PostgreSQL Source Code  git master
varlena.h File Reference
#include "nodes/pg_list.h"
#include "utils/sortsupport.h"
Include dependency graph for varlena.h:
This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Functions

int varstr_cmp (const char *arg1, int len1, const char *arg2, int len2, Oid collid)
 
void varstr_sortsupport (SortSupport ssup, Oid typid, Oid collid)
 
int varstr_levenshtein (const char *source, int slen, const char *target, int tlen, int ins_c, int del_c, int sub_c, bool trusted)
 
int varstr_levenshtein_less_equal (const char *source, int slen, const char *target, int tlen, int ins_c, int del_c, int sub_c, int max_d, bool trusted)
 
ListtextToQualifiedNameList (text *textval)
 
bool SplitIdentifierString (char *rawstring, char separator, List **namelist)
 
bool SplitDirectoriesString (char *rawstring, char separator, List **namelist)
 
bool SplitGUCList (char *rawstring, char separator, List **namelist)
 
textreplace_text_regexp (text *src_text, text *pattern_text, text *replace_text, int cflags, Oid collation, int search_start, int n)
 

Function Documentation

◆ replace_text_regexp()

text* replace_text_regexp ( text src_text,
text pattern_text,
text replace_text,
int  cflags,
Oid  collation,
int  search_start,
int  n 
)

Definition at line 4464 of file varlena.c.

4468 {
4469  text *ret_text;
4470  regex_t *re;
4471  int src_text_len = VARSIZE_ANY_EXHDR(src_text);
4472  int nmatches = 0;
4474  regmatch_t pmatch[10]; /* main match, plus \1 to \9 */
4475  int nmatch = lengthof(pmatch);
4476  pg_wchar *data;
4477  size_t data_len;
4478  int data_pos;
4479  char *start_ptr;
4480  int escape_status;
4481 
4482  initStringInfo(&buf);
4483 
4484  /* Convert data string to wide characters. */
4485  data = (pg_wchar *) palloc((src_text_len + 1) * sizeof(pg_wchar));
4486  data_len = pg_mb2wchar_with_len(VARDATA_ANY(src_text), data, src_text_len);
4487 
4488  /* Check whether replace_text has escapes, especially regexp submatches. */
4490 
4491  /* If no regexp submatches, we can use REG_NOSUB. */
4492  if (escape_status < 2)
4493  {
4494  cflags |= REG_NOSUB;
4495  /* Also tell pg_regexec we only want the whole-match location. */
4496  nmatch = 1;
4497  }
4498 
4499  /* Prepare the regexp. */
4500  re = RE_compile_and_cache(pattern_text, cflags, collation);
4501 
4502  /* start_ptr points to the data_pos'th character of src_text */
4503  start_ptr = (char *) VARDATA_ANY(src_text);
4504  data_pos = 0;
4505 
4506  while (search_start <= data_len)
4507  {
4508  int regexec_result;
4509 
4511 
4512  regexec_result = pg_regexec(re,
4513  data,
4514  data_len,
4515  search_start,
4516  NULL, /* no details */
4517  nmatch,
4518  pmatch,
4519  0);
4520 
4521  if (regexec_result == REG_NOMATCH)
4522  break;
4523 
4524  if (regexec_result != REG_OKAY)
4525  {
4526  char errMsg[100];
4527 
4529  pg_regerror(regexec_result, re, errMsg, sizeof(errMsg));
4530  ereport(ERROR,
4531  (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
4532  errmsg("regular expression failed: %s", errMsg)));
4533  }
4534 
4535  /*
4536  * Count matches, and decide whether to replace this match.
4537  */
4538  nmatches++;
4539  if (n > 0 && nmatches != n)
4540  {
4541  /*
4542  * No, so advance search_start, but not start_ptr/data_pos. (Thus,
4543  * we treat the matched text as if it weren't matched, and copy it
4544  * to the output later.)
4545  */
4546  search_start = pmatch[0].rm_eo;
4547  if (pmatch[0].rm_so == pmatch[0].rm_eo)
4548  search_start++;
4549  continue;
4550  }
4551 
4552  /*
4553  * Copy the text to the left of the match position. Note we are given
4554  * character not byte indexes.
4555  */
4556  if (pmatch[0].rm_so - data_pos > 0)
4557  {
4558  int chunk_len;
4559 
4560  chunk_len = charlen_to_bytelen(start_ptr,
4561  pmatch[0].rm_so - data_pos);
4562  appendBinaryStringInfo(&buf, start_ptr, chunk_len);
4563 
4564  /*
4565  * Advance start_ptr over that text, to avoid multiple rescans of
4566  * it if the replace_text contains multiple back-references.
4567  */
4568  start_ptr += chunk_len;
4569  data_pos = pmatch[0].rm_so;
4570  }
4571 
4572  /*
4573  * Copy the replace_text, processing escapes if any are present.
4574  */
4575  if (escape_status > 0)
4577  start_ptr, data_pos);
4578  else
4580 
4581  /* Advance start_ptr and data_pos over the matched text. */
4582  start_ptr += charlen_to_bytelen(start_ptr,
4583  pmatch[0].rm_eo - data_pos);
4584  data_pos = pmatch[0].rm_eo;
4585 
4586  /*
4587  * If we only want to replace one occurrence, we're done.
4588  */
4589  if (n > 0)
4590  break;
4591 
4592  /*
4593  * Advance search position. Normally we start the next search at the
4594  * end of the previous match; but if the match was of zero length, we
4595  * have to advance by one character, or we'd just find the same match
4596  * again.
4597  */
4598  search_start = data_pos;
4599  if (pmatch[0].rm_so == pmatch[0].rm_eo)
4600  search_start++;
4601  }
4602 
4603  /*
4604  * Copy the text to the right of the last match.
4605  */
4606  if (data_pos < data_len)
4607  {
4608  int chunk_len;
4609 
4610  chunk_len = ((char *) src_text + VARSIZE_ANY(src_text)) - start_ptr;
4611  appendBinaryStringInfo(&buf, start_ptr, chunk_len);
4612  }
4613 
4614  ret_text = cstring_to_text_with_len(buf.data, buf.len);
4615  pfree(buf.data);
4616  pfree(data);
4617 
4618  return ret_text;
4619 }
#define lengthof(array)
Definition: c.h:734
int errcode(int sqlerrcode)
Definition: elog.c:693
int errmsg(const char *fmt,...)
Definition: elog.c:904
#define ERROR
Definition: elog.h:33
#define ereport(elevel,...)
Definition: elog.h:143
unsigned int pg_wchar
Definition: mbprint.c:31
int pg_mb2wchar_with_len(const char *from, pg_wchar *to, int len)
Definition: mbutils.c:929
void pfree(void *pointer)
Definition: mcxt.c:1175
void * palloc(Size size)
Definition: mcxt.c:1068
#define CHECK_FOR_INTERRUPTS()
Definition: miscadmin.h:121
const void * data
static char * buf
Definition: pg_test_fsync.c:67
#define VARSIZE_ANY(PTR)
Definition: postgres.h:348
#define VARDATA_ANY(PTR)
Definition: postgres.h:361
#define VARSIZE_ANY_EXHDR(PTR)
Definition: postgres.h:354
size_t pg_regerror(int errcode, const regex_t *preg, char *errbuf, size_t errbuf_size)
Definition: regerror.c:60
#define REG_NOMATCH
Definition: regex.h:138
#define REG_OKAY
Definition: regex.h:137
#define REG_NOSUB
Definition: regex.h:107
int pg_regexec(regex_t *re, const chr *string, size_t len, size_t search_start, rm_detail_t *details, size_t nmatch, regmatch_t pmatch[], int flags)
Definition: regexec.c:176
regex_t * RE_compile_and_cache(text *text_re, int cflags, Oid collation)
Definition: regexp.c:138
void appendBinaryStringInfo(StringInfo str, const char *data, int datalen)
Definition: stringinfo.c:227
void initStringInfo(StringInfo str)
Definition: stringinfo.c:59
Definition: regex.h:56
regoff_t rm_eo
Definition: regex.h:86
regoff_t rm_so
Definition: regex.h:85
Definition: c.h:622
static void appendStringInfoText(StringInfo str, const text *t)
Definition: varlena.c:4240
static int check_replace_text_has_escape(const text *replace_text)
Definition: varlena.c:4331
static void appendStringInfoRegexpSubstr(StringInfo str, text *replace_text, regmatch_t *pmatch, char *start_ptr, int data_pos)
Definition: varlena.c:4364
static int charlen_to_bytelen(const char *p, int n)
Definition: varlena.c:785
text * cstring_to_text_with_len(const char *s, int len)
Definition: varlena.c:200
Datum replace_text(PG_FUNCTION_ARGS)
Definition: varlena.c:4254

