/*
* We do handle case-insensitive matching for single-byte encodings using
* fold-on-the-fly processing, however.
*/
static char
SB_lower_char(unsigned char c, pg_locale_t locale, bool locale_is_c)
{
if (locale_is_c)
return pg_ascii_tolower(c);
#ifdef HAVE_LOCALE_T
else if (locale)
return tolower_l(c, locale);
#endif
else
return pg_tolower(c);
}
/*
* dir_strcmp: strcmp except any two DIR_SEP characters are considered equal,
* and we honor filesystem case insensitivity if known
*/
static int
dir_strcmp(const char *s1, const char *s2)
{
while (*s1 && *s2)
{
if (
#ifndef WIN32
*s1 != *s2
#else
/* On windows, paths are case-insensitive */
pg_tolower((unsigned char) *s1) != pg_tolower((unsigned char) *s2)
#endif
&& !(IS_DIR_SEP(*s1) && IS_DIR_SEP(*s2)))
return (int) *s1 - (int) *s2;
s1++, s2++;
}
if (*s1)
return 1; /* s1 longer */
if (*s2)
return -1; /* s2 longer */
return 0;
}
/* * pg_an_to_ln -- return the local name corresponding to an authentication * name * * XXX Assumes that the first aname component is the user name. This is NOT * necessarily so, since an aname can actually be something out of your * worst X.400 nightmare, like * ORGANIZATION=U. C. Berkeley/NAME=Paul M. [email protected] * Note that the MIT an_to_ln code does the same thing if you don't * provide an aname mapping database...it may be a better idea to use * krb5_an_to_ln, except that it punts if multiple components are found, * and we can't afford to punt. * * For WIN32, convert username to lowercase because the Win32 kerberos library * generates tickets with the username as the user entered it instead of as * it is entered in the directory. */ static char * pg_an_to_ln(char *aname) { char *p; if ((p = strchr(aname, '/')) || (p = strchr(aname, '@'))) *p = '\0'; #ifdef WIN32 for (p = aname; *p; p++) *p = pg_tolower((unsigned char) *p); #endif return aname; }
/*
* Apply additional validation checks to a tzEntry
*
* Returns TRUE if OK, else false
*/
static bool
validateTzEntry(tzEntry *tzentry)
{
unsigned char *p;
/*
* Check restrictions imposed by datetkntbl storage format (see
* datetime.c)
*/
if (strlen(tzentry->abbrev) > TOKMAXLEN)
{
ereport(tz_elevel,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("time zone abbreviation \"%s\" is too long (maximum %d characters) in time zone file \"%s\", line %d",
tzentry->abbrev, TOKMAXLEN,
tzentry->filename, tzentry->lineno)));
return false;
}
if (tzentry->offset % 900 != 0)
{
ereport(tz_elevel,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("time zone offset %d is not a multiple of 900 sec (15 min) in time zone file \"%s\", line %d",
tzentry->offset,
tzentry->filename, tzentry->lineno)));
return false;
}
/*
* Sanity-check the offset: shouldn't exceed 14 hours
*/
if (tzentry->offset > 14 * 60 * 60 ||
tzentry->offset < -14 * 60 * 60)
{
ereport(tz_elevel,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("time zone offset %d is out of range in time zone file \"%s\", line %d",
tzentry->offset,
tzentry->filename, tzentry->lineno)));
return false;
}
/*
* Convert abbrev to lowercase (must match datetime.c's conversion)
*/
for (p = (unsigned char *) tzentry->abbrev; *p; p++)
*p = pg_tolower(*p);
return true;
}
/*
* Apply additional validation checks to a tzEntry
*
* Returns TRUE if OK, else false
*/
static bool
validateTzEntry(tzEntry *tzentry)
{
unsigned char *p;
/*
* Check restrictions imposed by datetkntbl storage format (see
* datetime.c)
*/
if (strlen(tzentry->abbrev) > TOKMAXLEN)
{
GUC_check_errmsg("time zone abbreviation \"%s\" is too long (maximum %d characters) in time zone file \"%s\", line %d",
tzentry->abbrev, TOKMAXLEN,
tzentry->filename, tzentry->lineno);
return false;
}
/*
* Sanity-check the offset: shouldn't exceed 14 hours
*/
if (tzentry->offset > 14 * 60 * 60 ||
tzentry->offset < -14 * 60 * 60)
{
GUC_check_errmsg("time zone offset %d is out of range in time zone file \"%s\", line %d",
tzentry->offset,
tzentry->filename, tzentry->lineno);
return false;
}
/*
* Convert abbrev to lowercase (must match datetime.c's conversion)
*/
for (p = (unsigned char *) tzentry->abbrev; *p; p++)
*p = pg_tolower(*p);
return true;
}
/*
* processSQLNamePattern
*
* Scan a wildcard-pattern string and generate appropriate WHERE clauses
* to limit the set of objects returned. The WHERE clauses are appended
* to the already-partially-constructed query in buf. Returns whether
* any clause was added.
