1 based */
+       if (fldnum < 1)
+               elog(ERROR, "field position must be > 0");
+
+       start_posn = text_position(PointerGetDatum(inputstring),
+                                                               PointerGetDatum(fldsep),
+                                                               fldnum - 1);
+       end_posn = text_position(PointerGetDatum(inputstring),
+                                                               PointerGetDatum(fldsep),
+                                                               fldnum);
+
+       if ((start_posn == 0) && (end_posn == 0))       /* fldsep not found */
+       {
+               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))
+       {
+               /* last field requested */
+               result_text = text_substring(PointerGetDatum(inputstring), start_posn + fldsep_len, -1, true);
+               PG_RETURN_TEXT_P(result_text);
+       }
+       else if ((start_posn == 0) && (end_posn != 0))
+       {
+               /* first field requested */
+               result_text = LEFT(inputstring, fldsep);
+               PG_RETURN_TEXT_P(result_text);
+       }
+       else
+       {
+               /* prior to last field requested */
+               result_text = text_substring(PointerGetDatum(inputstring), start_posn + fldsep_len, end_posn - start_posn - fldsep_len, false);
+               PG_RETURN_TEXT_P(result_text);
+       }
+}
+
+#define HEXBASE 16
+/*
+ * Convert a int32 to a string containing a base 16 (hex) representation of
+ * the number.
+ */
+Datum
+to_hex32(PG_FUNCTION_ARGS)
+{
+       static char             digits[] = "0123456789abcdef";
+       char                    buf[32];        /* bigger than needed, but reasonable */
+       char               *ptr,
+                                  *end;
+       text               *result_text;
+       int32                   value = PG_GETARG_INT32(0);
+
+       end = ptr = buf + sizeof(buf) - 1;
+       *ptr = '\0';
+
+       do
+       {
+               *--ptr = digits[value % HEXBASE];
+               value /= HEXBASE;
+       } while (ptr > buf && value);
+
+       result_text = PG_STR_GET_TEXT(ptr);
+       PG_RETURN_TEXT_P(result_text);
+}
+
+/*
+ * Convert a int64 to a string containing a base 16 (hex) representation of
+ * the number.
+ */
+Datum
+to_hex64(PG_FUNCTION_ARGS)
+{
+       static char             digits[] = "0123456789abcdef";
+       char                    buf[32];        /* bigger than needed, but reasonable */
+       char                    *ptr,
+                                       *end;
+       text                    *result_text;
+       int64                   value = PG_GETARG_INT64(0);
+
+       end = ptr = buf + sizeof(buf) - 1;
+       *ptr = '\0';
+
+       do
+       {
+               *--ptr = digits[value % HEXBASE];
+               value /= HEXBASE;
+       } while (ptr > buf && value);
+
+       result_text = PG_STR_GET_TEXT(ptr);
+       PG_RETURN_TEXT_P(result_text);
+}
+
index a9fe521845d21364aa868be55865a6cfb1afede7..0b45d85ec932774ccd28472ee2b127d4af1fbffb 100644 (file)
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2002, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $Id: pg_proc.h,v 1.260 2002/08/22 00:01:47 tgl Exp $
+ * $Id: pg_proc.h,v 1.261 2002/08/22 03:24:00 momjian Exp $
  *
  * NOTES
  *       The script catalog/genbki.sh reads this file and generates .bki
@@ -2120,7 +2120,7 @@ DATA(insert OID =  881 (  ltrim              PGNSP PGUID 14 f f t f i 1 25 "25"  "select
 DESCR("remove initial characters from string");
 DATA(insert OID =  882 (  rtrim                   PGNSP PGUID 14 f f t f i 1 25 "25"  "select rtrim($1, \' \')" - _null_ ));
 DESCR("remove trailing characters from string");
-DATA(insert OID =  883 (  substr          PGNSP PGUID 14 f f t f i 2 25 "25 23"        "select substr($1, $2, -1)" - _null_ ));
+DATA(insert OID =  883 (  substr          PGNSP PGUID 12 f f t f i 2 25 "25 23"        text_substr_no_len - _null_ ));
 DESCR("return portion of string");
 DATA(insert OID =  884 (  btrim                   PGNSP PGUID 12 f f t f i 2 25 "25 25"        btrim - _null_ ));
 DESCR("trim both ends of string");
@@ -2129,8 +2129,16 @@ DESCR("trim both ends of string");
 