References appendBinaryStringInfo(), appendStringInfoRegexpSubstr(), appendStringInfoText(), buf, charlen_to_bytelen(), CHECK_FOR_INTERRUPTS, check_replace_text_has_escape(), cstring_to_text_with_len(), data, ereport, errcode(), errmsg(), ERROR, initStringInfo(), lengthof, palloc(), pfree(), pg_mb2wchar_with_len(), pg_regerror(), pg_regexec(), RE_compile_and_cache(), REG_NOMATCH, REG_NOSUB, REG_OKAY, replace_text(), regmatch_t::rm_eo, regmatch_t::rm_so, VARDATA_ANY, VARSIZE_ANY, and VARSIZE_ANY_EXHDR.

Referenced by textregexreplace(), textregexreplace_extended(), and textregexreplace_noopt().

◆ SplitDirectoriesString()

bool SplitDirectoriesString ( char *  rawstring,
char  separator,
List **  namelist 
)

Definition at line 3842 of file varlena.c.

3844 {
3845  char *nextp = rawstring;
3846  bool done = false;
3847 
3848  *namelist = NIL;
3849 
3850  while (scanner_isspace(*nextp))
3851  nextp++; /* skip leading whitespace */
3852 
3853  if (*nextp == '\0')
3854  return true; /* allow empty string */
3855 
3856  /* At the top of the loop, we are at start of a new directory. */
3857  do
3858  {
3859  char *curname;
3860  char *endp;
3861 
3862  if (*nextp == '"')
3863  {
3864  /* Quoted name --- collapse quote-quote pairs */
3865  curname = nextp + 1;
3866  for (;;)
3867  {
3868  endp = strchr(nextp + 1, '"');
3869  if (endp == NULL)
3870  return false; /* mismatched quotes */
3871  if (endp[1] != '"')
3872  break; /* found end of quoted name */
3873  /* Collapse adjacent quotes into one quote, and look again */
3874  memmove(endp, endp + 1, strlen(endp));
3875  nextp = endp;
3876  }
3877  /* endp now points at the terminating quote */
3878  nextp = endp + 1;
3879  }
3880  else
3881  {
3882  /* Unquoted name --- extends to separator or end of string */
3883  curname = endp = nextp;
3884  while (*nextp && *nextp != separator)
3885  {
3886  /* trailing whitespace should not be included in name */
3887  if (!scanner_isspace(*nextp))
3888  endp = nextp + 1;
3889  nextp++;
3890  }
3891  if (curname == endp)
3892  return false; /* empty unquoted name not allowed */
3893  }
3894 
3895  while (scanner_isspace(*nextp))
3896  nextp++; /* skip trailing whitespace */
3897 
3898  if (*nextp == separator)
3899  {
3900  nextp++;
3901  while (scanner_isspace(*nextp))
3902  nextp++; /* skip leading whitespace for next */
3903  /* we expect another name, so done remains false */
3904  }
3905  else if (*nextp == '\0')
3906  done = true;
3907  else
3908  return false; /* invalid syntax */
3909 
3910  /* Now safe to overwrite separator with a null */
3911  *endp = '\0';
3912 
3913  /* Truncate path if it's overlength */
3914  if (strlen(curname) >= MAXPGPATH)
3915  curname[MAXPGPATH - 1] = '\0';
3916 
3917  /*
3918  * Finished isolating current name --- add it to list
3919  */
3920  curname = pstrdup(curname);
3921  canonicalize_path(curname);
3922  *namelist = lappend(*namelist, curname);
3923 
3924  /* Loop back if we didn't reach end of string */
3925  } while (!done);
3926 
3927  return true;
3928 }
List * lappend(List *list, void *datum)
Definition: list.c:336
char * pstrdup(const char *in)
Definition: mcxt.c:1305
#define MAXPGPATH
#define NIL
Definition: pg_list.h:65
void canonicalize_path(char *path)
Definition: path.c:264
bool scanner_isspace(char ch)
Definition: scansup.c:117

References canonicalize_path(), lappend(), MAXPGPATH, NIL, pstrdup(), and scanner_isspace().

Referenced by load_libraries(), and PostmasterMain().

◆ SplitGUCList()

bool SplitGUCList ( char *  rawstring,
char  separator,
List **  namelist 
)

Definition at line 3963 of file varlena.c.

