PostgreSQL Source Code  git master
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros
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 (char *arg1, int len1, char *arg2, int len2, Oid collid)
 
void varstr_sortsupport (SortSupport ssup, Oid collid, bool bpchar)
 
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)
 
textreplace_text_regexp (text *src_text, void *regexp, text *replace_text, bool glob)
 

Function Documentation

text* replace_text_regexp ( text src_text,
void *  regexp,
text replace_text,
bool  glob 
)

Definition at line 3893 of file varlena.c.

References appendBinaryStringInfo(), appendStringInfoRegexpSubstr(), appendStringInfoText(), buf, charlen_to_bytelen(), CHECK_FOR_INTERRUPTS, check_replace_text_has_escape_char(), cstring_to_text_with_len(), StringInfoData::data, ereport, errcode(), errmsg(), ERROR, initStringInfo(), StringInfoData::len, NULL, palloc(), pfree(), pg_mb2wchar_with_len(), pg_regerror(), pg_regexec(), REG_NOMATCH, REG_OKAY, REGEXP_REPLACE_BACKREF_CNT, regmatch_t::rm_eo, regmatch_t::rm_so, VARDATA_ANY, VARSIZE_ANY, and VARSIZE_ANY_EXHDR.

Referenced by textregexreplace(), and textregexreplace_noopt().

3895 {
3896  text *ret_text;
3897  regex_t *re = (regex_t *) regexp;
3898  int src_text_len = VARSIZE_ANY_EXHDR(src_text);
3901  pg_wchar *data;
3902  size_t data_len;
3903  int search_start;
3904  int data_pos;
3905  char *start_ptr;
3906  bool have_escape;
3907 
3908  initStringInfo(&buf);
3909 
3910  /* Convert data string to wide characters. */
3911  data = (pg_wchar *) palloc((src_text_len + 1) * sizeof(pg_wchar));
3912  data_len = pg_mb2wchar_with_len(VARDATA_ANY(src_text), data, src_text_len);
3913 
3914  /* Check whether replace_text has escape char. */
3915  have_escape = check_replace_text_has_escape_char(replace_text);
3916 
3917  /* start_ptr points to the data_pos'th character of src_text */
3918  start_ptr = (char *) VARDATA_ANY(src_text);
3919  data_pos = 0;
3920 
3921  search_start = 0;
3922  while (search_start <= data_len)
3923  {
3924  int regexec_result;
3925 
3927 
3928  regexec_result = pg_regexec(re,
3929  data,
3930  data_len,
3931  search_start,
3932  NULL, /* no details */
3934  pmatch,
3935  0);
3936 
3937  if (regexec_result == REG_NOMATCH)
3938  break;
3939 
3940  if (regexec_result != REG_OKAY)
3941  {
3942  char errMsg[100];
3943 
3945  pg_regerror(regexec_result, re, errMsg, sizeof(errMsg));
3946  ereport(ERROR,
3947  (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
3948  errmsg("regular expression failed: %s", errMsg)));
3949  }
3950 
3951  /*
3952  * Copy the text to the left of the match position. Note we are given
3953  * character not byte indexes.
3954  */
3955  if (pmatch[0].rm_so - data_pos > 0)
3956  {
3957  int chunk_len;
3958 
3959  chunk_len = charlen_to_bytelen(start_ptr,
3960  pmatch[0].rm_so - data_pos);
3961  appendBinaryStringInfo(&buf, start_ptr, chunk_len);
3962 
3963  /*
3964  * Advance start_ptr over that text, to avoid multiple rescans of
3965  * it if the replace_text contains multiple back-references.
3966  */
3967  start_ptr += chunk_len;
3968  data_pos = pmatch[0].rm_so;
3969  }
3970 
3971  /*
3972  * Copy the replace_text. Process back references when the
3973  * replace_text has escape characters.
3974  */
3975  if (have_escape)
3976  appendStringInfoRegexpSubstr(&buf, replace_text, pmatch,
3977  start_ptr, data_pos);
3978  else
3979  appendStringInfoText(&buf, replace_text);
3980 
3981  /* Advance start_ptr and data_pos over the matched text. */
3982  start_ptr += charlen_to_bytelen(start_ptr,
3983  pmatch[0].rm_eo - data_pos);
3984  data_pos = pmatch[0].rm_eo;
3985 
3986  /*
3987  * When global option is off, replace the first instance only.
3988  */
3989  if (!glob)
3990  break;
3991 
3992  /*
3993  * Advance search position. Normally we start the next search at the
3994  * end of the previous match; but if the match was of zero length, we
3995  * have to advance by one character, or we'd just find the same match
3996  * again.
3997  */
3998  search_start = data_pos;
3999  if (pmatch[0].rm_so == pmatch[0].rm_eo)
4000  search_start++;
4001  }
4002 
4003  /*
4004  * Copy the text to the right of the last match.
4005  */
4006  if (data_pos < data_len)
4007  {
4008  int chunk_len;
4009 
4010  chunk_len = ((char *) src_text + VARSIZE_ANY(src_text)) - start_ptr;
4011  appendBinaryStringInfo(&buf, start_ptr, chunk_len);
4012  }
4013 
4014  ret_text = cstring_to_text_with_len(buf.data, buf.len);
4015  pfree(buf.data);
4016  pfree(data);
4017 
4018  return ret_text;
4019 }
#define VARDATA_ANY(PTR)
Definition: postgres.h:347
regoff_t rm_so
Definition: regex.h:85
int errcode(int sqlerrcode)
Definition: elog.c:575
regoff_t rm_eo
Definition: regex.h:86
void pfree(void *pointer)
Definition: mcxt.c:950
#define REG_OKAY
Definition: regex.h:137
#define ERROR
Definition: elog.h:43
static bool check_replace_text_has_escape_char(const text *replace_text)
Definition: varlena.c:3755
static int charlen_to_bytelen(const char *p, int n)
Definition: varlena.c:735
static void appendStringInfoText(StringInfo str, const text *t)
Definition: varlena.c:3658
static char * buf
Definition: pg_test_fsync.c:66
text * cstring_to_text_with_len(const char *s, int len)
Definition: varlena.c:161
size_t pg_regerror(int errcode, const regex_t *preg, char *errbuf, size_t errbuf_size)
Definition: regerror.c:60
#define ereport(elevel, rest)
Definition: elog.h:122
unsigned int pg_wchar
Definition: mbprint.c:31
void initStringInfo(StringInfo str)
Definition: stringinfo.c:46
#define REGEXP_REPLACE_BACKREF_CNT
Definition: varlena.c:3882
#define VARSIZE_ANY(PTR)
Definition: postgres.h:334
int pg_mb2wchar_with_len(const char *from, pg_wchar *to, int len)
Definition: mbutils.c:734
#define NULL
Definition: c.h:229
static void appendStringInfoRegexpSubstr(StringInfo str, text *replace_text, regmatch_t *pmatch, char *start_ptr, int data_pos)
Definition: varlena.c:3788
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:172
#define VARSIZE_ANY_EXHDR(PTR)
Definition: postgres.h:340
void * palloc(Size size)
Definition: mcxt.c:849
int errmsg(const char *fmt,...)
Definition: elog.c:797
#define REG_NOMATCH
Definition: regex.h:138
Definition: c.h:439
#define CHECK_FOR_INTERRUPTS()
Definition: miscadmin.h:97
void appendBinaryStringInfo(StringInfo str, const char *data, int datalen)
Definition: stringinfo.c:208
Definition: regex.h:55
bool SplitDirectoriesString ( char *  rawstring,
char  separator,
List **  namelist 
)

