3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
11 /* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
14 #ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
18 # define Null(type) ((type)NULL)
21 =for apidoc_section $string
22 =for apidoc AmnU||Nullch
23 Null character pointer. (No longer available when C<PERL_CORE> is
26 =for apidoc_section $SV
27 =for apidoc AmnU||Nullsv
28 Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
32 Below are signatures of functions from config.h which can't easily be gleaned
33 from it, and are very unlikely to change
35 =for apidoc_section $signals
36 =for apidoc Am|int|Sigsetjmp|jmp_buf env|int savesigs
37 =for apidoc Am|void|Siglongjmp|jmp_buf env|int val
39 =for apidoc_section $filesystem
40 =for apidoc Am|void *|FILE_ptr|FILE * f
41 =for apidoc Am|Size_t|FILE_cnt|FILE * f
42 =for apidoc Am|void *|FILE_base|FILE * f
43 =for apidoc Am|Size_t|FILE_bufsiz|FILE *f
45 =for apidoc_section $string
46 =for apidoc Amu|token|CAT2|token x|token y
47 =for apidoc Amu|string|STRINGIFY|token x
49 =for apidoc_section $numeric
50 =for apidoc Am|double|Drand01
51 =for apidoc Am|void|seedDrand01|Rand_seed_t x
52 =for apidoc Am|char *|Gconvert|double x|Size_t n|bool t|char * b
57 # define Nullch Null(char*)
58 # define Nullfp Null(PerlIO*)
59 # define Nullsv Null(SV*)
72 =for apidoc_section $SV
73 =for apidoc Am |AV * |MUTABLE_AV |AV * p
74 =for apidoc_item |CV * |MUTABLE_CV |CV * p
75 =for apidoc_item |GV * |MUTABLE_GV |GV * p
76 =for apidoc_item |HV * |MUTABLE_HV |HV * p
77 =for apidoc_item |IO * |MUTABLE_IO |IO * p
78 =for apidoc_item |void *|MUTABLE_PTR|void * p
79 =for apidoc_item |SV * |MUTABLE_SV |SV * p
81 The C<MUTABLE_I<*>>() macros cast pointers to the types shown, in such a way
82 (compiler permitting) that casting away const-ness will give a warning;
86 AV *av1 = (AV*)sv; <== BAD: the const has been silently
88 AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
90 C<MUTABLE_PTR> is the base macro used to derive new casts. The other
91 already-built-in ones return pointers to what their names indicate.
95 The brace group version will raise a diagnostic if 'p' is const; the other
96 blindly casts away const.
98 #if defined(PERL_USE_GCC_BRACE_GROUPS)
99 # define MUTABLE_PTR(p) ({ void *p_ = (p); p_; })
101 # define MUTABLE_PTR(p) ((void *) (p))
104 #define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
105 #define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
106 #define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
107 #define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
108 #define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
109 #define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
112 =for apidoc_section $SV
113 =for apidoc Am |AV *|AV_FROM_REF|SV * ref
114 =for apidoc_item |CV *|CV_FROM_REF|SV * ref
115 =for apidoc_item |HV *|HV_FROM_REF|SV * ref
117 The C<I<*>V_FROM_REF> macros extract the C<SvRV()> from a given reference SV
118 and return a suitably-cast to pointer to the referenced SV. When running
119 under C<-DDEBUGGING>, assertions are also applied that check that I<ref> is
120 definitely a reference SV that refers to an SV of the right type.
125 #if defined(DEBUGGING) && defined(PERL_USE_GCC_BRACE_GROUPS)
126 # define xV_FROM_REF(XV, ref) \
128 assert(SvROK(_ref)); \
129 assert(SvTYPE(SvRV(_ref)) == SVt_PV ## XV); \
130 (XV *)(SvRV(_ref)); })
132 # define xV_FROM_REF(XV, ref) ((XV *)(SvRV(ref)))
135 #define AV_FROM_REF(ref) xV_FROM_REF(AV, ref)
136 #define CV_FROM_REF(ref) xV_FROM_REF(CV, ref)
137 #define HV_FROM_REF(ref) xV_FROM_REF(HV, ref)
140 # include <stdbool.h>
144 =for apidoc_section $casting
145 =for apidoc Am|bool|cBOOL|bool expr
147 Cast-to-bool. When Perl was able to be compiled on pre-C99 compilers, a
148 C<(bool)> cast didn't necessarily do the right thing, so this macro was
149 created (and made somewhat complicated to work around bugs in old
150 compilers). Now, many years later, and C99 is used, this is no longer
151 required, but is kept for backwards compatibility.