3965 {
3966  char *nextp = rawstring;
3967  bool done = false;
3968 
3969  *namelist = NIL;
3970 
3971  while (scanner_isspace(*nextp))
3972  nextp++; /* skip leading whitespace */
3973 
3974  if (*nextp == '\0')
3975  return true; /* allow empty string */
3976 
3977  /* At the top of the loop, we are at start of a new identifier. */
3978  do
3979  {
3980  char *curname;
3981  char *endp;
3982 
3983  if (*nextp == '"')
3984  {
3985  /* Quoted name --- collapse quote-quote pairs */
3986  curname = nextp + 1;
3987  for (;;)
3988  {
3989  endp = strchr(nextp + 1, '"');
3990  if (endp == NULL)
3991  return false; /* mismatched quotes */
3992  if (endp[1] != '"')
3993  break; /* found end of quoted name */
3994  /* Collapse adjacent quotes into one quote, and look again */
3995  memmove(endp, endp + 1, strlen(endp));
3996  nextp = endp;
3997  }
3998  /* endp now points at the terminating quote */
3999  nextp = endp + 1;
4000  }
4001  else
4002  {
4003  /* Unquoted name --- extends to separator or whitespace */
4004  curname = nextp;
4005  while (*nextp && *nextp != separator &&
4006  !scanner_isspace(*nextp))
4007  nextp++;
4008  endp = nextp;
4009  if (curname == nextp)
4010  return false; /* empty unquoted name not allowed */
4011  }
4012 
4013  while (scanner_isspace(*nextp))
4014  nextp++; /* skip trailing whitespace */
4015 
4016  if (*nextp == separator)
4017  {
4018  nextp++;
4019  while (scanner_isspace(*nextp))
4020  nextp++; /* skip leading whitespace for next */
4021  /* we expect another name, so done remains false */
4022  }
4023  else if (*nextp == '\0')
4024  done = true;
4025  else
4026  return false; /* invalid syntax */
4027 
4028  /* Now safe to overwrite separator with a null */
4029  *endp = '\0';
4030 
4031  /*
4032  * Finished isolating current name --- add it to list
4033  */
4034  *namelist = lappend(*namelist, curname);
4035 
4036  /* Loop back if we didn't reach end of string */
4037  } while (!done);
4038 
4039  return true;
4040 }

References lappend(), NIL, and scanner_isspace().

Referenced by dumpFunc(), parse_hba_auth_opt(), pg_get_functiondef(), and PostmasterMain().

◆ SplitIdentifierString()

bool SplitIdentifierString ( char *  rawstring,
char  separator,
List **  namelist 
)

Definition at line 3715 of file varlena.c.

3717 {
3718  char *nextp = rawstring;
3719  bool done = false;
3720 
3721  *namelist = NIL;
3722 
3723  while (scanner_isspace(*nextp))
3724  nextp++; /* skip leading whitespace */
3725 
3726  if (*nextp == '\0')
3727  return true; /* allow empty string */
3728 
3729  /* At the top of the loop, we are at start of a new identifier. */
3730  do
3731  {
3732  char *curname;
3733  char *endp;
3734 
3735  if (*nextp == '"')
3736  {
3737  /* Quoted name --- collapse quote-quote pairs, no downcasing */
3738  curname = nextp + 1;
3739  for (;;)
3740  {
3741  endp = strchr(nextp + 1, '"');
3742  if (endp == NULL)
3743  return false; /* mismatched quotes */
3744  if (endp[1] != '"')
3745  break; /* found end of quoted name */
3746  /* Collapse adjacent quotes into one quote, and look again */
3747  memmove(endp, endp + 1, strlen(endp));
3748  nextp = endp;
3749  }
3750  /* endp now points at the terminating quote */
3751  nextp = endp + 1;
3752  }
3753  else
3754  {
3755  /* Unquoted name --- extends to separator or whitespace */
3756  char *downname;
3757  int len;
3758 
3759  curname = nextp;
3760  while (*nextp && *nextp != separator &&
3761  !scanner_isspace(*nextp))
3762  nextp++;
3763  endp = nextp;
3764  if (curname == nextp)
3765  return false; /* empty unquoted name not allowed */
3766 
3767  /*
3768  * Downcase the identifier, using same code as main lexer does.
3769  *
3770  * XXX because we want to overwrite the input in-place, we cannot
3771  * support a downcasing transformation that increases the string
3772  * length. This is not a problem given the current implementation
3773  * of downcase_truncate_identifier, but we'll probably have to do
3774  * something about this someday.
3775  */
3776  len = endp - curname;
3777  downname = downcase_truncate_identifier(curname, len, false);
3778  Assert(strlen(downname) <= len);
3779  strncpy(curname, downname, len); /* strncpy is required here */
3780  pfree(downname);
3781  }
3782 
3783  while (scanner_isspace(*nextp))
3784  nextp++; /* skip trailing whitespace */
3785 
3786  if (*nextp == separator)
3787  {
3788  nextp++;
3789  while (scanner_isspace(*nextp))
3790  nextp++; /* skip leading whitespace for next */
3791  /* we expect another name, so done remains false */
3792  }
3793  else if (*nextp == '\0')
3794  done = true;
3795  else
3796  return false; /* invalid syntax */
3797 
3798  /* Now safe to overwrite separator with a null */
3799  *endp = '\0';
3800 
3801  /* Truncate name if it's overlength */
3802  truncate_identifier(curname, strlen(curname), false);
3803 
3804  /*
3805  * Finished isolating current name --- add it to list
3806  */
3807  *namelist = lappend(*namelist, curname);
3808 
3809  /* Loop back if we didn't reach end of string */
3810  } while (!done);
3811 
3812  return true;
3813 }
Assert(fmt[strlen(fmt) - 1] !='\n')
const void size_t len
void truncate_identifier(char *ident, int len, bool warn)
Definition: scansup.c:93
char * downcase_truncate_identifier(const char *ident, int len, bool warn)
Definition: scansup.c:37

References Assert(), downcase_truncate_identifier(), lappend(), len, NIL, pfree(), scanner_isspace(), and truncate_identifier().

Referenced by check_datestyle(), check_log_destination(), check_search_path(), check_temp_tablespaces(), check_wal_consistency_checking(), ExtractExtensionList(), parse_extension_control_file(), parse_output_parameters(), parse_publication_options(), plpgsql_extra_checks_check_hook(), PrepareTempTablespaces(), recomputeNamespacePath(), stringToQualifiedNameList(), and textToQualifiedNameList().

◆ textToQualifiedNameList()

List* textToQualifiedNameList ( text textval)

Definition at line 3657 of file varlena.c.