Definition at line 3372 of file varlena.c.

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

Referenced by PostmasterMain().

3374 {
3375  char *nextp = rawstring;
3376  bool done = false;
3377 
3378  *namelist = NIL;
3379 
3380  while (scanner_isspace(*nextp))
3381  nextp++; /* skip leading whitespace */
3382 
3383  if (*nextp == '\0')
3384  return true; /* allow empty string */
3385 
3386  /* At the top of the loop, we are at start of a new directory. */
3387  do
3388  {
3389  char *curname;
3390  char *endp;
3391 
3392  if (*nextp == '"')
3393  {
3394  /* Quoted name --- collapse quote-quote pairs */
3395  curname = nextp + 1;
3396  for (;;)
3397  {
3398  endp = strchr(nextp + 1, '"');
3399  if (endp == NULL)
3400  return false; /* mismatched quotes */
3401  if (endp[1] != '"')
3402  break; /* found end of quoted name */
3403  /* Collapse adjacent quotes into one quote, and look again */
3404  memmove(endp, endp + 1, strlen(endp));
3405  nextp = endp;
3406  }
3407  /* endp now points at the terminating quote */
3408  nextp = endp + 1;
3409  }
3410  else
3411  {
3412  /* Unquoted name --- extends to separator or end of string */
3413  curname = endp = nextp;
3414  while (*nextp && *nextp != separator)
3415  {
3416  /* trailing whitespace should not be included in name */
3417  if (!scanner_isspace(*nextp))
3418  endp = nextp + 1;
3419  nextp++;
3420  }
3421  if (curname == endp)
3422  return false; /* empty unquoted name not allowed */
3423  }
3424 
3425  while (scanner_isspace(*nextp))
3426  nextp++; /* skip trailing whitespace */
3427 
3428  if (*nextp == separator)
3429  {
3430  nextp++;
3431  while (scanner_isspace(*nextp))
3432  nextp++; /* skip leading whitespace for next */
3433  /* we expect another name, so done remains false */
3434  }
3435  else if (*nextp == '\0')
3436  done = true;
3437  else
3438  return false; /* invalid syntax */
3439 
3440  /* Now safe to overwrite separator with a null */
3441  *endp = '\0';
3442 
3443  /* Truncate path if it's overlength */
3444  if (strlen(curname) >= MAXPGPATH)
3445  curname[MAXPGPATH - 1] = '\0';
3446 
3447  /*
3448  * Finished isolating current name --- add it to list
3449  */
3450  curname = pstrdup(curname);
3451  canonicalize_path(curname);
3452  *namelist = lappend(*namelist, curname);
3453 
3454  /* Loop back if we didn't reach end of string */
3455  } while (!done);
3456 
3457  return true;
3458 }
#define NIL
Definition: pg_list.h:69
char * pstrdup(const char *in)
Definition: mcxt.c:1077
void canonicalize_path(char *path)
Definition: path.c:254
#define MAXPGPATH
#define memmove(d, s, c)
Definition: c.h:1058
List * lappend(List *list, void *datum)
Definition: list.c:128
bool scanner_isspace(char ch)
Definition: scansup.c:221
#define NULL
Definition: c.h:229
bool SplitIdentifierString ( char *  rawstring,
char  separator,
List **  namelist 
)