*
* conn: connection query will be sent to (consulted for escaping rules).
* buf: output parameter.
* pattern: user-specified pattern option, or NULL if none ("*" is implied).
* have_where: true if caller already emitted "WHERE" (clauses will be ANDed
* onto the existing WHERE clause).
* force_escape: always quote regexp special characters, even outside
* double quotes (else they are quoted only between double quotes).
* schemavar: name of query variable to match against a schema-name pattern.
* Can be NULL if no schema.
* namevar: name of query variable to match against an object-name pattern.
* altnamevar: NULL, or name of an alternative variable to match against name.
* visibilityrule: clause to use if we want to restrict to visible objects
* (for example, "pg_catalog.pg_table_is_visible(p.oid)"). Can be NULL.
*
* Formatting note: the text already present in buf should end with a newline.
* The appended text, if any, will end with one too.
*/
bool
processSQLNamePattern(PGconn *conn, PQExpBuffer buf, const char *pattern,
bool have_where, bool force_escape,
const char *schemavar, const char *namevar,
const char *altnamevar, const char *visibilityrule)
{
PQExpBufferData schemabuf;
PQExpBufferData namebuf;
int encoding = PQclientEncoding(conn);
bool inquotes;
const char *cp;
int i;
bool added_clause = false;
#define WHEREAND() \
(appendPQExpBufferStr(buf, have_where ? " AND " : "WHERE "), \
have_where = true, added_clause = true)
if (pattern == NULL)
{
/* Default: select all visible objects */
if (visibilityrule)
{
WHEREAND();
appendPQExpBuffer(buf, "%s\n", visibilityrule);
}
return added_clause;
}
initPQExpBuffer(&schemabuf);
initPQExpBuffer(&namebuf);
/*
* Parse the pattern, converting quotes and lower-casing unquoted letters.
* Also, adjust shell-style wildcard characters into regexp notation.
*
* We surround the pattern with "^(...)$" to force it to match the whole
* string, as per SQL practice. We have to have parens in case the string
* contains "|", else the "^" and "$" will be bound into the first and
* last alternatives which is not what we want.
*
* Note: the result of this pass is the actual regexp pattern(s) we want
* to execute. Quoting/escaping into SQL literal format will be done
* below using appendStringLiteralConn().
*/
appendPQExpBufferStr(&namebuf, "^(");
inquotes = false;
cp = pattern;
while (*cp)
{
char ch = *cp;
if (ch == '"')
{
if (inquotes && cp[1] == '"')
{
/* emit one quote, stay in inquotes mode */
appendPQExpBufferChar(&namebuf, '"');
cp++;
}
else
inquotes = !inquotes;
cp++;
}
else if (!inquotes && isupper((unsigned char) ch))
{
appendPQExpBufferChar(&namebuf,
pg_tolower((unsigned char) ch));
cp++;
}
else if (!inquotes && ch == '*')
{
appendPQExpBufferStr(&namebuf, ".*");
cp++;
}
else if (!inquotes && ch == '?')
{
appendPQExpBufferChar(&namebuf, '.');
cp++;
}
else if (!inquotes && ch == '.')
{
/* Found schema/name separator, move current pattern to schema */
resetPQExpBuffer(&schemabuf);
appendPQExpBufferStr(&schemabuf, namebuf.data);
resetPQExpBuffer(&namebuf);
appendPQExpBufferStr(&namebuf, "^(");
cp++;
}
else if (ch == '$')
{
/*
* Dollar is always quoted, whether inside quotes or not. The
* reason is that it's allowed in SQL identifiers, so there's a
* significant use-case for treating it literally, while because
* we anchor the pattern automatically there is no use-case for
* having it possess its regexp meaning.