3658 {
3659  char *rawname;
3660  List *result = NIL;
3661  List *namelist;
3662  ListCell *l;
3663 
3664  /* Convert to C string (handles possible detoasting). */
3665  /* Note we rely on being able to modify rawname below. */
3666  rawname = text_to_cstring(textval);
3667 
3668  if (!SplitIdentifierString(rawname, '.', &namelist))
3669  ereport(ERROR,
3670  (errcode(ERRCODE_INVALID_NAME),
3671  errmsg("invalid name syntax")));
3672 
3673  if (namelist == NIL)
3674  ereport(ERROR,
3675  (errcode(ERRCODE_INVALID_NAME),
3676  errmsg("invalid name syntax")));
3677 
3678  foreach(l, namelist)
3679  {
3680  char *curname = (char *) lfirst(l);
3681 
3682  result = lappend(result, makeString(pstrdup(curname)));
3683  }
3684 
3685  pfree(rawname);
3686  list_free(namelist);
3687 
3688  return result;
3689 }
void list_free(List *list)
Definition: list.c:1505
#define lfirst(lc)
Definition: pg_list.h:169
Definition: pg_list.h:51
String * makeString(char *str)
Definition: value.c:63
char * text_to_cstring(const text *t)
Definition: varlena.c:221
bool SplitIdentifierString(char *rawstring, char separator, List **namelist)
Definition: varlena.c:3715

References ereport, errcode(), errmsg(), ERROR, lappend(), lfirst, list_free(), makeString(), NIL, pfree(), pstrdup(), SplitIdentifierString(), and text_to_cstring().

Referenced by bt_metap(), bt_page_items_internal(), bt_page_stats_internal(), convert_table_name(), currtid_byrelname(), get_raw_page_internal(), get_rel_from_relname(), nextval(), pg_get_serial_sequence(), pg_get_viewdef_name(), pg_get_viewdef_name_ext(), pg_relpages(), pg_relpages_v1_5(), pgrowlocks(), pgstatindex(), pgstatindex_v1_5(), pgstattuple(), pgstattuple_v1_5(), row_security_active_name(), text_regclass(), ts_parse_byname(), and ts_token_type_byname().

◆ varstr_cmp()

int varstr_cmp ( const char *  arg1,
int  len1,
const char *  arg2,
int  len2,
Oid  collid 
)

Definition at line 1517 of file varlena.c.

1518 {
1519  int result;
1520 
1521  check_collation_set(collid);
1522 
1523  /*
1524  * Unfortunately, there is no strncoll(), so in the non-C locale case we
1525  * have to do some memory copying. This turns out to be significantly
1526  * slower, so we optimize the case where LC_COLLATE is C. We also try to
1527  * optimize relatively-short strings by avoiding palloc/pfree overhead.
1528  */
1529  if (lc_collate_is_c(collid))
1530  {
1531  result = memcmp(arg1, arg2, Min(len1, len2));
1532  if ((result == 0) && (len1 != len2))
1533  result = (len1 < len2) ? -1 : 1;
1534  }
1535  else
1536  {
1537  char a1buf[TEXTBUFLEN];
1538  char a2buf[TEXTBUFLEN];
1539  char *a1p,
1540  *a2p;
1541  pg_locale_t mylocale;
1542 
1543  mylocale = pg_newlocale_from_collation(collid);
1544 
1545  /*
1546  * memcmp() can't tell us which of two unequal strings sorts first,
1547  * but it's a cheap way to tell if they're equal. Testing shows that
1548  * memcmp() followed by strcoll() is only trivially slower than
1549  * strcoll() by itself, so we don't lose much if this doesn't work out
1550  * very often, and if it does - for example, because there are many
1551  * equal strings in the input - then we win big by avoiding expensive
1552  * collation-aware comparisons.
1553  */
1554  if (len1 == len2 && memcmp(arg1, arg2, len1) == 0)
1555  return 0;
1556 
1557 #ifdef WIN32
1558  /* Win32 does not have UTF-8, so we need to map to UTF-16 */
1559  if (GetDatabaseEncoding() == PG_UTF8
1560  && (!mylocale || mylocale->provider == COLLPROVIDER_LIBC))
1561  {
1562  int a1len;
1563  int a2len;
1564  int r;
1565 
1566  if (len1 >= TEXTBUFLEN / 2)
1567  {
1568  a1len = len1 * 2 + 2;
1569  a1p = palloc(a1len);
1570  }
1571  else
1572  {
1573  a1len = TEXTBUFLEN;
1574  a1p = a1buf;
1575  }
1576  if (len2 >= TEXTBUFLEN / 2)
1577  {
1578  a2len = len2 * 2 + 2;
1579  a2p = palloc(a2len);
1580  }
1581  else
1582  {
1583  a2len = TEXTBUFLEN;
1584  a2p = a2buf;
1585  }
1586 
1587  /* stupid Microsloth API does not work for zero-length input */
1588  if (len1 == 0)