Definition at line 3247 of file varlena.c.

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

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

3249 {
3250  char *nextp = rawstring;
3251  bool done = false;
3252 
3253  *namelist = NIL;
3254 
3255  while (scanner_isspace(*nextp))
3256  nextp++; /* skip leading whitespace */
3257 
3258  if (*nextp == '\0')
3259  return true; /* allow empty string */
3260 
3261  /* At the top of the loop, we are at start of a new identifier. */
3262  do
3263  {
3264  char *curname;
3265  char *endp;
3266 
3267  if (*nextp == '"')
3268  {
3269  /* Quoted name --- collapse quote-quote pairs, no downcasing */
3270  curname = nextp + 1;
3271  for (;;)
3272  {
3273  endp = strchr(nextp + 1, '"');
3274  if (endp == NULL)
3275  return false; /* mismatched quotes */
3276  if (endp[1] != '"')
3277  break; /* found end of quoted name */
3278  /* Collapse adjacent quotes into one quote, and look again */
3279  memmove(endp, endp + 1, strlen(endp));
3280  nextp = endp;
3281  }
3282  /* endp now points at the terminating quote */
3283  nextp = endp + 1;
3284  }
3285  else
3286  {
3287  /* Unquoted name --- extends to separator or whitespace */
3288  char *downname;
3289  int len;
3290 
3291  curname = nextp;
3292  while (*nextp && *nextp != separator &&
3293  !scanner_isspace(*nextp))
3294  nextp++;
3295  endp = nextp;
3296  if (curname == nextp)
3297  return false; /* empty unquoted name not allowed */
3298 
3299  /*
3300  * Downcase the identifier, using same code as main lexer does.
3301  *
3302  * XXX because we want to overwrite the input in-place, we cannot
3303  * support a downcasing transformation that increases the string
3304  * length. This is not a problem given the current implementation
3305  * of downcase_truncate_identifier, but we'll probably have to do
3306  * something about this someday.
3307  */
3308  len = endp - curname;
3309  downname = downcase_truncate_identifier(curname, len, false);
3310  Assert(strlen(downname) <= len);
3311  strncpy(curname, downname, len); /* strncpy is required here */
3312  pfree(downname);
3313  }
3314 
3315  while (scanner_isspace(*nextp))
3316  nextp++; /* skip trailing whitespace */
3317 
3318  if (*nextp == separator)
3319  {
3320  nextp++;
3321  while (scanner_isspace(*nextp))
3322  nextp++; /* skip leading whitespace for next */
3323  /* we expect another name, so done remains false */
3324  }
3325  else if (*nextp == '\0')
3326  done = true;
3327  else
3328  return false; /* invalid syntax */
3329 
3330  /* Now safe to overwrite separator with a null */
3331  *endp = '\0';
3332 
3333  /* Truncate name if it's overlength */
3334  truncate_identifier(curname, strlen(curname), false);
3335 
3336  /*
3337  * Finished isolating current name --- add it to list
3338  */
3339  *namelist = lappend(*namelist, curname);
3340 
3341  /* Loop back if we didn't reach end of string */
3342  } while (!done);
3343 
3344  return true;
3345 }
#define NIL
Definition: pg_list.h:69
char * downcase_truncate_identifier(const char *ident, int len, bool warn)
Definition: scansup.c:131
void truncate_identifier(char *ident, int len, bool warn)
Definition: scansup.c:187
void pfree(void *pointer)
Definition: mcxt.c:950
#define memmove(d, s, c)
Definition: c.h:1058
List * lappend(List *list, void *datum)
Definition: list.c:128
bool scanner_isspace(char ch)
Definition: scansup.c:221
#define NULL
Definition: c.h:229
#define Assert(condition)
Definition: c.h:675
List* textToQualifiedNameList ( text textval)

Definition at line 3189 of file varlena.c.