 DATA(insert OID =  936 (  substring    PGNSP PGUID 12 f f t f i 3 25 "25 23 23"  text_substr - _null_ ));
 DESCR("return portion of string");
-DATA(insert OID =  937 (  substring    PGNSP PGUID 14 f f t f i 2 25 "25 23"   "select substring($1, $2, -1)" - _null_ ));
+DATA(insert OID =  937 (  substring    PGNSP PGUID 12 f f t f i 2 25 "25 23"   text_substr_no_len - _null_ ));
 DESCR("return portion of string");
+DATA(insert OID =  2087 ( replace      PGNSP PGUID 12 f f t f i 3 25 "25 25 25"  replace_text - _null_ ));
+DESCR("replace all occurrences of old_substr with new_substr in string");
+DATA(insert OID =  2088 ( split        PGNSP PGUID 12 f f t f i 3 25 "25 25 23"  split_text - _null_ ));
+DESCR("split string by field_sep and return field_num");
+DATA(insert OID =  2089 ( to_hex       PGNSP PGUID 12 f f t f i 1 25 "23"  to_hex32 - _null_ ));
+DESCR("convert int32 number to hex");
+DATA(insert OID =  2090 ( to_hex       PGNSP PGUID 12 f f t f i 1 25 "20"  to_hex64 - _null_ ));
+DESCR("convert int64 number to hex");
 
 /* for multi-byte support */
 
@@ -2781,7 +2789,11 @@ DATA(insert OID = 2011 (  byteacat                  PGNSP PGUID 12 f f t f i 2 17 "17 17"        byt
 DESCR("concatenate");
 DATA(insert OID = 2012 (  substring               PGNSP PGUID 12 f f t f i 3 17 "17 23 23"  bytea_substr - _null_ ));
 DESCR("return portion of string");
-DATA(insert OID = 2013 (  substring               PGNSP PGUID 14 f f t f i 2 17 "17 23"        "select substring($1, $2, -1)" - _null_ ));
+DATA(insert OID = 2013 (  substring               PGNSP PGUID 12 f f t f i 2 17 "17 23"        bytea_substr_no_len - _null_ ));
+DESCR("return portion of string");
+DATA(insert OID = 2085 (  substr                  PGNSP PGUID 12 f f t f i 3 17 "17 23 23"  bytea_substr - _null_ ));
+DESCR("return portion of string");
+DATA(insert OID = 2086 (  substr                  PGNSP PGUID 12 f f t f i 2 17 "17 23"        bytea_substr_no_len - _null_ ));
 DESCR("return portion of string");
 DATA(insert OID = 2014 (  position                PGNSP PGUID 12 f f t f i 2 23 "17 17"        byteapos - _null_ ));
 DESCR("return position of substring");
index c18a17be9ae63c125ec07c5fa2f4d523fb0062a2..bdee336b002eeb8dffb8144e36b908de3514aa9e 100644 (file)
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2002, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $Id: builtins.h,v 1.194 2002/08/22 00:01:49 tgl Exp $
+ * $Id: builtins.h,v 1.195 2002/08/22 03:24:01 momjian Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -467,12 +467,17 @@ extern Datum textlen(PG_FUNCTION_ARGS);
 extern Datum textoctetlen(PG_FUNCTION_ARGS);
 extern Datum textpos(PG_FUNCTION_ARGS);
 extern Datum text_substr(PG_FUNCTION_ARGS);
+extern Datum text_substr_no_len(PG_FUNCTION_ARGS);
 extern Datum name_text(PG_FUNCTION_ARGS);
 extern Datum text_name(PG_FUNCTION_ARGS);
 extern int     varstr_cmp(char *arg1, int len1, char *arg2, int len2);
 extern List *textToQualifiedNameList(text *textval, const char *caller);
 extern bool SplitIdentifierString(char *rawstring, char separator,
                                                                  List **namelist);
+extern Datum replace_text(PG_FUNCTION_ARGS);
+extern Datum split_text(PG_FUNCTION_ARGS);
+extern Datum to_hex32(PG_FUNCTION_ARGS);
+extern Datum to_hex64(PG_FUNCTION_ARGS);
 
 extern Datum unknownin(PG_FUNCTION_ARGS);
 extern Datum unknownout(PG_FUNCTION_ARGS);
@@ -496,6 +501,7 @@ extern Datum byteacmp(PG_FUNCTION_ARGS);
 extern Datum byteacat(PG_FUNCTION_ARGS);
 extern Datum byteapos(PG_FUNCTION_ARGS);
 extern Datum bytea_substr(PG_FUNCTION_ARGS);
+extern Datum bytea_substr_no_len(PG_FUNCTION_ARGS);
 