1589  r = 0;
1590  else
1591  {
1592  r = MultiByteToWideChar(CP_UTF8, 0, arg1, len1,
1593  (LPWSTR) a1p, a1len / 2);
1594  if (!r)
1595  ereport(ERROR,
1596  (errmsg("could not convert string to UTF-16: error code %lu",
1597  GetLastError())));
1598  }
1599  ((LPWSTR) a1p)[r] = 0;
1600 
1601  if (len2 == 0)
1602  r = 0;
1603  else
1604  {
1605  r = MultiByteToWideChar(CP_UTF8, 0, arg2, len2,
1606  (LPWSTR) a2p, a2len / 2);
1607  if (!r)
1608  ereport(ERROR,
1609  (errmsg("could not convert string to UTF-16: error code %lu",
1610  GetLastError())));
1611  }
1612  ((LPWSTR) a2p)[r] = 0;
1613 
1614  errno = 0;
1615 #ifdef HAVE_LOCALE_T
1616  if (mylocale)
1617  result = wcscoll_l((LPWSTR) a1p, (LPWSTR) a2p, mylocale->info.lt);
1618  else
1619 #endif
1620  result = wcscoll((LPWSTR) a1p, (LPWSTR) a2p);
1621  if (result == 2147483647) /* _NLSCMPERROR; missing from mingw
1622  * headers */
1623  ereport(ERROR,
1624  (errmsg("could not compare Unicode strings: %m")));
1625 
1626  /* Break tie if necessary. */
1627  if (result == 0 &&
1628  (!mylocale || mylocale->deterministic))
1629  {
1630  result = memcmp(arg1, arg2, Min(len1, len2));
1631  if ((result == 0) && (len1 != len2))
1632  result = (len1 < len2) ? -1 : 1;
1633  }
1634 
1635  if (a1p != a1buf)
1636  pfree(a1p);
1637  if (a2p != a2buf)
1638  pfree(a2p);
1639 
1640  return result;
1641  }
1642 #endif /* WIN32 */
1643 
1644  if (len1 >= TEXTBUFLEN)
1645  a1p = (char *) palloc(len1 + 1);
1646  else
1647  a1p = a1buf;
1648  if (len2 >= TEXTBUFLEN)
1649  a2p = (char *) palloc(len2 + 1);
1650  else
1651  a2p = a2buf;
1652 
1653  memcpy(a1p, arg1, len1);
1654  a1p[len1] = '\0';
1655  memcpy(a2p, arg2, len2);
1656  a2p[len2] = '\0';
1657 
1658  if (mylocale)
1659  {
1660  if (mylocale->provider == COLLPROVIDER_ICU)
1661  {
1662 #ifdef USE_ICU
1663 #ifdef HAVE_UCOL_STRCOLLUTF8
1664  if (GetDatabaseEncoding() == PG_UTF8)
1665  {
1666  UErrorCode status;
1667 
1668  status = U_ZERO_ERROR;
1669  result = ucol_strcollUTF8(mylocale->info.icu.ucol,
1670  arg1, len1,
1671  arg2, len2,
1672  &status);
1673  if (U_FAILURE(status))
1674  ereport(ERROR,
1675  (errmsg("collation failed: %s", u_errorName(status))));
1676  }
1677  else
1678 #endif
1679  {
1680  int32_t ulen1,
1681  ulen2;
1682  UChar *uchar1,
1683  *uchar2;
1684 
1685  ulen1 = icu_to_uchar(&uchar1, arg1, len1);
1686  ulen2 = icu_to_uchar(&uchar2, arg2, len2);
1687 
1688  result = ucol_strcoll(mylocale->info.icu.ucol,
1689  uchar1, ulen1,
1690  uchar2, ulen2);
1691 
1692  pfree(uchar1);
1693  pfree(uchar2);
1694  }
1695 #else /* not USE_ICU */
1696  /* shouldn't happen */
1697  elog(ERROR, "unsupported collprovider: %c", mylocale->provider);
1698 #endif /* not USE_ICU */
1699  }
1700  else
1701  {
1702 #ifdef HAVE_LOCALE_T
1703  result = strcoll_l(a1p, a2p, mylocale->info.lt);
1704 #else
1705  /* shouldn't happen */
1706  elog(ERROR, "unsupported collprovider: %c", mylocale->provider);
1707 #endif
1708  }
1709  }
1710  else
1711  result = strcoll(a1p, a2p);
1712 
1713  /* Break tie if necessary. */
1714  if (result == 0 &&
1715  (!mylocale || mylocale->deterministic))
1716  result = strcmp(a1p, a2p);
1717 
1718  if (a1p != a1buf)
1719  pfree(a1p);
1720  if (a2p != a2buf)
1721  pfree(a2p);
1722  }
1723 
1724  return result;
1725 }
#define Min(x, y)
Definition: c.h:986
#define elog(elevel,...)
Definition: elog.h:218
int GetDatabaseEncoding(void)
Definition: mbutils.c:1210
bool lc_collate_is_c(Oid collation)
Definition: pg_locale.c:1336
pg_locale_t pg_newlocale_from_collation(Oid collid)
Definition: pg_locale.c:1518
static void static void status(const char *fmt,...) pg_attribute_printf(1
Definition: pg_regress.c:229
@ PG_UTF8
Definition: pg_wchar.h:230
union pg_locale_struct::@150 info
bool deterministic
Definition: pg_locale.h:87
static void check_collation_set(Oid collid)
Definition: varlena.c:1488
#define TEXTBUFLEN
Definition: varlena.c:112
#define strcoll_l
Definition: win32_port.h:424
#define wcscoll_l
Definition: win32_port.h:426

