invert = 1;
/* FALLTHROUGH */
case ASCII:
- my_invlist = invlist_clone(PL_XPosix_ptrs[_CC_ASCII]);
+ my_invlist = invlist_clone(PL_Posix_ptrs[_CC_ASCII]);
/* This can be handled as a Posix class */
goto join_posix_and_ascii;
/* FALLTHROUGH */
case POSIXA:
assert(FLAGS(scan) != _CC_ASCII);
- _invlist_intersection(PL_XPosix_ptrs[FLAGS(scan)],
- PL_XPosix_ptrs[_CC_ASCII],
- &my_invlist);
+ my_invlist = invlist_clone(PL_Posix_ptrs[FLAGS(scan)]);
goto join_posix_and_ascii;
case NPOSIXD:
RExC_parse++;
is_neg = TRUE;
}
+ endptr = RExC_end;
if (grok_atoUV(RExC_parse, &unum, &endptr)
&& unum <= I32_MAX
) {
}
else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
UV uv;
+ endptr = RExC_end;
if (grok_atoUV(RExC_parse, &uv, &endptr)
&& uv <= I32_MAX
) {
/* (?(1)...) */
char c;
UV uv;
+ endptr = RExC_end;
if (grok_atoUV(RExC_parse, &uv, &endptr)
&& uv <= I32_MAX
) {
maxpos = next;
RExC_parse++;
if (isDIGIT(*RExC_parse)) {
+ endptr = RExC_end;
if (!grok_atoUV(RExC_parse, &uv, &endptr))
vFAIL("Invalid quantifier in {,}");
if (uv >= REG_INFTY)
else
maxpos = RExC_parse;
if (isDIGIT(*maxpos)) {
+ endptr = RExC_end;
if (!grok_atoUV(maxpos, &uv, &endptr))
vFAIL("Invalid quantifier in {,}");
if (uv >= REG_INFTY)
* *node_p, nor *code_point_p, nor *flagp.
*
* If <cp_count> is not NULL, the caller wants to know the length (in code
- * points) that this \N sequence matches. This is set even if the function
- * returns FALSE, as detailed below.
+ * points) that this \N sequence matches. This is set, and the input is
+ * parsed for errors, even if the function returns FALSE, as detailed below.
*
* There are 5 possibilities here, as detailed in the next 5 paragraphs.
*
}
/* Here, looks like its really a multiple character sequence. Fail
- * if that's not what the caller wants. */
- if (! node_p) {
-
- /* But even if failing, we count the code points if requested, and
- * don't back up up the pointer as the caller is expected to
- * handle this situation */
- if (cp_count) {
- char * dot = RExC_parse + 1;
- do {
- dot = (char *) memchr(dot, '.', endbrace - dot);
- if (! dot) {
- break;
- }
- count++;
- dot++;
- } while (dot < endbrace);
- count++;
-
- *cp_count = count;
- RExC_parse = endbrace;
- nextchar(pRExC_state);
- }
- else { /* Back up the pointer. */
- RExC_parse = p;
- }
+ * if that's not what the caller wants. But continue with counting
+ * and error checking if they still want a count */
+ if (! node_p && ! cp_count) {
return FALSE;
}
* form \x{char1}\x{char2}... and then call reg recursively to
* parse it (enclosing in "(?: ... )" ). That way, it retains its
* atomicness, while not having to worry about special handling
- * that some code points may have. */
+ * that some code points may have. We don't create a subpattern,
+ * but go through the motions of code point counting and error
+ * checking, if the caller doesn't want a node returned. */
- if (count == 1) {
+ if (node_p && count == 1) {
substitute_parse = newSVpvs("?:");
}
do_concat:
- /* Convert to notation the rest of the code understands */
- sv_catpv(substitute_parse, "\\x{");
- sv_catpvn(substitute_parse, start_digit, RExC_parse - start_digit);
- sv_catpv(substitute_parse, "}");
+ if (node_p) {
+ /* Convert to notation the rest of the code understands */
+ sv_catpv(substitute_parse, "\\x{");
+ sv_catpvn(substitute_parse, start_digit,
+ RExC_parse - start_digit);
+ sv_catpv(substitute_parse, "}");
+ }
/* Move to after the dot (or ending brace the final time through.)