*/
appendPQExpBufferStr(&namebuf, "\\$");
cp++;
}
else
{
/*
* Ordinary data character, transfer to pattern
*
* Inside double quotes, or at all times if force_escape is true,
* quote regexp special characters with a backslash to avoid
* regexp errors. Outside quotes, however, let them pass through
* as-is; this lets knowledgeable users build regexp expressions
* that are more powerful than shell-style patterns.
*/
if ((inquotes || force_escape) &&
strchr("|*+?()[]{}.^$\\", ch))
appendPQExpBufferChar(&namebuf, '\\');
i = PQmblen(cp, encoding);
while (i-- && *cp)
{
appendPQExpBufferChar(&namebuf, *cp);
cp++;
}
}
}
/*
* Now decide what we need to emit. Note there will be a leading "^(" in
* the patterns in any case.
*/
if (namebuf.len > 2)
{
/* We have a name pattern, so constrain the namevar(s) */
appendPQExpBufferStr(&namebuf, ")$");
/* Optimize away a "*" pattern */
if (strcmp(namebuf.data, "^(.*)$") != 0)
{
WHEREAND();
if (altnamevar)
{
appendPQExpBuffer(buf, "(%s ~ ", namevar);
appendStringLiteralConn(buf, namebuf.data, conn);
appendPQExpBuffer(buf, "\n OR %s ~ ", altnamevar);
appendStringLiteralConn(buf, namebuf.data, conn);
appendPQExpBufferStr(buf, ")\n");
}
else
{
appendPQExpBuffer(buf, "%s ~ ", namevar);
appendStringLiteralConn(buf, namebuf.data, conn);
appendPQExpBufferChar(buf, '\n');
}
}
}
if (schemabuf.len > 2)
{
/* We have a schema pattern, so constrain the schemavar */
appendPQExpBufferStr(&schemabuf, ")$");
/* Optimize away a "*" pattern */
if (strcmp(schemabuf.data, "^(.*)$") != 0 && schemavar)
{
WHEREAND();
appendPQExpBuffer(buf, "%s ~ ", schemavar);
appendStringLiteralConn(buf, schemabuf.data, conn);
appendPQExpBufferChar(buf, '\n');
}
}
else
{
/* No schema pattern given, so select only visible objects */
if (visibilityrule)
{
WHEREAND();
appendPQExpBuffer(buf, "%s\n", visibilityrule);
}
}
termPQExpBuffer(&schemabuf);
termPQExpBuffer(&namebuf);
return added_clause;
#undef WHEREAND
}
/*
* PGTYPESdate_defmt_asc
*
* function works as follows:
* - first we analyze the paramters
* - if this is a special case with no delimiters, add delimters
* - find the tokens. First we look for numerical values. If we have found
* less than 3 tokens, we check for the months' names and thereafter for
* the abbreviations of the months' names.