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

Referenced by bt_metap(), bt_page_items(), bt_page_stats(), 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().

3190 {
3191  char *rawname;
3192  List *result = NIL;
3193  List *namelist;
3194  ListCell *l;
3195 
3196  /* Convert to C string (handles possible detoasting). */
3197  /* Note we rely on being able to modify rawname below. */
3198  rawname = text_to_cstring(textval);
3199 
3200  if (!SplitIdentifierString(rawname, '.', &namelist))
3201  ereport(ERROR,
3202  (errcode(ERRCODE_INVALID_NAME),
3203  errmsg("invalid name syntax")));
3204 
3205  if (namelist == NIL)
3206  ereport(ERROR,
3207  (errcode(ERRCODE_INVALID_NAME),
3208  errmsg("invalid name syntax")));
3209 
3210  foreach(l, namelist)
3211  {
3212  char *curname = (char *) lfirst(l);
3213 
3214  result = lappend(result, makeString(pstrdup(curname)));
3215  }
3216 
3217  pfree(rawname);
3218  list_free(namelist);
3219 
3220  return result;
3221 }
Value * makeString(char *str)
Definition: value.c:53
#define NIL
Definition: pg_list.h:69
char * pstrdup(const char *in)
Definition: mcxt.c:1077
int errcode(int sqlerrcode)
Definition: elog.c:575
return result
Definition: formatting.c:1632
void pfree(void *pointer)
Definition: mcxt.c:950
#define ERROR
Definition: elog.h:43
bool SplitIdentifierString(char *rawstring, char separator, List **namelist)
Definition: varlena.c:3247
#define ereport(elevel, rest)
Definition: elog.h:122
List * lappend(List *list, void *datum)
Definition: list.c:128
#define lfirst(lc)
Definition: pg_list.h:106
char * text_to_cstring(const text *t)
Definition: varlena.c:182
int errmsg(const char *fmt,...)
Definition: elog.c:797
void list_free(List *list)
Definition: list.c:1133
Definition: pg_list.h:45
int varstr_cmp ( char *  arg1,
int  len1,
char *  arg2,
int  len2,
Oid  collid 
)

Definition at line 1382 of file varlena.c.

References COLLPROVIDER_ICU, DEFAULT_COLLATION_OID, elog, ereport, errcode(), errhint(), errmsg(), ERROR, GetDatabaseEncoding(), pg_locale_struct::info, lc_collate_is_c(), Min, OidIsValid, palloc(), pfree(), pg_newlocale_from_collation(), PG_UTF8, pg_locale_struct::provider, result, status(), strcoll_l, TEXTBUFLEN, and wcscoll_l.

Referenced by bpchar_larger(), bpchar_smaller(), bpcharcmp(), bpcharge(), bpchargt(), bpcharle(), bpcharlt(), citextcmp(), compareJsonbScalarValue(), gin_compare_jsonb(), make_greater_string(), spg_text_leaf_consistent(), and text_cmp().