* */
RExC_parse++;
+ count++;
} while (RExC_parse < endbrace);
+ if (! node_p) { /* Doesn't want the node */
+ assert (cp_count);
+
+ *cp_count = count;
+ return FALSE;
+ }
+
sv_catpv(substitute_parse, ")");
#ifdef EBCDIC
* in which case return I32_MAX (rather than possibly 32-bit wrapping) */
static I32
-S_backref_value(char *p)
+S_backref_value(char *p, char *e)
{
- const char* endptr;
+ const char* endptr = e;
UV val;
if (grok_atoUV(p, &val, &endptr) && val <= I32_MAX)
return (I32)val;
if (RExC_parse >= RExC_end) {
goto unterminated_g;
}
- num = S_backref_value(RExC_parse);
+ num = S_backref_value(RExC_parse, RExC_end);
if (num == 0)
vFAIL("Reference to invalid group 0");
else if (num == I32_MAX) {
}
}
else {
- num = S_backref_value(RExC_parse);
+ num = S_backref_value(RExC_parse, RExC_end);
/* bare \NNN might be backref or octal - if it is larger
* than or equal RExC_npar then it is assumed to be an
* octal escape. Note RExC_npar is +1 from the actual
/* NOTE, RExC_npar is 1 more than the actual number of
* parens we have seen so far, hence the < RExC_npar below. */
- if ( !isDIGIT(p[1]) || S_backref_value(p) < RExC_npar)
+ if ( !isDIGIT(p[1]) || S_backref_value(p, RExC_end) < RExC_npar)
{ /* Not to be treated as an octal constant, go
find backref */
--p;
} /* End of switch on '\' */
break;
case '{':
- /* Currently we allow an lbrace at the start of a construct
- * without raising a warning. This is because we think we
- * will never want such a brace to be meant to be other
- * than taken literally. */
+ /* Trying to gain new uses for '{' without breaking too
+ * much existing code is hard. The solution currently
+ * adopted is:
+ * 1) If there is no ambiguity that a '{' should always
+ * be taken literally, at the start of a construct, we
+ * just do so.
+ * 2) If the literal '{' conflicts with our desired use
+ * of it as a metacharacter, we die. The deprecation
+ * cycles for this have come and gone.
+ * 3) If there is ambiguity, we raise a simple warning.
+ * This could happen, for example, if the user
+ * intended it to introduce a quantifier, but slightly
+ * misspelled the quantifier. Without this warning,
+ * the quantifier would silently be taken as a literal
+ * string of characters instead of a meta construct */
if (len || (p > RExC_start && isALPHA_A(*(p - 1)))) {
-
- /* But, we raise a fatal warning otherwise, as the
- * deprecation cycle has come and gone. Except that it
- * turns out that some heavily-relied on upstream
- * software, notably GNU Autoconf, have failed to fix
- * their uses. For these, don't make it fatal unless
- * we anticipate using the '{' for something else.
- * This happens after any alpha, and for a looser {m,n}
- * quantifier specification */
if ( RExC_strict
|| ( p > parse_start + 1
&& isALPHA_A(*(p - 1))
"illegal here");
}
if (PASS2) {
- ckWARNregdep(p + 1,
- "Unescaped left brace in regex is "
- "deprecated here (and will be fatal "
- "in Perl 5.30), passed through");
+ ckWARNreg(p + 1, "Unescaped left brace in regex is"
+ " passed through");
}
}
goto normal_default;
/* Position parse to next real character */
skip_to_be_ignored_text(pRExC_state, &RExC_parse,
FALSE /* Don't force to /x */ );
- if (PASS2 && *RExC_parse == '{' && OP(ret) != SBOL && ! regcurly(RExC_parse)) {
- ckWARNregdep(RExC_parse + 1, "Unescaped left brace in regex is deprecated here (and will be fatal in Perl 5.30), passed through");
+ if ( PASS2 && *RExC_parse == '{'
+ && OP(ret) != SBOL && ! regcurly(RExC_parse))
+ {
+ if (RExC_strict || new_regcurly(RExC_parse, RExC_end)) {
+ RExC_parse++;
+ vFAIL("Unescaped left brace in regex is illegal here");
+ }
+ ckWARNreg(RExC_parse + 1, "Unescaped left brace in regex is"
+ " passed through");
}
return(ret);
}
}
else if ( UNI_SEMANTICS
+ || AT_LEAST_ASCII_RESTRICTED
|| classnum == _CC_ASCII
|| (DEPENDS_SEMANTICS && ( classnum == _CC_DIGIT
|| classnum == _CC_XDIGIT)))
{
- /* We usually have to worry about /d and /a affecting what
- * POSIX classes match, with special code needed for /d
- * because we won't know until runtime what all matches.