* - then we see which parameter should be the date, the month and the
* year and from these values we calculate the date
*/
#define PGTYPES_DATE_MONTH_MAXLENGTH 20 /* probably even less :-) */
int
PGTYPESdate_defmt_asc(date * d, char *fmt, char *str)
{
/*
* token[2] = { 4,6 } means that token 2 starts at position 4 and ends at
* (including) position 6
*/
int token[3][2];
int token_values[3] = {-1, -1, -1};
char *fmt_token_order;
char *fmt_ystart,
*fmt_mstart,
*fmt_dstart;
int i;
int reading_digit;
int token_count;
char *str_copy;
struct tm tm;
tm.tm_year = tm.tm_mon = tm.tm_mday = 0; /* keep compiler quiet */
if (!d || !str || !fmt)
{
errno = PGTYPES_DATE_ERR_EARGS;
return -1;
}
/* analyze the fmt string */
fmt_ystart = strstr(fmt, "yy");
fmt_mstart = strstr(fmt, "mm");
fmt_dstart = strstr(fmt, "dd");
if (!fmt_ystart || !fmt_mstart || !fmt_dstart)
{
errno = PGTYPES_DATE_ERR_EARGS;
return -1;
}
if (fmt_ystart < fmt_mstart)
{
/* y m */
if (fmt_dstart < fmt_ystart)
{
/* d y m */
fmt_token_order = "dym";
}
else if (fmt_dstart > fmt_mstart)
{
/* y m d */
fmt_token_order = "ymd";
}
else
{
/* y d m */
fmt_token_order = "ydm";
}
}
else
{
/* fmt_ystart > fmt_mstart */
/* m y */
if (fmt_dstart < fmt_mstart)
{
/* d m y */
fmt_token_order = "dmy";
}
else if (fmt_dstart > fmt_ystart)
{
/* m y d */
fmt_token_order = "myd";
}
else
{
/* m d y */
fmt_token_order = "mdy";
}
}
/*
* handle the special cases where there is no delimiter between the
* digits. If we see this:
*
* only digits, 6 or 8 bytes then it might be ddmmyy and ddmmyyyy (or
* similar)
*
* we reduce it to a string with delimiters and continue processing
*/
/* check if we have only digits */
reading_digit = 1;
for (i = 0; str[i]; i++)
{
if (!isdigit((unsigned char) str[i]))
{
reading_digit = 0;
break;
}
}
if (reading_digit)
{
int frag_length[3];
int target_pos;
i = strlen(str);
if (i != 8 && i != 6)
{
errno = PGTYPES_DATE_ERR_ENOSHORTDATE;
return -1;
}
/* okay, this really is the special case */
/*
* as long as the string, one additional byte for the terminator and 2
* for the delimiters between the 3 fiedls
*/
str_copy = pgtypes_alloc(strlen(str) + 1 + 2);
if (!str_copy)
return -1;
/* determine length of the fragments */
if (i == 6)
{
frag_length[0] = 2;
frag_length[1] = 2;
frag_length[2] = 2;
}
else
{
if (fmt_token_order[0] == 'y')
{
frag_length[0] = 4;
frag_length[1] = 2;
frag_length[2] = 2;
}
else if (fmt_token_order[1] == 'y')
{
frag_length[0] = 2;
frag_length[1] = 4;
frag_length[2] = 2;
}
else
{
frag_length[0] = 2;
frag_length[1] = 2;
frag_length[2] = 4;
}
}
target_pos = 0;
/*
* XXX: Here we could calculate the positions of the tokens and save
* the for loop down there where we again check with isdigit() for
* digits.
*/
for (i = 0; i < 3; i++)
{
int start_pos = 0;
if (i >= 1)
start_pos += frag_length[0];
if (i == 2)
start_pos += frag_length[1];
strncpy(str_copy + target_pos, str + start_pos,
frag_length[i]);
target_pos += frag_length[i];
if (i != 2)
{
str_copy[target_pos] = ' ';
target_pos++;
}
}
str_copy[target_pos] = '\0';
}
else
{
str_copy = pgtypes_strdup(str);
if (!str_copy)
return -1;
/* convert the whole string to lower case */
for (i = 0; str_copy[i]; i++)
str_copy[i] = (char) pg_tolower((unsigned char) str_copy[i]);
}
/* look for numerical tokens */
reading_digit = 0;
token_count = 0;
for (i = 0; i < strlen(str_copy); i++)
{
if (!isdigit((unsigned char) str_copy[i]) && reading_digit)
{
/* the token is finished */
token[token_count][1] = i - 1;
reading_digit = 0;
token_count++;
}
else if (isdigit((unsigned char) str_copy[i]) && !reading_digit)
{
/* we have found a token */
token[token_count][0] = i;
reading_digit = 1;
}
}
/*
* we're at the end of the input string, but maybe we are still reading a
* number...