1383 {
1384  int result;
1385 
1386  /*
1387  * Unfortunately, there is no strncoll(), so in the non-C locale case we
1388  * have to do some memory copying. This turns out to be significantly
1389  * slower, so we optimize the case where LC_COLLATE is C. We also try to
1390  * optimize relatively-short strings by avoiding palloc/pfree overhead.
1391  */
1392  if (lc_collate_is_c(collid))
1393  {
1394  result = memcmp(arg1, arg2, Min(len1, len2));
1395  if ((result == 0) && (len1 != len2))
1396  result = (len1 < len2) ? -1 : 1;
1397  }
1398  else
1399  {
1400  char a1buf[TEXTBUFLEN];
1401  char a2buf[TEXTBUFLEN];
1402  char *a1p,
1403  *a2p;
1404  pg_locale_t mylocale = 0;
1405 
1406  if (collid != DEFAULT_COLLATION_OID)
1407  {
1408  if (!OidIsValid(collid))
1409  {
1410  /*
1411  * This typically means that the parser could not resolve a
1412  * conflict of implicit collations, so report it that way.
1413  */
1414  ereport(ERROR,
1415  (errcode(ERRCODE_INDETERMINATE_COLLATION),
1416  errmsg("could not determine which collation to use for string comparison"),
1417  errhint("Use the COLLATE clause to set the collation explicitly.")));
1418  }
1419  mylocale = pg_newlocale_from_collation(collid);
1420  }
1421 
1422  /*
1423  * memcmp() can't tell us which of two unequal strings sorts first,
1424  * but it's a cheap way to tell if they're equal. Testing shows that
1425  * memcmp() followed by strcoll() is only trivially slower than
1426  * strcoll() by itself, so we don't lose much if this doesn't work out
1427  * very often, and if it does - for example, because there are many
1428  * equal strings in the input - then we win big by avoiding expensive
1429  * collation-aware comparisons.
1430  */
1431  if (len1 == len2 && memcmp(arg1, arg2, len1) == 0)
1432  return 0;
1433 
1434 #ifdef WIN32
1435  /* Win32 does not have UTF-8, so we need to map to UTF-16 */
1436  if (GetDatabaseEncoding() == PG_UTF8)
1437  {
1438  int a1len;
1439  int a2len;
1440  int r;
1441 
1442  if (len1 >= TEXTBUFLEN / 2)
1443  {
1444  a1len = len1 * 2 + 2;
1445  a1p = palloc(a1len);
1446  }
1447  else
1448  {
1449  a1len = TEXTBUFLEN;
1450  a1p = a1buf;
1451  }
1452  if (len2 >= TEXTBUFLEN / 2)
1453  {
1454  a2len = len2 * 2 + 2;
1455  a2p = palloc(a2len);
1456  }
1457  else
1458  {
1459  a2len = TEXTBUFLEN;
1460  a2p = a2buf;
1461  }
1462 
1463  /* stupid Microsloth API does not work for zero-length input */
1464  if (len1 == 0)
1465  r = 0;
1466  else
1467  {
1468  r = MultiByteToWideChar(CP_UTF8, 0, arg1, len1,
1469  (LPWSTR) a1p, a1len / 2);
1470  if (!r)
1471  ereport(ERROR,
1472  (errmsg("could not convert string to UTF-16: error code %lu",
1473  GetLastError())));
1474  }
1475  ((LPWSTR) a1p)[r] = 0;
1476 
1477  if (len2 == 0)
1478  r = 0;
1479  else
1480  {
1481  r = MultiByteToWideChar(CP_UTF8, 0, arg2, len2,
1482  (LPWSTR) a2p, a2len / 2);
1483  if (!r)
1484  ereport(ERROR,
1485  (errmsg("could not convert string to UTF-16: error code %lu",
1486  GetLastError())));
1487  }
1488  ((LPWSTR) a2p)[r] = 0;
1489 
1490  errno = 0;
1491 #ifdef HAVE_LOCALE_T
1492  if (mylocale)
1493  result = wcscoll_l((LPWSTR) a1p, (LPWSTR) a2p, mylocale->info.lt);
1494  else
1495 #endif
1496  result = wcscoll((LPWSTR) a1p, (LPWSTR) a2p);
1497  if (result == 2147483647) /* _NLSCMPERROR; missing from mingw
1498  * headers */
1499  ereport(ERROR,
1500  (errmsg("could not compare Unicode strings: %m")));
1501 
1502  /*
1503  * In some locales wcscoll() can claim that nonidentical strings
1504  * are equal. Believing that would be bad news for a number of
1505  * reasons, so we follow Perl's lead and sort "equal" strings
1506  * according to strcmp (on the UTF-8 representation).
1507  */
1508  if (result == 0)
1509  {
1510  result = memcmp(arg1, arg2, Min(len1, len2));
1511  if ((result == 0) && (len1 != len2))
1512  result = (len1 < len2) ? -1 : 1;
1513  }
1514 
1515  if (a1p != a1buf)
1516  pfree(a1p);
1517  if (a2p != a2buf)
1518  pfree(a2p);
1519 
1520  return result;
1521  }
1522 #endif /* WIN32 */
1523 
1524  if (len1 >= TEXTBUFLEN)
1525  a1p = (char *) palloc(len1 + 1);
1526  else
1527  a1p = a1buf;
1528  if (len2 >= TEXTBUFLEN)
1529  a2p = (char *) palloc(len2 + 1);
1530  else
1531  a2p = a2buf;
1532 
1533  memcpy(a1p, arg1, len1);
1534  a1p[len1] = '\0';
1535  memcpy(a2p, arg2, len2);
1536  a2p[len2] = '\0';
1537 
1538  if (mylocale)
1539  {
1540  if (mylocale->provider == COLLPROVIDER_ICU)
1541  {
1542 #ifdef USE_ICU
1543 #ifdef HAVE_UCOL_STRCOLLUTF8
1544  if (GetDatabaseEncoding() == PG_UTF8)
1545  {
1546  UErrorCode status;
1547 
1548  status = U_ZERO_ERROR;
1549  result = ucol_strcollUTF8(mylocale->info.icu.ucol,
1550  arg1, len1,
1551  arg2, len2,
1552  &status);
1553  if (U_FAILURE(status))
1554  ereport(ERROR,
1555  (errmsg("collation failed: %s", u_errorName(status))));
1556  }
1557  else
1558 #endif
1559  {
1560  int32_t ulen1,
1561  ulen2;
1562  UChar *uchar1,
1563  *uchar2;
1564 
1565  ulen1 = icu_to_uchar(&uchar1, arg1, len1);
1566  ulen2 = icu_to_uchar(&uchar2, arg2, len2);
1567 
1568  result = ucol_strcoll(mylocale->info.icu.ucol,
1569  uchar1, ulen1,
1570  uchar2, ulen2);
1571  }
1572 #else /* not USE_ICU */
1573  /* shouldn't happen */
1574  elog(ERROR, "unsupported collprovider: %c", mylocale->provider);
1575 #endif /* not USE_ICU */
1576  }
1577  else
1578  {
1579 #ifdef HAVE_LOCALE_T
1580  result = strcoll_l(a1p, a2p, mylocale->info.lt);
1581 #else
1582  /* shouldn't happen */
1583  elog(ERROR, "unsupported collprovider: %c", mylocale->provider);
1584 #endif
1585  }
1586  }
1587  else
1588  result = strcoll(a1p, a2p);
1589 
1590  /*
1591  * In some locales strcoll() can claim that nonidentical strings are
1592  * equal. Believing that would be bad news for a number of reasons,
1593  * so we follow Perl's lead and sort "equal" strings according to
1594  * strcmp().
1595  */
1596  if (result == 0)
1597  result = strcmp(a1p, a2p);
1598 
1599  if (a1p != a1buf)
1600  pfree(a1p);
1601  if (a2p != a2buf)
1602  pfree(a2p);
1603  }
1604 
1605  return result;
1606 }
#define COLLPROVIDER_ICU
Definition: pg_collation.h:85
int errhint(const char *fmt,...)
Definition: elog.c:987
#define Min(x, y)
Definition: c.h:806
union pg_locale_struct::@120 info
int errcode(int sqlerrcode)
Definition: elog.c:575
return result
Definition: formatting.c:1632
#define OidIsValid(objectId)
Definition: c.h:538
#define wcscoll_l
Definition: win32.h:348
void pfree(void *pointer)
Definition: mcxt.c:950
#define ERROR
Definition: elog.h:43
bool lc_collate_is_c(Oid collation)
Definition: pg_locale.c:1128
#define DEFAULT_COLLATION_OID
Definition: pg_collation.h:75
#define ereport(elevel, rest)
Definition: elog.h:122
pg_locale_t pg_newlocale_from_collation(Oid collid)
Definition: pg_locale.c:1265
int GetDatabaseEncoding(void)
Definition: mbutils.c:1015
#define TEXTBUFLEN
Definition: varlena.c:83
void * palloc(Size size)
Definition: mcxt.c:849
int errmsg(const char *fmt,...)
Definition: elog.c:797
int int32_t
Definition: crypt.c:108
#define elog
Definition: elog.h:219
static void static void status(const char *fmt,...) pg_attribute_printf(1
Definition: pg_regress.c:224
#define strcoll_l
Definition: win32.h:346
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.