155 #define cBOOL(cbool) ((bool) (cbool))
157 /* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
158 * XXX Should really be a Configure probe, with HAS__FUNCTION__
159 * and FUNCTION__ as results.
160 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
161 #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
162 # define FUNCTION__ __func__
163 # define SAFE_FUNCTION__ __func__
164 #elif (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tru64, -c99 not known, only -std1). */
165 # define FUNCTION__ ("")
166 # define SAFE_FUNCTION__ ("UNKNOWN")
168 # define FUNCTION__ __FUNCTION__ /* Common extension. */
169 # define SAFE_FUNCTION__ __FUNCTION__ /* Common extension. */
172 /* XXX A note on the perl source internal type system. The
173 original intent was that I32 be *exactly* 32 bits.
175 Currently, we only guarantee that I32 is *at least* 32 bits.
176 Specifically, if int is 64 bits, then so is I32. (This is the case
177 for the Cray.) This has the advantage of meshing nicely with
178 standard library calls (where we pass an I32 and the library is
179 expecting an int), but the disadvantage that an I32 is not 32 bits.
180 Andy Dougherty August 1996
182 There is no guarantee that there is *any* integral type with
183 exactly 32 bits. It is perfectly legal for a system to have
184 sizeof(short) == sizeof(int) == sizeof(long) == 8.
186 Similarly, there is no guarantee that I16 and U16 have exactly 16
189 For dealing with issues that may arise from various 32/64-bit
190 systems, we will ask Configure to check out
192 SHORTSIZE == sizeof(short)
193 INTSIZE == sizeof(int)
194 LONGSIZE == sizeof(long)
195 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
196 PTRSIZE == sizeof(void *)
197 DOUBLESIZE == sizeof(double)
198 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
202 #ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
203 # include <inttypes.h>
204 # ifdef INT32_MIN_BROKEN
206 # define INT32_MIN (-2147483647-1)
208 # ifdef INT64_MIN_BROKEN
210 # define INT64_MIN (-9223372036854775807LL-1)
226 /* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
227 Please search CHAR_MAX in perl.h for further details. */
229 # define U8_MAX UINT8_MAX
231 # define U8_MAX PERL_UCHAR_MAX
234 # define U8_MIN UINT8_MIN
236 # define U8_MIN PERL_UCHAR_MIN
240 # define I16_MAX INT16_MAX
242 # define I16_MAX PERL_SHORT_MAX
245 # define I16_MIN INT16_MIN
247 # define I16_MIN PERL_SHORT_MIN
250 # define U16_MAX UINT16_MAX
252 # define U16_MAX PERL_USHORT_MAX
255 # define U16_MIN UINT16_MIN
257 # define U16_MIN PERL_USHORT_MIN
261 # define I32_MAX INT32_MAX
263 # define I32_MAX PERL_INT_MAX
265 # define I32_MAX PERL_LONG_MAX
268 # define I32_MIN INT32_MIN
270 # define I32_MIN PERL_INT_MIN
272 # define I32_MIN PERL_LONG_MIN
275 # ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
276 # define U32_MAX UINT_MAX
278 # define U32_MAX 4294967295U
281 # define U32_MAX PERL_UINT_MAX
283 # define U32_MAX PERL_ULONG_MAX
286 # define U32_MIN UINT32_MIN
288 # define U32_MIN PERL_UINT_MIN
290 # define U32_MIN PERL_ULONG_MIN
294 =for apidoc_section $integer
295 =for apidoc Ay|| PERL_INT_FAST8_T
296 =for apidoc_item PERL_INT_FAST16_T
297 =for apidoc_item PERL_UINT_FAST8_T
298 =for apidoc_item PERL_UINT_FAST16_T
300 These are equivalent to the correspondingly-named C99 typedefs on platforms
301 that have those; they evaluate to C<int> and C<unsigned int> on platforms that
302 don't, so that you can portably take advantage of this C99 feature.