*/
if (reading_digit)
{
token[token_count][1] = i - 1;
token_count++;
}
if (token_count < 2)
{
/*
* not all tokens found, no way to find 2 missing tokens with string
* matches
*/
free(str_copy);
errno = PGTYPES_DATE_ERR_ENOSHORTDATE;
return -1;
}
if (token_count != 3)
{
/*
* not all tokens found but we may find another one with string
* matches by testing for the months names and months abbreviations
*/
char *month_lower_tmp = pgtypes_alloc(PGTYPES_DATE_MONTH_MAXLENGTH);
char *start_pos;
int j;
int offset;
int found = 0;
char **list;
if (!month_lower_tmp)
{
/* free variables we alloc'ed before */
free(str_copy);
return -1;
}
list = pgtypes_date_months;
for (i = 0; list[i]; i++)
{
for (j = 0; j < PGTYPES_DATE_MONTH_MAXLENGTH; j++)
{
month_lower_tmp[j] = (char) pg_tolower((unsigned char) list[i][j]);
if (!month_lower_tmp[j])
{
/* properly terminated */
break;
}
}
if ((start_pos = strstr(str_copy, month_lower_tmp)))
{
offset = start_pos - str_copy;
/*
* sort the new token into the numeric tokens, shift them if
* necessary
*/
if (offset < token[0][0])
{
token[2][0] = token[1][0];
token[2][1] = token[1][1];
token[1][0] = token[0][0];
token[1][1] = token[0][1];
token_count = 0;
}
else if (offset < token[1][0])
{
token[2][0] = token[1][0];
token[2][1] = token[1][1];
token_count = 1;
}
else
token_count = 2;
token[token_count][0] = offset;
token[token_count][1] = offset + strlen(month_lower_tmp) - 1;
/*
* the value is the index of the month in the array of months
* + 1 (January is month 0)
*/
token_values[token_count] = i + 1;
found = 1;
break;
}
/*
* evil[tm] hack: if we read the pgtypes_date_months and haven't
* found a match, reset list to point to pgtypes_date_months_short
* and reset the counter variable i
*/
if (list == pgtypes_date_months)
{
if (list[i + 1] == NULL)
{
list = months;
i = -1;
}
}
}
if (!found)
{
free(month_lower_tmp);
free(str_copy);
errno = PGTYPES_DATE_ERR_ENOTDMY;
return -1;
}
/*
* here we found a month. token[token_count] and
* token_values[token_count] reflect the month's details.
*
* only the month can be specified with a literal. Here we can do a
* quick check if the month is at the right position according to the
* format string because we can check if the token that we expect to
* be the month is at the position of the only token that already has
* a value. If we wouldn't check here we could say "December 4 1990"
* with a fmt string of "dd mm yy" for 12 April 1990.
*/
if (fmt_token_order[token_count] != 'm')
{
/* deal with the error later on */
token_values[token_count] = -1;
}
free(month_lower_tmp);
}
/* terminate the tokens with ASCII-0 and get their values */
for (i = 0; i < 3; i++)
{
*(str_copy + token[i][1] + 1) = '\0';
/* A month already has a value set, check for token_value == -1 */
if (token_values[i] == -1)
{
errno = 0;
token_values[i] = strtol(str_copy + token[i][0], (char **) NULL, 10);
/* strtol sets errno in case of an error */
if (errno)
token_values[i] = -1;
}
if (fmt_token_order[i] == 'd')
tm.tm_mday = token_values[i];
else if (fmt_token_order[i] == 'm')
tm.tm_mon = token_values[i];
else if (fmt_token_order[i] == 'y')
tm.tm_year = token_values[i];
}
free(str_copy);
if (tm.tm_mday < 1 || tm.tm_mday > 31)
{
errno = PGTYPES_DATE_BAD_DAY;
return -1;
}
if (tm.tm_mon < 1 || tm.tm_mon > MONTHS_PER_YEAR)
{
errno = PGTYPES_DATE_BAD_MONTH;
return -1;
}
if (tm.tm_mday == 31 && (tm.tm_mon == 4 || tm.tm_mon == 6 || tm.tm_mon == 9 || tm.tm_mon == 11))
{
errno = PGTYPES_DATE_BAD_DAY;
return -1;
}
if (tm.tm_mon == 2 && tm.tm_mday > 29)
{
errno = PGTYPES_DATE_BAD_DAY;
return -1;
}
*d = date2j(tm.tm_year, tm.tm_mon, tm.tm_mday) - date2j(2000, 1, 1);
return 0;
}
void
validate_and_adjust_options(StdRdOptions *result,