References check_collation_set(), pg_locale_struct::deterministic, elog, ereport, errmsg(), ERROR, GetDatabaseEncoding(), pg_locale_struct::info, lc_collate_is_c(), Min, palloc(), pfree(), pg_newlocale_from_collation(), PG_UTF8, pg_locale_struct::provider, status(), strcoll_l, TEXTBUFLEN, and wcscoll_l.

Referenced by bpchar_larger(), bpchar_smaller(), bpcharcmp(), bpchareq(), bpcharge(), bpchargt(), bpcharle(), bpcharlt(), bpcharne(), btnametextcmp(), bttextnamecmp(), citextcmp(), compareJsonbScalarValue(), gin_compare_jsonb(), make_greater_string(), namecmp(), nameeqtext(), namenetext(), spg_text_leaf_consistent(), text_cmp(), texteqname(), and textnename().

◆ varstr_levenshtein()

int varstr_levenshtein ( const char *  source,
int  slen,
const char *  target,
int  tlen,
int  ins_c,
int  del_c,
int  sub_c,
bool  trusted 
)

Definition at line 73 of file levenshtein.c.

78 {
79  int m,
80  n;
81  int *prev;
82  int *curr;
83  int *s_char_len = NULL;
84  int i,
85  j;
86  const char *y;
87 
88  /*
89  * For varstr_levenshtein_less_equal, we have real variables called
90  * start_column and stop_column; otherwise it's just short-hand for 0 and
91  * m.
92  */
93 #ifdef LEVENSHTEIN_LESS_EQUAL
94  int start_column,
95  stop_column;
96 
97 #undef START_COLUMN
98 #undef STOP_COLUMN
99 #define START_COLUMN start_column
100 #define STOP_COLUMN stop_column
101 #else
102 #undef START_COLUMN
103 #undef STOP_COLUMN
104 #define START_COLUMN 0
105 #define STOP_COLUMN m
106 #endif
107 
108  /* Convert string lengths (in bytes) to lengths in characters */
109  m = pg_mbstrlen_with_len(source, slen);
110  n = pg_mbstrlen_with_len(target, tlen);
111 
112  /*
113  * We can transform an empty s into t with n insertions, or a non-empty t
114  * into an empty s with m deletions.
115  */
116  if (!m)
117  return n * ins_c;
118  if (!n)
119  return m * del_c;
120 
121  /*
122  * For security concerns, restrict excessive CPU+RAM usage. (This
123  * implementation uses O(m) memory and has O(mn) complexity.) If
124  * "trusted" is true, caller is responsible for not making excessive
125  * requests, typically by using a small max_d along with strings that are
126  * bounded, though not necessarily to MAX_LEVENSHTEIN_STRLEN exactly.
127  */
128  if (!trusted &&
129  (m > MAX_LEVENSHTEIN_STRLEN ||
131  ereport(ERROR,
132  (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
133  errmsg("levenshtein argument exceeds maximum length of %d characters",
135 
136 #ifdef LEVENSHTEIN_LESS_EQUAL
137  /* Initialize start and stop columns. */
138  start_column = 0;
139  stop_column = m + 1;
140 
141  /*
142  * If max_d >= 0, determine whether the bound is impossibly tight. If so,
143  * return max_d + 1 immediately. Otherwise, determine whether it's tight
144  * enough to limit the computation we must perform. If so, figure out
145  * initial stop column.
146  */
147  if (max_d >= 0)
148  {
149  int min_theo_d; /* Theoretical minimum distance. */
150  int max_theo_d; /* Theoretical maximum distance. */
151  int net_inserts = n - m;
152 
153  min_theo_d = net_inserts < 0 ?
154  -net_inserts * del_c : net_inserts * ins_c;
155  if (min_theo_d > max_d)
156  return max_d + 1;
157  if (ins_c + del_c < sub_c)
158  sub_c = ins_c + del_c;
159  max_theo_d = min_theo_d + sub_c * Min(m, n);
160  if (max_d >= max_theo_d)
161  max_d = -1;
162  else if (ins_c + del_c > 0)
163  {
164  /*
165  * Figure out how much of the first row of the notional matrix we
166  * need to fill in. If the string is growing, the theoretical
167  * minimum distance already incorporates the cost of deleting the
168  * number of characters necessary to make the two strings equal in
169  * length. Each additional deletion forces another insertion, so
170  * the best-case total cost increases by ins_c + del_c. If the
171  * string is shrinking, the minimum theoretical cost assumes no
172  * excess deletions; that is, we're starting no further right than
173  * column n - m. If we do start further right, the best-case
174  * total cost increases by ins_c + del_c for each move right.
175  */
176  int slack_d = max_d - min_theo_d;
177  int best_column = net_inserts < 0 ? -net_inserts : 0;
178 
179  stop_column = best_column + (slack_d / (ins_c + del_c)) + 1;
180  if (stop_column > m)
181  stop_column = m + 1;
182  }
183  }
184 #endif
185 
186  /*
187  * In order to avoid calling pg_mblen() repeatedly on each character in s,
188  * we cache all the lengths before starting the main loop -- but if all
189  * the characters in both strings are single byte, then we skip this and
190  * use a fast-path in the main loop. If only one string contains
191  * multi-byte characters, we still build the array, so that the fast-path
192  * needn't deal with the case where the array hasn't been initialized.
193  */
194  if (m != slen || n != tlen)
195  {
196  int i;
197  const char *cp = source;
198 
199  s_char_len = (int *) palloc((m + 1) * sizeof(int));
200  for (i = 0; i < m; ++i)
201  {
202  s_char_len[i] = pg_mblen(cp);
203  cp += s_char_len[i];
204  }
205  s_char_len[i] = 0;
206  }
207 
208  /* One more cell for initialization column and row. */
209  ++m;
210  ++n;
211 
212  /* Previous and current rows of notional array. */
213  prev = (int *) palloc(2 * m * sizeof(int));
214  curr = prev + m;
215 
216  /*
217  * To transform the first i characters of s into the first 0 characters of
218  * t, we must perform i deletions.
219  */
220  for (i = START_COLUMN; i < STOP_COLUMN; i++)
221  prev[i] = i * del_c;
222 
223  /* Loop through rows of the notional array */
224  for (y = target, j = 1; j < n; j++)