 /* version.c */
 extern Datum pgsql_version(PG_FUNCTION_ARGS);
index 67dbd1a2d922d1efb5557dd20a66f9b3190fc406..373746ca7ac48fc8be835eca77918b8461fa006a 100644 (file)
@@ -573,3 +573,166 @@ SELECT text 'text' || varchar ' and varchar' AS "Concat text to varchar";
  text and varchar
 (1 row)
 
+--
+-- test substr with toasted text values
+--
+CREATE TABLE toasttest(f1 text);
+insert into toasttest values(repeat('1234567890',10000));
+insert into toasttest values(repeat('1234567890',10000));
+-- If the starting position is zero or less, then return from the start of the string
+-- adjusting the length to be consistent with the "negative start" per SQL92.
+SELECT substr(f1, -1, 5) from toasttest;
+ substr 
+--------
+ 123
+ 123
+(2 rows)
+
+-- If the length is less than zero, an ERROR is thrown.
+SELECT substr(f1, 5, -1) from toasttest;
+ERROR:  negative substring length not allowed
+-- If no third argument (length) is provided, the length to the end of the
+-- string is assumed.
+SELECT substr(f1, 99995) from toasttest;
+ substr 
+--------
+ 567890
+ 567890
+(2 rows)
+
+-- If start plus length is > string length, the result is truncated to
+-- string length
+SELECT substr(f1, 99995, 10) from toasttest;
+ substr 
+--------
+ 567890
+ 567890
+(2 rows)
+
+DROP TABLE toasttest;
+--
+-- test substr with toasted bytea values
+--
+CREATE TABLE toasttest(f1 bytea);
+insert into toasttest values(decode(repeat('1234567890',10000),'escape'));
+insert into toasttest values(decode(repeat('1234567890',10000),'escape'));
+-- If the starting position is zero or less, then return from the start of the string
+-- adjusting the length to be consistent with the "negative start" per SQL92.
+SELECT substr(f1, -1, 5) from toasttest;
+ substr 
+--------
+ 123
+ 123
+(2 rows)
+
+-- If the length is less than zero, an ERROR is thrown.
+SELECT substr(f1, 5, -1) from toasttest;
+ERROR:  negative substring length not allowed
+-- If no third argument (length) is provided, the length to the end of the
+-- string is assumed.
+SELECT substr(f1, 99995) from toasttest;
+ substr 
+--------
+ 567890
+ 567890
+(2 rows)
+
+-- If start plus length is > string length, the result is truncated to
+-- string length
+SELECT substr(f1, 99995, 10) from toasttest;
+ substr 
+--------
+ 567890
+ 567890
+(2 rows)
+
+DROP TABLE toasttest;
+--
+-- test length
+--
+SELECT length('abcdef') AS "length_6";
+ length_6 
+----------
+        6
+(1 row)
+
+--
+-- test strpos
+--
+SELECT strpos('abcdef', 'cd') AS "pos_3";
+ pos_3 
+-------
+     3
+(1 row)
+
+SELECT strpos('abcdef', 'xy') AS "pos_0";
+ pos_0 
+-------
+     0
+(1 row)
+
+--
+-- test replace
+--
+SELECT replace('abcdef', 'de', '45') AS "abc45f";
+ abc45f 
+--------
+ abc45f
+(1 row)
+
+SELECT replace('yabadabadoo', 'ba', '123') AS "ya123da123doo";
+ ya123da123doo 
+---------------
+ ya123da123doo
+(1 row)
+
+SELECT replace('yabadoo', 'bad', '') AS "yaoo";
+ yaoo 
+------
+ yaoo
+(1 row)
+
+--
+-- test split
+--
+select split('joeuser@mydatabase','@',0) AS "an error";
+ERROR:  field position must be > 0
+select split('joeuser@mydatabase','@',1) AS "joeuser";
+ joeuser 
+---------
+ joeuser
+(1 row)
+
+select split('joeuser@mydatabase','@',2) AS "mydatabase";
+ mydatabase 
+------------
+ mydatabase
+(1 row)
+
+select split('joeuser@mydatabase','@',3) AS "empty string";
+ empty string 
+--------------
+(1 row)
+
+select split('@joeuser@mydatabase@','@',2) AS "joeuser";
+ joeuser 
+---------
+ joeuser
+(1 row)
+
+--
+-- test to_hex
+--
+select to_hex(256*256*256 - 1) AS "ffffff";
+ ffffff 
+--------
+ ffffff
+(1 row)
+
+select to_hex(256::bigint*256::bigint*256::bigint*256::bigint - 1) AS "ffffffff";
+ ffffffff 
+----------
+ ffffffff
+(1 row)
+
index 3127d0ebffcc5e932c72d0da9e1c26cb77f15b93..af83a0b5715b8c67f17e3db8da29698414b6ba0e 100644 (file)
@@ -197,3 +197,96 @@ SELECT char(20) 'characters' || 'and text' AS "Concat char to unknown type";
 SELECT text 'text' || char(20) ' and characters' AS "Concat text to char";
 