relopt_value *options,
int num_options, relopt_kind kind, bool validate)
{
int i;
relopt_value *fillfactor_opt;
relopt_value *appendonly_opt;
relopt_value *blocksize_opt;
relopt_value *comptype_opt;
relopt_value *complevel_opt;
relopt_value *checksum_opt;
relopt_value *orientation_opt;
/* fillfactor */
fillfactor_opt = get_option_set(options, num_options, SOPT_FILLFACTOR);
if (fillfactor_opt != NULL)
{
result->fillfactor = fillfactor_opt->values.int_val;
}
/* appendonly */
appendonly_opt = get_option_set(options, num_options, SOPT_APPENDONLY);
if (appendonly_opt != NULL)
{
if (!KIND_IS_RELATION(kind))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("usage of parameter \"appendonly\" in a non relation object is not supported")));
result->appendonly = appendonly_opt->values.bool_val;
}
/* blocksize */
blocksize_opt = get_option_set(options, num_options, SOPT_BLOCKSIZE);
if (blocksize_opt != NULL)
{
if (!KIND_IS_RELATION(kind) && validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("usage of parameter \"blocksize\" in a non relation object is not supported")));
if (!result->appendonly && validate)
ereport(ERROR,
(errcode(ERRCODE_GP_FEATURE_NOT_SUPPORTED),
errmsg("invalid option 'blocksize' for base relation. "
"Only valid for Append Only relations")));
result->blocksize = blocksize_opt->values.int_val;
if (result->blocksize < MIN_APPENDONLY_BLOCK_SIZE ||
result->blocksize > MAX_APPENDONLY_BLOCK_SIZE ||
result->blocksize % MIN_APPENDONLY_BLOCK_SIZE != 0)
{
if (validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("block size must be between 8KB and 2MB and be"
" an 8KB multiple. Got %d", result->blocksize)));
result->blocksize = DEFAULT_APPENDONLY_BLOCK_SIZE;
}
}
/* compression type */
comptype_opt = get_option_set(options, num_options, SOPT_COMPTYPE);
if (comptype_opt != NULL)
{
if (!KIND_IS_RELATION(kind) && validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("usage of parameter \"compresstype\" in a non relation object is not supported")));
if (!result->appendonly && validate)
ereport(ERROR,
(errcode(ERRCODE_GP_FEATURE_NOT_SUPPORTED),
errmsg("invalid option \"compresstype\" for base relation."
" Only valid for Append Only relations")));
result->compresstype = pstrdup(comptype_opt->values.string_val);
if (!compresstype_is_valid(result->compresstype))
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("unknown compresstype \"%s\"",
result->compresstype)));
for (i = 0; i < strlen(result->compresstype); i++)
result->compresstype[i] = pg_tolower(result->compresstype[i]);
}
/* compression level */
complevel_opt = get_option_set(options, num_options, SOPT_COMPLEVEL);
if (complevel_opt != NULL)
{
if (!KIND_IS_RELATION(kind) && validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("usage of parameter \"compresslevel\" in a non relation object is not supported")));
if (!result->appendonly && validate)
ereport(ERROR,
(errcode(ERRCODE_GP_FEATURE_NOT_SUPPORTED),
errmsg("invalid option 'compresslevel' for base "
"relation. Only valid for Append Only relations")));
result->compresslevel = complevel_opt->values.int_val;
if (result->compresstype &&
pg_strcasecmp(result->compresstype, "none") != 0 &&
result->compresslevel == 0 && validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("compresstype can\'t be used with compresslevel 0")));
if (result->compresslevel < 0)
{
if (validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("compresslevel=%d is out of range (should be positive)",
result->compresslevel)));
result->compresslevel = setDefaultCompressionLevel(result->compresstype);
}
/*
* use the default compressor if compresslevel was indicated but not
* compresstype. must make a copy otherwise str_tolower below will
* crash.