225  {
226  int *temp;
227  const char *x = source;
228  int y_char_len = n != tlen + 1 ? pg_mblen(y) : 1;
229 
230 #ifdef LEVENSHTEIN_LESS_EQUAL
231 
232  /*
233  * In the best case, values percolate down the diagonal unchanged, so
234  * we must increment stop_column unless it's already on the right end
235  * of the array. The inner loop will read prev[stop_column], so we
236  * have to initialize it even though it shouldn't affect the result.
237  */
238  if (stop_column < m)
239  {
240  prev[stop_column] = max_d + 1;
241  ++stop_column;
242  }
243 
244  /*
245  * The main loop fills in curr, but curr[0] needs a special case: to
246  * transform the first 0 characters of s into the first j characters
247  * of t, we must perform j insertions. However, if start_column > 0,
248  * this special case does not apply.
249  */
250  if (start_column == 0)
251  {
252  curr[0] = j * ins_c;
253  i = 1;
254  }
255  else
256  i = start_column;
257 #else
258  curr[0] = j * ins_c;
259  i = 1;
260 #endif
261 
262  /*
263  * This inner loop is critical to performance, so we include a
264  * fast-path to handle the (fairly common) case where no multibyte
265  * characters are in the mix. The fast-path is entitled to assume
266  * that if s_char_len is not initialized then BOTH strings contain
267  * only single-byte characters.
268  */
269  if (s_char_len != NULL)
270  {
271  for (; i < STOP_COLUMN; i++)
272  {
273  int ins;
274  int del;
275  int sub;
276  int x_char_len = s_char_len[i - 1];
277 
278  /*
279  * Calculate costs for insertion, deletion, and substitution.
280  *
281  * When calculating cost for substitution, we compare the last
282  * character of each possibly-multibyte character first,
283  * because that's enough to rule out most mis-matches. If we
284  * get past that test, then we compare the lengths and the
285  * remaining bytes.
286  */
287  ins = prev[i] + ins_c;
288  del = curr[i - 1] + del_c;
289  if (x[x_char_len - 1] == y[y_char_len - 1]
290  && x_char_len == y_char_len &&
291  (x_char_len == 1 || rest_of_char_same(x, y, x_char_len)))
292  sub = prev[i - 1];
293  else
294  sub = prev[i - 1] + sub_c;
295 
296  /* Take the one with minimum cost. */
297  curr[i] = Min(ins, del);
298  curr[i] = Min(curr[i], sub);
299 
300  /* Point to next character. */
301  x += x_char_len;
302  }
303  }
304  else
305  {
306  for (; i < STOP_COLUMN; i++)
307  {
308  int ins;
309  int del;
310  int sub;
311 
312  /* Calculate costs for insertion, deletion, and substitution. */
313  ins = prev[i] + ins_c;
314  del = curr[i - 1] + del_c;
315  sub = prev[i - 1] + ((*x == *y) ? 0 : sub_c);
316 
317  /* Take the one with minimum cost. */
318  curr[i] = Min(ins, del);
319  curr[i] = Min(curr[i], sub);
320 
321  /* Point to next character. */
322  x++;
323  }
324  }
325 
326  /* Swap current row with previous row. */
327  temp = curr;
328  curr = prev;
329  prev = temp;
330 
331  /* Point to next character. */
332  y += y_char_len;
333 
334 #ifdef LEVENSHTEIN_LESS_EQUAL
335 
336  /*
337  * This chunk of code represents a significant performance hit if used
338  * in the case where there is no max_d bound. This is probably not
339  * because the max_d >= 0 test itself is expensive, but rather because
340  * the possibility of needing to execute this code prevents tight
341  * optimization of the loop as a whole.
342  */
343  if (max_d >= 0)
344  {
345  /*
346  * The "zero point" is the column of the current row where the
347  * remaining portions of the strings are of equal length. There
348  * are (n - 1) characters in the target string, of which j have
349  * been transformed. There are (m - 1) characters in the source
350  * string, so we want to find the value for zp where (n - 1) - j =
351  * (m - 1) - zp.
352  */
353  int zp = j - (n - m);
354 
355  /* Check whether the stop column can slide left. */
356  while (stop_column > 0)
357  {
358  int ii = stop_column - 1;
359  int net_inserts = ii - zp;
360 
361  if (prev[ii] + (net_inserts > 0 ? net_inserts * ins_c :
362  -net_inserts * del_c) <= max_d)
363  break;
364  stop_column--;
365  }
366 
367  /* Check whether the start column can slide right. */
368  while (start_column < stop_column)
369  {
370  int net_inserts = start_column - zp;
371 
372  if (prev[start_column] +
373  (net_inserts > 0 ? net_inserts * ins_c :
374  -net_inserts * del_c) <= max_d)
375  break;
376 
377  /*
378  * We'll never again update these values, so we must make sure
379  * there's nothing here that could confuse any future
380  * iteration of the outer loop.
381  */
382  prev[start_column] = max_d + 1;
383  curr[start_column] = max_d + 1;
384  if (start_column != 0)
385  source += (s_char_len != NULL) ? s_char_len[start_column - 1] : 1;
386  start_column++;
387  }
388 
389  /* If they cross, we're going to exceed the bound. */
390  if (start_column >= stop_column)
391  return max_d + 1;
392  }
393 #endif
394  }
395 
396  /*
397  * Because the final value was swapped from the previous row to the
398  * current row, that's where we'll find it.
399  */
400  return prev[m - 1];
401 }
int y
Definition: isn.c:72
int x
Definition: isn.c:71
int j
Definition: isn.c:74
int i
Definition: isn.c:73
#define START_COLUMN
#define MAX_LEVENSHTEIN_STRLEN
Definition: levenshtein.c:26
#define STOP_COLUMN
int pg_mbstrlen_with_len(const char *mbstr, int limit)
Definition: mbutils.c:1000
int pg_mblen(const char *mbstr)
Definition: mbutils.c:966
static rewind_source * source
Definition: pg_rewind.c:81
static bool rest_of_char_same(const char *s1, const char *s2, int len)
Definition: varlena.c:6186