- * But there is no extra work needed under /u, and
- * [:ascii:] is unaffected by /a and /d; and :digit: and
- * :xdigit: don't have runtime differences under /d. So we
- * can special case these, and avoid some extra work below,
- * and at runtime. */
+ /* We usually have to worry about /d a affecting what POSIX
+ * classes match, with special code needed because we won't
+ * know until runtime what all matches. But there is no
+ * extra work needed under /u and /a; and [:ascii:] is
+ * unaffected by /d; and :digit: and :xdigit: don't have
+ * runtime differences under /d. So we can special case
+ * these, and avoid some extra work below, and at runtime.
+ * */
_invlist_union_maybe_complement_2nd(
simple_posixes,
- PL_XPosix_ptrs[classnum],
+ ((AT_LEAST_ASCII_RESTRICTED)
+ ? PL_Posix_ptrs[classnum]
+ : PL_XPosix_ptrs[classnum]),
namedclass % 2 != 0,
&simple_posixes);
}
}
}
if (posixes || nposixes) {
-
- /* We have to adjust /a and /aa */
- if (AT_LEAST_ASCII_RESTRICTED) {
-
- /* Under /a and /aa, nothing above ASCII matches these */
- if (posixes) {
- _invlist_intersection(posixes,
- PL_XPosix_ptrs[_CC_ASCII],
- &posixes);
- }
-
- /* Under /a and /aa, everything above ASCII matches these
- * complements */
- if (nposixes) {
- _invlist_union_complement_2nd(nposixes,
- PL_XPosix_ptrs[_CC_ASCII],
- &nposixes);
- }
- }
-
if (! DEPENDS_SEMANTICS) {
/* For everything but /d, we can just add the current 'posixes' and
#endif /* DEBUGGING */
+#ifndef PERL_IN_XSUB_RE
+
+#include "uni_keywords.h"
+
+void
+Perl_init_uniprops(pTHX)
+{
+ /* Set up the inversion list global variables */
+
+ PL_XPosix_ptrs[_CC_ASCII] = _new_invlist_C_array(uni_prop_ptrs[UNI_ASCII]);
+ PL_XPosix_ptrs[_CC_ALPHANUMERIC] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXALNUM]);
+ PL_XPosix_ptrs[_CC_ALPHA] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXALPHA]);
+ PL_XPosix_ptrs[_CC_BLANK] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXBLANK]);
+ PL_XPosix_ptrs[_CC_CASED] = _new_invlist_C_array(uni_prop_ptrs[UNI_CASED]);
+ PL_XPosix_ptrs[_CC_CNTRL] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXCNTRL]);
+ PL_XPosix_ptrs[_CC_DIGIT] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXDIGIT]);
+ PL_XPosix_ptrs[_CC_GRAPH] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXGRAPH]);
+ PL_XPosix_ptrs[_CC_LOWER] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXLOWER]);
+ PL_XPosix_ptrs[_CC_PRINT] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXPRINT]);
+ PL_XPosix_ptrs[_CC_PUNCT] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXPUNCT]);
+ PL_XPosix_ptrs[_CC_SPACE] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXSPACE]);
+ PL_XPosix_ptrs[_CC_UPPER] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXUPPER]);
+ PL_XPosix_ptrs[_CC_VERTSPACE] = _new_invlist_C_array(uni_prop_ptrs[UNI_VERTSPACE]);
+ PL_XPosix_ptrs[_CC_WORDCHAR] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXWORD]);
+ PL_XPosix_ptrs[_CC_XDIGIT] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXXDIGIT]);
+
+ PL_Posix_ptrs[_CC_ASCII] = _new_invlist_C_array(uni_prop_ptrs[UNI_ASCII]);
+ PL_Posix_ptrs[_CC_ALPHANUMERIC] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXALNUM]);
+ PL_Posix_ptrs[_CC_ALPHA] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXALPHA]);
+ PL_Posix_ptrs[_CC_BLANK] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXBLANK]);