References ereport, errcode(), errmsg(), ERROR, i, MAX_LEVENSHTEIN_STRLEN, Min, NULL, palloc(), pg_mblen(), pg_mbstrlen_with_len(), rest_of_char_same(), START_COLUMN, and STOP_COLUMN.

Referenced by levenshtein(), and levenshtein_with_costs().

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 }
#define Min(x, y)
Definition: c.h:806
int errcode(int sqlerrcode)
Definition: elog.c:575
int pg_mbstrlen_with_len(const char *mbstr, int limit)
Definition: mbutils.c:805
#define STOP_COLUMN
#define ERROR
Definition: elog.h:43
#define START_COLUMN
#define ereport(elevel, rest)
Definition: elog.h:122
#define NULL
Definition: c.h:229
#define MAX_LEVENSHTEIN_STRLEN
Definition: levenshtein.c:26
static bool rest_of_char_same(const char *s1, const char *s2, int len)
Definition: varlena.c:5456
int pg_mblen(const char *mbstr)
Definition: mbutils.c:771
void * palloc(Size size)
Definition: mcxt.c:849
int errmsg(const char *fmt,...)
Definition: elog.c:797
int i
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 
)
void varstr_sortsupport ( SortSupport  ssup,
Oid  collid,
bool  bpchar 
)

Definition at line 1804 of file varlena.c.