References ereport, errcode(), errmsg(), ERROR, i, j, MAX_LEVENSHTEIN_STRLEN, Min, palloc(), pg_mblen(), pg_mbstrlen_with_len(), rest_of_char_same(), source, START_COLUMN, STOP_COLUMN, x, and y.

Referenced by levenshtein(), and levenshtein_with_costs().

◆ varstr_levenshtein_less_equal()

int varstr_levenshtein_less_equal ( const char *  source,
int  slen,
const char *  target,
int  tlen,
int  ins_c,
int  del_c,
int  sub_c,
int  max_d,
bool  trusted 
)

◆ varstr_sortsupport()

void varstr_sortsupport ( SortSupport  ssup,
Oid  typid,
Oid  collid 
)

Definition at line 2011 of file varlena.c.

2012 {
2013  bool abbreviate = ssup->abbreviate;
2014  bool collate_c = false;
2015  VarStringSortSupport *sss;
2016  pg_locale_t locale = 0;
2017 
2018  check_collation_set(collid);
2019 
2020  /*
2021  * If possible, set ssup->comparator to a function which can be used to
2022  * directly compare two datums. If we can do this, we'll avoid the
2023  * overhead of a trip through the fmgr layer for every comparison, which
2024  * can be substantial.
2025  *
2026  * Most typically, we'll set the comparator to varlenafastcmp_locale,
2027  * which uses strcoll() to perform comparisons. We use that for the
2028  * BpChar case too, but type NAME uses namefastcmp_locale. However, if
2029  * LC_COLLATE = C, we can make things quite a bit faster with
2030  * varstrfastcmp_c, bpcharfastcmp_c, or namefastcmp_c, all of which use
2031  * memcmp() rather than strcoll().
2032  */
2033  if (lc_collate_is_c(collid))
2034  {
2035  if (typid == BPCHAROID)
2036  ssup->comparator = bpcharfastcmp_c;
2037  else if (typid == NAMEOID)
2038  {
2039  ssup->comparator = namefastcmp_c;
2040  /* Not supporting abbreviation with type NAME, for now */
2041  abbreviate = false;
2042  }
2043  else
2044  ssup->comparator = varstrfastcmp_c;
2045 
2046  collate_c = true;
2047  }
2048  else
2049  {
2050  /*
2051  * We need a collation-sensitive comparison. To make things faster,
2052  * we'll figure out the collation based on the locale id and cache the
2053  * result.
2054  */
2056 
2057  /*
2058  * There is a further exception on Windows. When the database
2059  * encoding is UTF-8 and we are not using the C collation, complex
2060  * hacks are required. We don't currently have a comparator that
2061  * handles that case, so we fall back on the slow method of having the
2062  * sort code invoke bttextcmp() (in the case of text) via the fmgr
2063  * trampoline. ICU locales work just the same on Windows, however.
2064  */
2065 #ifdef WIN32
2066  if (GetDatabaseEncoding() == PG_UTF8 &&
2067  !(locale && locale->provider == COLLPROVIDER_ICU))
2068  return;
2069 #endif
2070 
2071  /*
2072  * We use varlenafastcmp_locale except for type NAME.
2073  */
2074  if (typid == NAMEOID)
2075  {
2077  /* Not supporting abbreviation with type NAME, for now */
2078  abbreviate = false;
2079  }
2080  else
2082  }
2083 
2084  /*
2085  * Unfortunately, it seems that abbreviation for non-C collations is
2086  * broken on many common platforms; testing of multiple versions of glibc
2087  * reveals that, for many locales, strcoll() and strxfrm() do not return
2088  * consistent results, which is fatal to this optimization. While no
2089  * other libc other than Cygwin has so far been shown to have a problem,
2090  * we take the conservative course of action for right now and disable
2091  * this categorically. (Users who are certain this isn't a problem on
2092  * their system can define TRUST_STRXFRM.)
2093  *
2094  * Even apart from the risk of broken locales, it's possible that there
2095  * are platforms where the use of abbreviated keys should be disabled at
2096  * compile time. Having only 4 byte datums could make worst-case
2097  * performance drastically more likely, for example. Moreover, macOS's
2098  * strxfrm() implementation is known to not effectively concentrate a
2099  * significant amount of entropy from the original string in earlier
2100  * transformed blobs. It's possible that other supported platforms are
2101  * similarly encumbered. So, if we ever get past disabling this
2102  * categorically, we may still want or need to disable it for particular
2103  * platforms.
2104  */
2105 #ifndef TRUST_STRXFRM
2106  if (!collate_c && !(locale && locale->provider == COLLPROVIDER_ICU))
2107  abbreviate = false;
2108 #endif
2109 
2110  /*
2111  * If we're using abbreviated keys, or if we're using a locale-aware
2112  * comparison, we need to initialize a VarStringSortSupport object. Both
2113  * cases will make use of the temporary buffers we initialize here for
2114  * scratch space (and to detect requirement for BpChar semantics from
2115  * caller), and the abbreviation case requires additional state.
2116  */
2117  if (abbreviate || !collate_c)
2118  {
2119  sss = palloc(sizeof(VarStringSortSupport));
2120  sss->buf1 = palloc(TEXTBUFLEN);
2121  sss->buflen1 = TEXTBUFLEN;
2122  sss->buf2 = palloc(TEXTBUFLEN);
2123  sss->buflen2 = TEXTBUFLEN;
2124  /* Start with invalid values */
2125  sss->last_len1 = -1;
2126  sss->last_len2 = -1;
2127  /* Initialize */
2128  sss->last_returned = 0;
2129  sss->locale = locale;
2130 
2131  /*
2132  * To avoid somehow confusing a strxfrm() blob and an original string,
2133  * constantly keep track of the variety of data that buf1 and buf2
2134  * currently contain.
2135  *
2136  * Comparisons may be interleaved with conversion calls. Frequently,
2137  * conversions and comparisons are batched into two distinct phases,
2138  * but the correctness of caching cannot hinge upon this. For
2139  * comparison caching, buffer state is only trusted if cache_blob is
2140  * found set to false, whereas strxfrm() caching only trusts the state
2141  * when cache_blob is found set to true.
2142  *
2143  * Arbitrarily initialize cache_blob to true.
2144  */
2145  sss->cache_blob = true;
2146  sss->collate_c = collate_c;
2147  sss->typid = typid;
2148  ssup->ssup_extra = sss;
2149 
2150  /*
2151  * If possible, plan to use the abbreviated keys optimization. The
2152  * core code may switch back to authoritative comparator should
2153  * abbreviation be aborted.
2154  */
2155  if (abbreviate)
2156  {
2157  sss->prop_card = 0.20;
2158  initHyperLogLog(&sss->abbr_card, 10);
2159  initHyperLogLog(&sss->full_card, 10);
2160  ssup->abbrev_full_comparator = ssup->comparator;
2164  }
2165  }
2166 }
void initHyperLogLog(hyperLogLogState *cState, uint8 bwidth)
Definition: hyperloglog.c:66
static char * locale
Definition: initdb.c:128
int(* comparator)(Datum x, Datum y, SortSupport ssup)
Definition: sortsupport.h:106
Datum(* abbrev_converter)(Datum original, SortSupport ssup)
Definition: sortsupport.h:172
void * ssup_extra
Definition: sortsupport.h:87
int(* abbrev_full_comparator)(Datum x, Datum y, SortSupport ssup)
Definition: sortsupport.h:191
bool(* abbrev_abort)(int memtupcount, SortSupport ssup)
Definition: sortsupport.h:182
pg_locale_t locale
Definition: varlena.c:94
hyperLogLogState full_card
Definition: varlena.c:92
hyperLogLogState abbr_card
Definition: varlena.c:91
int ssup_datum_unsigned_cmp(Datum x, Datum y, SortSupport ssup)
Definition: tuplesort.c:4904
static bool varstr_abbrev_abort(int memtupcount, SortSupport ssup)
Definition: varlena.c:2695
static int varlenafastcmp_locale(Datum x, Datum y, SortSupport ssup)
Definition: varlena.c:2254
static int bpcharfastcmp_c(Datum x, Datum y, SortSupport ssup)
Definition: varlena.c:2209
static int namefastcmp_c(Datum x, Datum y, SortSupport ssup)
Definition: varlena.c:2242
static int namefastcmp_locale(Datum x, Datum y, SortSupport ssup)
Definition: varlena.c:2285
static Datum varstr_abbrev_convert(Datum original, SortSupport ssup)
Definition: varlena.c:2455
static int varstrfastcmp_c(Datum x, Datum y, SortSupport ssup)
Definition: varlena.c:2172

References VarStringSortSupport::abbr_card, SortSupportData::abbrev_abort, SortSupportData::abbrev_converter, SortSupportData::abbrev_full_comparator, SortSupportData::abbreviate, bpcharfastcmp_c(), VarStringSortSupport::buf1, VarStringSortSupport::buf2, VarStringSortSupport::buflen1, VarStringSortSupport::buflen2, VarStringSortSupport::cache_blob, check_collation_set(), VarStringSortSupport::collate_c, SortSupportData::comparator, VarStringSortSupport::full_card, GetDatabaseEncoding(), initHyperLogLog(), VarStringSortSupport::last_len1, VarStringSortSupport::last_len2, VarStringSortSupport::last_returned, lc_collate_is_c(), VarStringSortSupport::locale, locale, namefastcmp_c(), namefastcmp_locale(), palloc(), pg_newlocale_from_collation(), PG_UTF8, VarStringSortSupport::prop_card, ssup_datum_unsigned_cmp(), SortSupportData::ssup_extra, TEXTBUFLEN, VarStringSortSupport::typid, varlenafastcmp_locale(), varstr_abbrev_abort(), varstr_abbrev_convert(), and varstrfastcmp_c().

Referenced by bpchar_sortsupport(), btbpchar_pattern_sortsupport(), btnamesortsupport(), bttext_pattern_sortsupport(), bttextsortsupport(), and bytea_sortsupport().