+ PL_Posix_ptrs[_CC_CASED] = _new_invlist_C_array(uni_prop_ptrs[UNI_CASED]);
+ PL_Posix_ptrs[_CC_CNTRL] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXCNTRL]);
+ PL_Posix_ptrs[_CC_DIGIT] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXDIGIT]);
+ PL_Posix_ptrs[_CC_GRAPH] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXGRAPH]);
+ PL_Posix_ptrs[_CC_LOWER] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXLOWER]);
+ PL_Posix_ptrs[_CC_PRINT] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXPRINT]);
+ PL_Posix_ptrs[_CC_PUNCT] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXPUNCT]);
+ PL_Posix_ptrs[_CC_SPACE] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXSPACE]);
+ PL_Posix_ptrs[_CC_UPPER] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXUPPER]);
+ PL_Posix_ptrs[_CC_VERTSPACE] = _new_invlist_C_array(uni_prop_ptrs[UNI_VERTSPACE]);
+ PL_Posix_ptrs[_CC_WORDCHAR] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXWORD]);
+ PL_Posix_ptrs[_CC_XDIGIT] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXXDIGIT]);
+
+ PL_GCB_invlist = _new_invlist_C_array(_Perl_GCB_invlist);
+ PL_SB_invlist = _new_invlist_C_array(_Perl_SB_invlist);
+ PL_WB_invlist = _new_invlist_C_array(_Perl_WB_invlist);
+ PL_LB_invlist = _new_invlist_C_array(_Perl_LB_invlist);
+ PL_SCX_invlist = _new_invlist_C_array(_Perl_SCX_invlist);
+
+ PL_AboveLatin1 = _new_invlist_C_array(AboveLatin1_invlist);
+ PL_Latin1 = _new_invlist_C_array(Latin1_invlist);
+ PL_UpperLatin1 = _new_invlist_C_array(UpperLatin1_invlist);
+
+ PL_Assigned_invlist = _new_invlist_C_array(uni_prop_ptrs[UNI_ASSIGNED]);
+
+ PL_utf8_perl_idstart = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_IDSTART]);
+ PL_utf8_perl_idcont = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_IDCONT]);
+
+ PL_utf8_charname_begin = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_CHARNAME_BEGIN]);
+ PL_utf8_charname_continue = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_CHARNAME_CONTINUE]);
+
+ PL_utf8_foldable = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_ANY_FOLDS]);
+ PL_HasMultiCharFold = _new_invlist_C_array(uni_prop_ptrs[
+ UNI__PERL_FOLDS_TO_MULTI_CHAR]);
+ PL_NonL1NonFinalFold = _new_invlist_C_array(
+ NonL1_Perl_Non_Final_Folds_invlist);
+
+ PL_utf8_toupper = _new_invlist_C_array(Uppercase_Mapping_invlist);
+ PL_utf8_tolower = _new_invlist_C_array(Lowercase_Mapping_invlist);
+ PL_utf8_totitle = _new_invlist_C_array(Titlecase_Mapping_invlist);
+ PL_utf8_tofold = _new_invlist_C_array(Case_Folding_invlist);
+ PL_utf8_tosimplefold = _new_invlist_C_array(Simple_Case_Folding_invlist);
+ PL_utf8_foldclosures = _new_invlist_C_array(_Perl_IVCF_invlist);
+ PL_utf8_mark = _new_invlist_C_array(uni_prop_ptrs[UNI_M]);
+
+ /* The below are used only by deprecated functions. They could be removed */
+ PL_utf8_xidcont = _new_invlist_C_array(uni_prop_ptrs[UNI_XIDC]);
+ PL_utf8_idcont = _new_invlist_C_array(uni_prop_ptrs[UNI_IDC]);
+ PL_utf8_xidstart = _new_invlist_C_array(uni_prop_ptrs[UNI_XIDS]);
+}
+
+SV *
+Perl_parse_uniprop_string(pTHX_ const char * const name, const Size_t len, const bool to_fold, bool * invert)
+{
+ /* Parse the interior meat of \p{} passed to this in 'name' with length 'len',
+ * and return an inversion list if a property with 'name' is found, or NULL
+ * if not. 'name' point to the input with leading and trailing space trimmed.