References VarStringSortSupport::abbr_card, SortSupportData::abbrev_abort, SortSupportData::abbrev_converter, SortSupportData::abbrev_full_comparator, SortSupportData::abbreviate, VarStringSortSupport::bpchar, bpchar(), bpcharfastcmp_c(), VarStringSortSupport::buf1, VarStringSortSupport::buf2, VarStringSortSupport::buflen1, VarStringSortSupport::buflen2, VarStringSortSupport::cache_blob, VarStringSortSupport::collate_c, COLLPROVIDER_ICU, SortSupportData::comparator, DEFAULT_COLLATION_OID, ereport, errcode(), errhint(), errmsg(), ERROR, VarStringSortSupport::full_card, GetDatabaseEncoding(), initHyperLogLog(), VarStringSortSupport::last_len1, VarStringSortSupport::last_len2, VarStringSortSupport::last_returned, lc_collate_is_c(), VarStringSortSupport::locale, locale, OidIsValid, palloc(), pg_newlocale_from_collation(), PG_UTF8, VarStringSortSupport::prop_card, pg_locale_struct::provider, SortSupportData::ssup_extra, TEXTBUFLEN, varstr_abbrev_abort(), varstr_abbrev_convert(), varstrcmp_abbrev(), varstrfastcmp_c(), and varstrfastcmp_locale().