307 typedef int_fast8_t PERL_INT_FAST8_T;
308 typedef uint_fast8_t PERL_UINT_FAST8_T;
309 typedef int_fast16_t PERL_INT_FAST16_T;
310 typedef uint_fast16_t PERL_UINT_FAST16_T;
312 typedef int PERL_INT_FAST8_T;
313 typedef unsigned int PERL_UINT_FAST8_T;
314 typedef int PERL_INT_FAST16_T;
315 typedef unsigned int PERL_UINT_FAST16_T;
318 /* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
319 * anyone is grepping for it. So BIT_DIGITS gives the number of decimal digits
320 * required to represent any possible unsigned number containing N bits.
321 * TYPE_DIGITS gives the number of decimal digits required to represent any
322 * possible unsigned number of type T. */
323 #define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
324 #define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
325 #define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
327 /* Unused by core; should be deprecated */
328 #define Ctl(ch) ((ch) & 037)
330 #if defined(PERL_CORE) || defined(PERL_EXT)
332 # define MIN(a,b) ((a) < (b) ? (a) : (b))
335 # define MAX(a,b) ((a) > (b) ? (a) : (b))
339 /* Returns a boolean as to whether the input unsigned number is a power of 2
340 * (2**0, 2**1, etc). In other words if it has just a single bit set.
341 * If not, subtracting 1 would leave the uppermost bit set, so the & would
343 #if defined(PERL_CORE) || defined(PERL_EXT)
344 # define isPOWER_OF_2(n) ((n) && ((n) & ((n)-1)) == 0)
347 /* Returns a mask with the lowest n bits set */
348 #define nBIT_MASK(n) ((UINTMAX_C(1) << (n)) - 1)
350 /* The largest unsigned number that will fit into n bits */
351 #define nBIT_UMAX(n) nBIT_MASK(n)
354 =for apidoc_section $directives
355 =for apidoc Am||__ASSERT_|bool expr
357 This is a helper macro to avoid preprocessor issues, replaced by nothing
358 unless under DEBUGGING, where it expands to an assert of its argument,
359 followed by a comma (hence the comma operator). If we just used a straight
360 assert(), we would get a comma with nothing before it when not DEBUGGING.
364 We also use empty definition under Coverity since the __ASSERT_
365 checks often check for things that Really Cannot Happen, and Coverity
366 detects that and gets all excited. */
368 #if defined(DEBUGGING) && !defined(__COVERITY__) \
369 && ! defined(PERL_SMALL_MACRO_BUFFER)
370 # define __ASSERT_(statement) assert(statement),
372 # define __ASSERT_(statement)
376 =for apidoc_section $SV
378 =for apidoc Ama|SV*|newSVpvs|"literal string"
379 Like C<newSVpvn>, but takes a literal string instead of a
382 =for apidoc Ama|SV*|newSVpvs_flags|"literal string"|U32 flags
383 Like C<newSVpvn_flags>, but takes a literal string instead of
384 a string/length pair.
386 =for apidoc Ama|SV*|newSVpvs_share|"literal string"
387 Like C<newSVpvn_share>, but takes a literal string instead of
388 a string/length pair and omits the hash parameter.
390 =for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string"|I32 flags
391 Like C<sv_catpvn_flags>, but takes a literal string instead
392 of a string/length pair.
394 =for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string"
395 Like C<sv_catpvn_nomg>, but takes a literal string instead of
396 a string/length pair.
398 =for apidoc Am|void|sv_catpvs|SV* sv|"literal string"
399 Like C<sv_catpvn>, but takes a literal string instead of a
402 =for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string"
403 Like C<sv_catpvn_mg>, but takes a literal string instead of a
406 =for apidoc Am|SV *|sv_setref_pvs|SV *const rv|const char *const classname|"literal string"
407 Like C<sv_setref_pvn>, but takes a literal string instead of
408 a string/length pair.