+ * 'to_fold' indicates if /i is in effect.
+ *
+ * When the return is an inversion list, '*invert' will be set to a boolean
+ * indicating if it should be inverted or not
+ *
+ * This currently doesn't handle all cases. A NULL return indicates the
+ * caller should try a different approach
+ */
+
+ char* lookup_name;
+ bool stricter = FALSE;
+ bool is_nv_type = FALSE; /* nv= or numeric_value=, or possibly one
+ of the cjk numeric properties (though
+ it requires extra effort to compile
+ them) */
+ unsigned int i;
+ unsigned int j = 0, lookup_len;
+ int equals_pos = -1; /* Where the '=' is found, or negative if none */
+ int slash_pos = -1; /* Where the '/' is found, or negative if none */
+ int table_index = 0;
+ bool starts_with_In_or_Is = FALSE;
+ Size_t lookup_offset = 0;
+
+ PERL_ARGS_ASSERT_PARSE_UNIPROP_STRING;
+
+ /* The input will be modified into 'lookup_name' */
+ Newx(lookup_name, len, char);
+ SAVEFREEPV(lookup_name);
+
+ /* Parse the input. */
+ for (i = 0; i < len; i++) {
+ char cur = name[i];
+
+ /* These characters can be freely ignored in most situations. Later it
+ * may turn out we shouldn't have ignored them, and we have to reparse,
+ * but we don't have enough information yet to make that decision */
+ if (cur == '-' || cur == '_' || isSPACE(cur)) {
+ continue;
+ }
+
+ /* Case differences are also ignored. Our lookup routine assumes
+ * everything is lowercase */
+ if (isUPPER(cur)) {
+ lookup_name[j++] = toLOWER(cur);
+ continue;
+ }
+
+ /* A double colon is either an error, or a package qualifier to a
+ * subroutine user-defined property; neither of which do we currently
+ * handle
+ *
+ * But a single colon is a synonym for '=' */
+ if (cur == ':') {
+ if (i < len - 1 && name[i+1] == ':') {
+ return NULL;
+ }
+ cur = '=';
+ }
+
+ /* Otherwise, this character is part of the name. */
+ lookup_name[j++] = cur;
+
+ /* Only the equals sign needs further processing */
+ if (cur == '=') {
+ equals_pos = j; /* Note where it occurred in the input */
+ break;
+ }
+ }
+
+ /* Here, we are either done with the whole property name, if it was simple;
+ * or are positioned just after the '=' if it is compound. */
+
+ if (equals_pos >= 0) {
+ assert(! stricter); /* We shouldn't have set this yet */
+
+ /* Space immediately after the '=' is ignored */
+ i++;
+ for (; i < len; i++) {
+ if (! isSPACE(name[i])) {
+ break;
+ }
+ }
+
+ /* Certain properties need special handling. They may optionally be
+ * prefixed by 'is'. Ignore that prefix for the purposes of checking
+ * if this is one of those properties */
+ if (memBEGINPs(lookup_name, len, "is")) {
+ lookup_offset = 2;
+ }
+
+ /* Then check if it is one of these properties. This is hard-coded
+ * because easier this way, and the list is unlikely to change. There
+ * are several properties like this in the Unihan DB, which is unlikely
+ * to be compiled, and they all end with 'numeric'. The interiors
+ * aren't checked for the precise property. This would stop working if
+ * a cjk property were to be created that ended with 'numeric' and
+ * wasn't a numeric type */
+ is_nv_type = memEQs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "numericvalue")
+ || memEQs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "nv")
+ || ( memENDPs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "numeric")
+ && ( memBEGINPs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "cjk")
+ || memBEGINPs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "k")));
+ if ( is_nv_type
+ || memEQs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "canonicalcombiningclass")
+ || memEQs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "ccc")
+ || memEQs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "age")
+ || memEQs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "in")
+ || memEQs(lookup_name + lookup_offset,
+ j - 1 - lookup_offset, "presentin"))
+ {
+ unsigned int k;
+
+ /* What makes these properties special is that the stuff after the
+ * '=' is a number. Therefore, we can't throw away '-'
+ * willy-nilly, as those could be a minus sign. Other stricter
+ * rules also apply. However, these properties all can have the
+ * rhs not be a number, in which case they contain at least one
+ * alphabetic. In those cases, the stricter rules don't apply.