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

1805 {
1806  bool abbreviate = ssup->abbreviate;
1807  bool collate_c = false;
1808  VarStringSortSupport *sss;
1809  pg_locale_t locale = 0;
1810 
1811  /*
1812  * If possible, set ssup->comparator to a function which can be used to
1813  * directly compare two datums. If we can do this, we'll avoid the
1814  * overhead of a trip through the fmgr layer for every comparison, which
1815  * can be substantial.
1816  *
1817  * Most typically, we'll set the comparator to varstrfastcmp_locale, which
1818  * uses strcoll() to perform comparisons and knows about the special
1819  * requirements of BpChar callers. However, if LC_COLLATE = C, we can
1820  * make things quite a bit faster with varstrfastcmp_c or bpcharfastcmp_c,
1821  * both of which use memcmp() rather than strcoll().
1822  *
1823  * There is a further exception on Windows. When the database encoding is
1824  * UTF-8 and we are not using the C collation, complex hacks are required.
1825  * We don't currently have a comparator that handles that case, so we fall
1826  * back on the slow method of having the sort code invoke bttextcmp() (in
1827  * the case of text) via the fmgr trampoline.
1828  */
1829  if (lc_collate_is_c(collid))
1830  {
1831  if (!bpchar)
1832  ssup->comparator = varstrfastcmp_c;
1833  else
1834  ssup->comparator = bpcharfastcmp_c;
1835 
1836  collate_c = true;
1837  }
1838 #ifdef WIN32
1839  else if (GetDatabaseEncoding() == PG_UTF8)
1840  return;
1841 #endif
1842  else
1843  {
1845 
1846  /*
1847  * We need a collation-sensitive comparison. To make things faster,
1848  * we'll figure out the collation based on the locale id and cache the
1849  * result.
1850  */
1851  if (collid != DEFAULT_COLLATION_OID)
1852  {
1853  if (!OidIsValid(collid))
1854  {
1855  /*
1856  * This typically means that the parser could not resolve a
1857  * conflict of implicit collations, so report it that way.
1858  */
1859  ereport(ERROR,
1860  (errcode(ERRCODE_INDETERMINATE_COLLATION),
1861  errmsg("could not determine which collation to use for string comparison"),
1862  errhint("Use the COLLATE clause to set the collation explicitly.")));
1863  }
1864  locale = pg_newlocale_from_collation(collid);
1865  }
1866  }
1867 
1868  /*
1869  * Unfortunately, it seems that abbreviation for non-C collations is
1870  * broken on many common platforms; testing of multiple versions of glibc
1871  * reveals that, for many locales, strcoll() and strxfrm() do not return
1872  * consistent results, which is fatal to this optimization. While no
1873  * other libc other than Cygwin has so far been shown to have a problem,
1874  * we take the conservative course of action for right now and disable
1875  * this categorically. (Users who are certain this isn't a problem on
1876  * their system can define TRUST_STRXFRM.)
1877  *
1878  * Even apart from the risk of broken locales, it's possible that there
1879  * are platforms where the use of abbreviated keys should be disabled at
1880  * compile time. Having only 4 byte datums could make worst-case
1881  * performance drastically more likely, for example. Moreover, macOS's
1882  * strxfrm() implementation is known to not effectively concentrate a
1883  * significant amount of entropy from the original string in earlier
1884  * transformed blobs. It's possible that other supported platforms are
1885  * similarly encumbered. So, if we ever get past disabling this
1886  * categorically, we may still want or need to disable it for particular
1887  * platforms.
1888  */
1889 #ifndef TRUST_STRXFRM
1890  if (!collate_c && !(locale && locale->provider == COLLPROVIDER_ICU))
1891  abbreviate = false;
1892 #endif
1893 
1894  /*
1895  * If we're using abbreviated keys, or if we're using a locale-aware
1896  * comparison, we need to initialize a StringSortSupport object. Both
1897  * cases will make use of the temporary buffers we initialize here for
1898  * scratch space (and to detect requirement for BpChar semantics from
1899  * caller), and the abbreviation case requires additional state.
1900  */
1901  if (abbreviate || !collate_c)
1902  {
1903  sss = palloc(sizeof(VarStringSortSupport));
1904  sss->buf1 = palloc(TEXTBUFLEN);
1905  sss->buflen1 = TEXTBUFLEN;
1906  sss->buf2 = palloc(TEXTBUFLEN);
1907  sss->buflen2 = TEXTBUFLEN;
1908  /* Start with invalid values */
1909  sss->last_len1 = -1;
1910  sss->last_len2 = -1;
1911  /* Initialize */
1912  sss->last_returned = 0;
1913  sss->locale = locale;
1914 
1915  /*
1916  * To avoid somehow confusing a strxfrm() blob and an original string,
1917  * constantly keep track of the variety of data that buf1 and buf2
1918  * currently contain.
1919  *
1920  * Comparisons may be interleaved with conversion calls. Frequently,
1921  * conversions and comparisons are batched into two distinct phases,
1922  * but the correctness of caching cannot hinge upon this. For
1923  * comparison caching, buffer state is only trusted if cache_blob is
1924  * found set to false, whereas strxfrm() caching only trusts the state
1925  * when cache_blob is found set to true.
1926  *
1927  * Arbitrarily initialize cache_blob to true.
1928  */
1929  sss->cache_blob = true;
1930  sss->collate_c = collate_c;
1931  sss->bpchar = bpchar;
1932  ssup->ssup_extra = sss;
1933 
1934  /*
1935  * If possible, plan to use the abbreviated keys optimization. The
1936  * core code may switch back to authoritative comparator should
1937  * abbreviation be aborted.
1938  */
1939  if (abbreviate)
1940  {
1941  sss->prop_card = 0.20;
1942  initHyperLogLog(&sss->abbr_card, 10);
1943  initHyperLogLog(&sss->full_card, 10);
1944  ssup->abbrev_full_comparator = ssup->comparator;
1945  ssup->comparator = varstrcmp_abbrev;
1948  }
1949  }
1950 }
static int varstrfastcmp_locale(Datum x, Datum y, SortSupport ssup)
Definition: varlena.c:2026
#define COLLPROVIDER_ICU
Definition: pg_collation.h:85
int(* comparator)(Datum x, Datum y, SortSupport ssup)
Definition: sortsupport.h:107
static int bpcharfastcmp_c(Datum x, Datum y, SortSupport ssup)
Definition: varlena.c:1993
int errhint(const char *fmt,...)
Definition: elog.c:987
static Datum varstr_abbrev_convert(Datum original, SortSupport ssup)
Definition: varlena.c:2226
static int varstrcmp_abbrev(Datum x, Datum y, SortSupport ssup)
Definition: varlena.c:2201
int errcode(int sqlerrcode)
Definition: elog.c:575
#define OidIsValid(objectId)
Definition: c.h:538
static int varstrfastcmp_c(Datum x, Datum y, SortSupport ssup)
Definition: varlena.c:1956
void initHyperLogLog(hyperLogLogState *cState, uint8 bwidth)
Definition: hyperloglog.c:65
#define ERROR
Definition: elog.h:43
bool lc_collate_is_c(Oid collation)
Definition: pg_locale.c:1128
hyperLogLogState abbr_card
Definition: varlena.c:73
pg_locale_t locale
Definition: varlena.c:76
#define DEFAULT_COLLATION_OID
Definition: pg_collation.h:75
int(* abbrev_full_comparator)(Datum x, Datum y, SortSupport ssup)
Definition: sortsupport.h:192
void * ssup_extra
Definition: sortsupport.h:87
#define ereport(elevel, rest)
Definition: elog.h:122
pg_locale_t pg_newlocale_from_collation(Oid collid)
Definition: pg_locale.c:1265
int GetDatabaseEncoding(void)
Definition: mbutils.c:1015
hyperLogLogState full_card
Definition: varlena.c:74
static bool varstr_abbrev_abort(int memtupcount, SortSupport ssup)
Definition: varlena.c:2460
#define TEXTBUFLEN
Definition: varlena.c:83
void * palloc(Size size)
Definition: mcxt.c:849
int errmsg(const char *fmt,...)
Definition: elog.c:797
static char * locale
Definition: initdb.c:123
Datum bpchar(PG_FUNCTION_ARGS)
Definition: varchar.c:267
bool(* abbrev_abort)(int memtupcount, SortSupport ssup)
Definition: sortsupport.h:183
Datum(* abbrev_converter)(Datum original, SortSupport ssup)
Definition: sortsupport.h:173