410 =for apidoc_section $string
412 =for apidoc Ama|char*|savepvs|"literal string"
413 Like C<savepvn>, but takes a literal string instead of a
416 =for apidoc Ama|char*|savesharedpvs|"literal string"
417 A version of C<savepvs()> which allocates the duplicate string in memory
418 which is shared between threads.
420 =for apidoc_section $GV
422 =for apidoc Am|HV*|gv_stashpvs|"name"|I32 create
423 Like C<gv_stashpvn>, but takes a literal string instead of a
426 =for apidoc_section $HV
428 =for apidoc Am|SV**|hv_fetchs|HV* tb|"key"|I32 lval
429 Like C<hv_fetch>, but takes a literal string instead of a
431 =for apidoc_section $lexer
433 =for apidoc Amx|void|lex_stuff_pvs|"pv"|U32 flags
435 Like L</lex_stuff_pvn>, but takes a literal string instead of
436 a string/length pair.
441 #define ASSERT_IS_LITERAL(s) ("" s "")
444 =for apidoc_section $string
446 =for apidoc Amu|pair|STR_WITH_LEN|"literal string"
448 Returns two comma separated tokens of the input literal string, and its length.
449 This is convenience macro which helps out in some API calls.
450 Note that it can't be used as an argument to macros or functions that under
451 some configurations might be macros, which means that it requires the full
452 Perl_xxx(aTHX_ ...) form for any API calls where it's used.
457 #define STR_WITH_LEN(s) ASSERT_IS_LITERAL(s), (sizeof(s)-1)
459 /* STR_WITH_LEN() shortcuts */
460 #define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
461 #define newSVpvs_flags(str,flags) \
462 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
463 #define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
464 #define sv_catpvs_flags(sv, str, flags) \
465 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
466 #define sv_catpvs_nomg(sv, str) \
467 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
468 #define sv_catpvs(sv, str) \
469 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
470 #define sv_catpvs_mg(sv, str) \
471 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
472 #define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
473 #define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
474 #define sv_setref_pvs(rv, classname, str) \
475 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
476 #define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
477 #define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
478 #define gv_stashpvs(str, create) \
479 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
481 #define gv_fetchpvs(namebeg, flags, sv_type) \
482 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), flags, sv_type)
483 #define gv_fetchpvn gv_fetchpvn_flags
484 #define sv_catxmlpvs(dsv, str, utf8) \
485 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
488 #define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
490 #define get_cvs(str, flags) \
491 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
493 /* internal helpers */
495 #ifndef PERL_VERSION_MAJOR
496 # define PERL_VERSION_MAJOR PERL_REVISION
498 # undef PERL_REVISION /* We don't want code to be using these */
500 #ifndef PERL_VERSION_MINOR
501 # define PERL_VERSION_MINOR PERL_VERSION
505 #ifndef PERL_VERSION_PATCH
506 # define PERL_VERSION_PATCH PERL_SUBVERSION
508 # undef PERL_SUBVERSION
511 #define PERL_JNP_TO_DECIMAL_(maJor,miNor,Patch) \
512 /* '10*' leaves room for things like alpha, beta, releases */ \
513 (10 * ((maJor) * 1000000) + ((miNor) * 1000) + (Patch))
514 #define PERL_DECIMAL_VERSION_ \
515 PERL_JNP_TO_DECIMAL_(PERL_VERSION_MAJOR, PERL_VERSION_MINOR, \
519 =for apidoc_section $versioning
520 =for apidoc AmR|bool|PERL_VERSION_EQ|const U8 major|const U8 minor|const U8 patch
521 =for apidoc_item PERL_VERSION_GE
522 =for apidoc_item PERL_VERSION_GT
523 =for apidoc_item PERL_VERSION_LE
524 =for apidoc_item PERL_VERSION_LT
525 =for apidoc_item PERL_VERSION_NE
527 Returns whether or not the perl currently being compiled has the specified