+ * But the numeric type properties can have the alphas [Ee] to
+ * signify an exponent, and it is still a number with stricter
+ * rules. So look for an alpha that signifys not-strict */
+ stricter = TRUE;
+ for (k = i; k < len; k++) {
+ if ( isALPHA(name[k])
+ && (! is_nv_type || ! isALPHA_FOLD_EQ(name[k], 'E')))
+ {
+ stricter = FALSE;
+ break;
+ }
+ }
+ }
+
+ if (stricter) {
+
+ /* A number may have a leading '+' or '-'. The latter is retained
+ * */
+ if (name[i] == '+') {
+ i++;
+ }
+ else if (name[i] == '-') {
+ lookup_name[j++] = '-';
+ i++;
+ }
+
+ /* Skip leading zeros including single underscores separating the
+ * zeros, or between the final leading zero and the first other
+ * digit */
+ for (; i < len - 1; i++) {
+ if ( name[i] != '0'
+ && (name[i] != '_' || ! isDIGIT(name[i+1])))
+ {
+ break;
+ }
+ }
+ }
+ }
+ else { /* No '=' */
+
+ /* We are now in a position to determine if this property should have
+ * been parsed using stricter rules. Only a few are like that, and
+ * unlikely to change. */
+ if ( memBEGINPs(lookup_name, j, "perl")
+ && memNEs(lookup_name + 4, j - 4, "space")
+ && memNEs(lookup_name + 4, j - 4, "word"))
+ {
+ stricter = TRUE;
+
+ /* We set the inputs back to 0 and the code below will reparse,
+ * using strict */
+ i = j = 0;
+ }
+ }
+
+ /* Here, we have either finished the property, or are positioned to parse
+ * the remainder, and we know if stricter rules apply. Finish out, if not
+ * already done */
+ for (; i < len; i++) {
+ char cur = name[i];
+
+ /* In all instances, case differences are ignored, and we normalize to
+ * lowercase */
+ if (isUPPER(cur)) {
+ lookup_name[j++] = toLOWER(cur);
+ continue;
+ }
+
+ /* An underscore is skipped, but not under strict rules unless it
+ * separates two digits */
+ if (cur == '_') {
+ if ( stricter
+ && ( i == 0 || (int) i == equals_pos || i == len- 1
+ || ! isDIGIT(name[i-1]) || ! isDIGIT(name[i+1])))
+ {
+ lookup_name[j++] = '_';
+ }
+ continue;
+ }
+
+ /* Hyphens are skipped except under strict */
+ if (cur == '-' && ! stricter) {
+ continue;
+ }
+
+ /* XXX Bug in documentation. It says white space skipped adjacent to
+ * non-word char. Maybe we should, but shouldn't skip it next to a dot
+ * in a number */
+ if (isSPACE(cur) && ! stricter) {
+ continue;
+ }
+
+ lookup_name[j++] = cur;
+
+ /* Unless this is a non-trailing slash, we are done with it */
+ if (i >= len - 1 || cur != '/') {
+ continue;
+ }
+
+ slash_pos = j;
+
+ /* A slash in the 'numeric value' property indicates that what follows
+ * is a denominator. It can have a leading '+' and '0's that should be
+ * skipped. But we have never allowed a negative denominator, so treat
+ * a minus like every other character. (No need to rule out a second
+ * '/', as that won't match anything anyway */
+ if (is_nv_type) {
+ i++;
+ if (i < len && name[i] == '+') {
+ i++;
+ }
+
+ /* Skip leading zeros including underscores separating digits */
+ for (; i < len - 1; i++) {
+ if ( name[i] != '0'
+ && (name[i] != '_' || ! isDIGIT(name[i+1])))
+ {
+ break;
+ }
+ }
+
+ /* Store the first real character in the denominator */
+ lookup_name[j++] = name[i];
+ }
+ }
+
+ /* Here are completely done parsing the input 'name', and 'lookup_name'
+ * contains a copy, normalized.