*/
if (result->compresslevel > 0 && !result->compresstype)
result->compresstype = pstrdup(AO_DEFAULT_COMPRESSTYPE);
/* Check upper bound of compresslevel for each compression type */
if (result->compresstype &&
(pg_strcasecmp(result->compresstype, "zlib") == 0) &&
(result->compresslevel > 9))
{
if (validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("compresslevel=%d is out of range for zlib "
"(should be in the range 1 to 9)",
result->compresslevel)));
result->compresslevel = setDefaultCompressionLevel(result->compresstype);
}
if (result->compresstype &&
(pg_strcasecmp(result->compresstype, "zstd") == 0) &&
(result->compresslevel > 19))
{
if (validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("compresslevel=%d is out of range for zstd "
"(should be in the range 1 to 19)",
result->compresslevel)));
result->compresslevel = setDefaultCompressionLevel(result->compresstype);
}
if (result->compresstype &&
(pg_strcasecmp(result->compresstype, "quicklz") == 0) &&
(result->compresslevel != 1))
{
if (validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("compresslevel=%d is out of range for quicklz "
"(should be 1)",
result->compresslevel)));
result->compresslevel = setDefaultCompressionLevel(result->compresstype);
}
if (result->compresstype &&
(pg_strcasecmp(result->compresstype, "rle_type") == 0) &&
(result->compresslevel > 4))
{
if (validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("compresslevel=%d is out of range for rle_type "
"(should be in the range 1 to 4)",
result->compresslevel)));
result->compresslevel = setDefaultCompressionLevel(result->compresstype);
}
}
/* checksum */
checksum_opt = get_option_set(options, num_options, SOPT_CHECKSUM);
if (checksum_opt != NULL)
{
if (!KIND_IS_RELATION(kind) && validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("usage of parameter \"checksum\" in a non relation "
"object is not supported")));
if (!result->appendonly && validate)
ereport(ERROR,
(errcode(ERRCODE_GP_FEATURE_NOT_SUPPORTED),
errmsg("invalid option \"checksum\" for base relation. "
"Only valid for Append Only relations")));
result->checksum = checksum_opt->values.bool_val;
}
/* Disable checksum for heap relations. */
else if (result->appendonly == false)
result->checksum = false;
/* columnstore */
orientation_opt = get_option_set(options, num_options, SOPT_ORIENTATION);
if (orientation_opt != NULL)
{
if (!KIND_IS_RELATION(kind) && validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("usage of parameter \"orientation\" in a non "
"relation object is not supported")));
if (!result->appendonly && validate)
ereport(ERROR,
(errcode(ERRCODE_GP_FEATURE_NOT_SUPPORTED),
errmsg("invalid option \"orientation\" for base relation. "
"Only valid for Append Only relations")));
if (!(pg_strcasecmp(orientation_opt->values.string_val, "column") == 0 ||
pg_strcasecmp(orientation_opt->values.string_val, "row") == 0) &&
validate)
{
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("invalid parameter value for \"orientation\": "
"\"%s\"", orientation_opt->values.string_val)));
}
result->columnstore = (pg_strcasecmp(orientation_opt->values.string_val, "column") == 0 ?
true : false);
if (result->compresstype &&
(pg_strcasecmp(result->compresstype, "rle_type") == 0) &&
!result->columnstore)
{
if (validate)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("%s cannot be used with Append Only relations "
"row orientation", result->compresstype)));
}
}
if (result->appendonly && result->compresstype != NULL)
if (result->compresslevel == AO_DEFAULT_COMPRESSLEVEL)
result->compresslevel = setDefaultCompressionLevel(result->compresstype);
}
/*
* Accept a string of the form "name=value,name=value,...". Space
* around ',' and '=' is allowed. Parsed values are stored in
* corresponding fields of StdRdOptions object. The parser is a
* finite state machine that changes states for each input character
* scanned.
*/
Datum
parseAOStorageOpts(const char *opts_str, bool *aovalue)
{
int dims[1];
int lbs[1];
Datum result;
ArrayBuildState *astate;
bool foundAO = false;
const char *cp;
const char *name_st = NULL;
const char *value_st = NULL;
char *name = NULL,
*value = NULL;
enum state
{
/*
* Consume whitespace at the beginning of a name token.
*/
LEADING_NAME,
/*
* Name token is being scanned. Allowed characters are alphabets,
* whitespace and '='.
*/
NAME_TOKEN,
/*
* Name token was terminated by whitespace. This state scans the
* trailing whitespace after name token.
*/
TRAILING_NAME,
/*
* Whitespace after '=' and before value token.
*/
LEADING_VALUE,
/*
* Value token is being scanned. Allowed characters are alphabets,
* digits, '_'. Value should be delimited by a ',', whitespace or end
* of string '\0'.