 SELECT text 'text' || varchar ' and varchar' AS "Concat text to varchar";
+
+--
+-- test substr with toasted text values
+--
+CREATE TABLE toasttest(f1 text);
+
+insert into toasttest values(repeat('1234567890',10000));
+insert into toasttest values(repeat('1234567890',10000));
+
+-- If the starting position is zero or less, then return from the start of the string
+-- adjusting the length to be consistent with the "negative start" per SQL92.
+SELECT substr(f1, -1, 5) from toasttest;
+
+-- If the length is less than zero, an ERROR is thrown.
+SELECT substr(f1, 5, -1) from toasttest;
+
+-- If no third argument (length) is provided, the length to the end of the
+-- string is assumed.
+SELECT substr(f1, 99995) from toasttest;
+
+-- If start plus length is > string length, the result is truncated to
+-- string length
+SELECT substr(f1, 99995, 10) from toasttest;
+
+DROP TABLE toasttest;
+
+--
+-- test substr with toasted bytea values
+--
+CREATE TABLE toasttest(f1 bytea);
+
+insert into toasttest values(decode(repeat('1234567890',10000),'escape'));
+insert into toasttest values(decode(repeat('1234567890',10000),'escape'));
+
+-- If the starting position is zero or less, then return from the start of the string
+-- adjusting the length to be consistent with the "negative start" per SQL92.
+SELECT substr(f1, -1, 5) from toasttest;
+
+-- If the length is less than zero, an ERROR is thrown.
+SELECT substr(f1, 5, -1) from toasttest;
+
+-- If no third argument (length) is provided, the length to the end of the
+-- string is assumed.
+SELECT substr(f1, 99995) from toasttest;
+
+-- If start plus length is > string length, the result is truncated to
+-- string length
+SELECT substr(f1, 99995, 10) from toasttest;
+
+DROP TABLE toasttest;
+
+--
+-- test length
+--
+
+SELECT length('abcdef') AS "length_6";
+
+--
+-- test strpos
+--
+
+SELECT strpos('abcdef', 'cd') AS "pos_3";
+
+SELECT strpos('abcdef', 'xy') AS "pos_0";
+
+--
+-- test replace
+--
+SELECT replace('abcdef', 'de', '45') AS "abc45f";
+
+SELECT replace('yabadabadoo', 'ba', '123') AS "ya123da123doo";
+
+SELECT replace('yabadoo', 'bad', '') AS "yaoo";
+
+--
+-- test split
+--
+select split('joeuser@mydatabase','@',0) AS "an error";
+
+select split('joeuser@mydatabase','@',1) AS "joeuser";
+
+select split('joeuser@mydatabase','@',2) AS "mydatabase";
+
+select split('joeuser@mydatabase','@',3) AS "empty string";
+
+select split('@joeuser@mydatabase@','@',2) AS "joeuser";
+
+--
+-- test to_hex
+--
+select to_hex(256*256*256 - 1) AS "ffffff";
+
+select to_hex(256::bigint*256::bigint*256::bigint*256::bigint - 1) AS "ffffffff";