+ *
+ * This special case is grandfathered in: 'L_' and 'GC=L_' are accepted and
+ * different from without the underscores. */
+ if ( ( UNLIKELY(memEQs(lookup_name, j, "l"))
+ || UNLIKELY(memEQs(lookup_name, j, "gc=l")))
+ && UNLIKELY(name[len-1] == '_'))
+ {
+ lookup_name[j++] = '&';
+ }
+ else if (len > 2 && name[0] == 'I' && ( name[1] == 'n' || name[1] == 's'))
+ {
+
+ /* Also, if the original input began with 'In' or 'Is', it could be a
+ * subroutine call instead of a property names, which currently isn't
+ * handled by this function. Subroutine calls can't happen if there is
+ * an '=' in the name */
+ if (equals_pos < 0 && get_cvn_flags(name, len, GV_NOTQUAL) != NULL) {
+ return NULL;
+ }
+
+ starts_with_In_or_Is = TRUE;
+ }
+
+ lookup_len = j; /* Use a more mnemonic name starting here */
+
+ /* Get the index into our pointer table of the inversion list corresponding
+ * to the property */
+ table_index = match_uniprop((U8 *) lookup_name, lookup_len);
+
+ /* If it didn't find the property */
+ if (table_index == 0) {
+
+ /* If didn't find the property, we try again stripping off any initial
+ * 'In' or 'Is' */
+ if (starts_with_In_or_Is) {
+ lookup_name += 2;
+ lookup_len -= 2;
+ equals_pos -= 2;
+ slash_pos -= 2;
+
+ table_index = match_uniprop((U8 *) lookup_name, lookup_len);
+ }
+
+ if (table_index == 0) {
+ char * canonical;
+
+ /* If not found, and not a numeric type property, isn't a legal
+ * property */
+ if (! is_nv_type) {
+ return NULL;
+ }
+
+ /* But the numeric type properties need more work to decide. What
+ * we do is make sure we have the number in canonical form and look
+ * that up. */
+
+ if (slash_pos < 0) { /* No slash */
+
+ /* When it isn't a rational, take the input, convert it to a
+ * NV, then create a canonical string representation of that
+ * NV. */
+
+ NV value;
+
+ /* Get the value */
+ if (my_atof3(lookup_name + equals_pos, &value,
+ lookup_len - equals_pos)
+ != lookup_name + lookup_len)
+ {
+ return NULL;
+ }
+
+ /* If the value is an integer, the canonical value is integral */
+ if (Perl_ceil(value) == value) {
+ canonical = Perl_form(aTHX_ "%.*s%.0" NVff,
+ equals_pos, lookup_name, value);
+ }
+ else { /* Otherwise, it is %e with a known precision */
+ canonical = Perl_form(aTHX_ "%.*s%.*" NVef,
+ equals_pos, lookup_name,
+ PL_E_FORMAT_PRECISION, value);
+ }
+ }
+ else { /* Has a slash. Create a rational in canonical form */
+ UV numerator, denominator, gcd, trial;
+ const char * end_ptr;
+ const char * sign = "";
+
+ /* We can't just find the numerator, denominator, and do the
+ * division, then use the method above, because that is
+ * inexact. And the input could be a rational that is within
+ * epsilon (given our precision) of a valid rational, and would
+ * then incorrectly compare valid.