*/
VALUE_TOKEN,
/*
* Whitespace after value token.
*/
TRAILING_VALUE,
/*
* End of string. This state can only be entered from VALUE_TOKEN or
* TRAILING_VALUE.
*/
EOS
};
enum state st = LEADING_NAME;
/*
* Initialize ArrayBuildState ourselves rather than leaving it to
* accumArrayResult(). This aviods the catalog lookup (pg_type) performed
* by accumArrayResult().
*/
astate = (ArrayBuildState *) palloc(sizeof(ArrayBuildState));
astate->mcontext = CurrentMemoryContext;
astate->alen = 10; /* Initial number of name=value pairs. */
astate->dvalues = (Datum *) palloc(astate->alen * sizeof(Datum));
astate->dnulls = (bool *) palloc(astate->alen * sizeof(bool));
astate->nelems = 0;
astate->element_type = TEXTOID;
astate->typlen = -1;
astate->typbyval = false;
astate->typalign = 'i';
cp = opts_str - 1;
do
{
++cp;
switch (st)
{
case LEADING_NAME:
if (isalpha(*cp))
{
st = NAME_TOKEN;
name_st = cp;
}
else if (!isspace(*cp))
{
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("invalid storage option name in \"%s\"",
opts_str)));
}
break;
case NAME_TOKEN:
if (isspace(*cp))
st = TRAILING_NAME;
else if (*cp == '=')
st = LEADING_VALUE;
else if (!isalpha(*cp))
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("invalid storage option name in \"%s\"",
opts_str)));
if (st != NAME_TOKEN)
{
name = palloc(cp - name_st + 1);
strncpy(name, name_st, cp - name_st);
name[cp - name_st] = '\0';
for (name_st = name; *name_st != '\0'; ++name_st)
*(char *) name_st = pg_tolower(*name_st);
}
break;
case TRAILING_NAME:
if (*cp == '=')
st = LEADING_VALUE;
else if (!isspace(*cp))
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("invalid value for option \"%s\", expected \"=\"", name)));
break;
case LEADING_VALUE:
if (isalnum(*cp))
{
st = VALUE_TOKEN;
value_st = cp;
}
else if (!isspace(*cp))
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("invalid value for option \"%s\"", name)));
break;
case VALUE_TOKEN:
if (isspace(*cp))
st = TRAILING_VALUE;
else if (*cp == '\0')
st = EOS;
else if (*cp == ',')
st = LEADING_NAME;
/* Need to check '_' for rle_type */
else if (!(isalnum(*cp) || *cp == '_'))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("invalid value for option \"%s\"", name)));
if (st != VALUE_TOKEN)
{
value = palloc(cp - value_st + 1);
strncpy(value, value_st, cp - value_st);
value[cp - value_st] = '\0';
for (value_st = value; *value_st != '\0'; ++value_st)
*(char *) value_st = pg_tolower(*value_st);
Assert(name);
accumAOStorageOpt(name, value, astate,
&foundAO, aovalue);
pfree(name);
name = NULL;
pfree(value);
value = NULL;
}
break;
case TRAILING_VALUE:
if (*cp == ',')
st = LEADING_NAME;
else if (*cp == '\0')
st = EOS;
else if (!isspace(*cp))
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("syntax error after \"%s\"", value)));
break;
case EOS:
/*
* We better get out of the loop right after entering this
* state. Therefore, we should never get here.
*/
elog(ERROR, "invalid value \"%s\" for GUC", opts_str);
break;
};
} while (*cp != '\0');
if (st != EOS)
elog(ERROR, "invalid value \"%s\" for GUC", opts_str);
if (!foundAO)
{
/*
* Add "appendonly=true" datum if it was not explicitly specified by
* user. This is needed to validate the array of datums constructed
* from user specified options.
*/
accumAOStorageOpt(SOPT_APPENDONLY, "true", astate, NULL, NULL);
}
lbs[0] = 1;
dims[0] = astate->nelems;
result = makeMdArrayResult(astate, 1, dims, lbs, CurrentMemoryContext, false);
pfree(astate->dvalues);
pfree(astate->dnulls);
pfree(astate);
return result;
}