+ *
+ * We're only interested in the part after the '=' */
+ const char * this_lookup_name = lookup_name + equals_pos;
+ lookup_len -= equals_pos;
+ slash_pos -= equals_pos;
+
+ /* Handle any leading minus */
+ if (this_lookup_name[0] == '-') {
+ sign = "-";
+ this_lookup_name++;
+ lookup_len--;
+ slash_pos--;
+ }
+
+ /* Convert the numerator to numeric */
+ end_ptr = this_lookup_name + slash_pos;
+ if (! grok_atoUV(this_lookup_name, &numerator, &end_ptr)) {
+ return NULL;
+ }
+
+ /* It better have included all characters before the slash */
+ if (*end_ptr != '/') {
+ return NULL;
+ }
+
+ /* Set to look at just the denominator */
+ this_lookup_name += slash_pos;
+ lookup_len -= slash_pos;
+ end_ptr = this_lookup_name + lookup_len;
+
+ /* Convert the denominator to numeric */
+ if (! grok_atoUV(this_lookup_name, &denominator, &end_ptr)) {
+ return NULL;
+ }
+
+ /* It better be the rest of the characters, and don't divide by
+ * 0 */
+ if ( end_ptr != this_lookup_name + lookup_len
+ || denominator == 0)
+ {
+ return NULL;
+ }
+
+ /* Get the greatest common denominator using
+ http://en.wikipedia.org/wiki/Euclidean_algorithm */
+ gcd = numerator;
+ trial = denominator;
+ while (trial != 0) {
+ UV temp = trial;
+ trial = gcd % trial;
+ gcd = temp;
+ }
+
+ /* If already in lowest possible terms, we have already tried
+ * looking this up */
+ if (gcd == 1) {
+ return NULL;
+ }
+
+ /* Reduce the rational, which should put it in canonical form.
+ * Then look it up */
+ numerator /= gcd;
+ denominator /= gcd;
+
+ canonical = Perl_form(aTHX_ "%.*s%s%" UVuf "/%" UVuf,
+ equals_pos, lookup_name, sign, numerator, denominator);
+ }
+
+ /* Here, we have the number in canonical form. Try that */
+ table_index = match_uniprop((U8 *) canonical, strlen(canonical));
+ if (table_index == 0) {
+ return NULL;
+ }
+ }
+ }
+
+ /* The return is an index into a table of ptrs. A negative return
+ * signifies that the real index is the absolute value, but the result
+ * needs to be inverted */
+ if (table_index < 0) {
+ *invert = TRUE;
+ table_index = -table_index;
+ }
+ else {
+ *invert = FALSE;
+ }
+
+ /* Out-of band indices indicate a deprecated property. The proper index is
+ * modulo it with the table size. And dividing by the table size yields
+ * an offset into a table constructed to contain the corresponding warning
+ * message */
+ if (table_index > MAX_UNI_KEYWORD_INDEX) {
+ Size_t warning_offset = table_index / MAX_UNI_KEYWORD_INDEX;
+ table_index %= MAX_UNI_KEYWORD_INDEX;
+ Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
+ "Use of '%.*s' in \\p{} or \\P{} is deprecated because: %s",
+ (int) len, name, deprecated_property_msgs[warning_offset]);
+ }
+
+ /* In a few properties, a different property is used under /i. These are
+ * unlikely to change, so are hard-coded here. */
+ if (to_fold) {
+ if ( table_index == UNI_XPOSIXUPPER
+ || table_index == UNI_XPOSIXLOWER
+ || table_index == UNI_TITLE)
+ {
+ table_index = UNI_CASED;
+ }
+ else if ( table_index == UNI_UPPERCASELETTER
+ || table_index == UNI_LOWERCASELETTER
+# ifdef UNI_TITLECASELETTER /* Missing from early Unicodes */
+ || table_index == UNI_TITLECASELETTER
+# endif
+ ) {
+ table_index = UNI_CASEDLETTER;
+ }
+ else if ( table_index == UNI_POSIXUPPER
+ || table_index == UNI_POSIXLOWER)
+ {
+ table_index = UNI_POSIXALPHA;
+ }
+ }
+
+ /* Create and return the inversion list */
+ return _new_invlist_C_array(uni_prop_ptrs[table_index]);
+}
+
+#endif
+
/*
* ex: set ts=8 sts=4 sw=4 et:
*/