528 relationship to the perl given by the parameters. For example,
530 #if PERL_VERSION_GT(5,24,2)
531 code that will only be compiled on perls after v5.24.2
536 Note that this is usable in making compile-time decisions
538 You may use the special value '*' for the final number to mean ALL possible
541 #if PERL_VERSION_EQ(5,31,'*')
543 means all perls in the 5.31 series. And
545 #if PERL_VERSION_NE(5,24,'*')
547 means all perls EXCEPT 5.24 ones. And
549 #if PERL_VERSION_LE(5,9,'*')
553 #if PERL_VERSION_LT(5,10,0)
555 This means you don't have to think so much when converting from the existing
556 deprecated C<PERL_VERSION> to using this macro:
558 #if PERL_VERSION <= 9
562 #if PERL_VERSION_LE(5,9,'*')
567 /* N.B. These don't work if the patch version is 42 or 92, as those are what
568 * '*' is in ASCII and EBCDIC respectively */
569 # define PERL_VERSION_EQ(j,n,p) \
571 ? ( (j) == PERL_VERSION_MAJOR \
572 && (n) == PERL_VERSION_MINOR) \
573 : (PERL_DECIMAL_VERSION_ == PERL_JNP_TO_DECIMAL_(j,n,p)))
574 # define PERL_VERSION_NE(j,n,p) (! PERL_VERSION_EQ(j,n,p))
576 # define PERL_VERSION_LT(j,n,p) /* < '*' effectively means < 0 */ \
577 (PERL_DECIMAL_VERSION_ < PERL_JNP_TO_DECIMAL_( (j), \
579 (((p) == '*') ? 0 : p)))
580 # define PERL_VERSION_GE(j,n,p) (! PERL_VERSION_LT(j,n,p))
582 # define PERL_VERSION_LE(j,n,p) /* <= '*' effectively means < n+1 */ \
583 (PERL_DECIMAL_VERSION_ < PERL_JNP_TO_DECIMAL_( (j), \
584 (((p) == '*') ? ((n)+1) : (n)), \
585 (((p) == '*') ? 0 : p)))
586 # define PERL_VERSION_GT(j,n,p) (! PERL_VERSION_LE(j,n,p))
589 =for apidoc_section $string
591 =for apidoc Am|bool|strNE|char* s1|char* s2
592 Test two C<NUL>-terminated strings to see if they are different. Returns true
595 =for apidoc Am|bool|strEQ|char* s1|char* s2
596 Test two C<NUL>-terminated strings to see if they are equal. Returns true or
599 =for apidoc Am|bool|strLT|char* s1|char* s2
600 Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
601 second, C<s2>. Returns true or false.
603 =for apidoc Am|bool|strLE|char* s1|char* s2
604 Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
605 equal to the second, C<s2>. Returns true or false.
607 =for apidoc Am|bool|strGT|char* s1|char* s2
608 Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
609 the second, C<s2>. Returns true or false.
611 =for apidoc Am|bool|strGE|char* s1|char* s2
612 Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
613 or equal to the second, C<s2>. Returns true or false.
615 =for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
616 Test two C<NUL>-terminated strings to see if they are different. The C<len>
617 parameter indicates the number of bytes to compare. Returns true or false. (A
618 wrapper for C<strncmp>).
620 =for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
621 Test two C<NUL>-terminated strings to see if they are equal. The C<len>
622 parameter indicates the number of bytes to compare. Returns true or false. (A
623 wrapper for C<strncmp>).
625 =for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
626 Test two buffers (which may contain embedded C<NUL> characters, to see if they
627 are equal. The C<len> parameter indicates the number of bytes to compare.
628 Returns true or false. It is undefined behavior if either of the buffers
629 doesn't contain at least C<len> bytes.
631 =for apidoc Am|bool|memEQs|char* s1|STRLEN l1|"s2"
632 Like L</memEQ>, but the second string is a literal enclosed in double quotes,
633 C<l1> gives the number of bytes in C<s1>.
634 Returns true or false.
636 =for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
637 Test two buffers (which may contain embedded C<NUL> characters, to see if they
638 are not equal. The C<len> parameter indicates the number of bytes to compare.
639 Returns true or false. It is undefined behavior if either of the buffers
640 doesn't contain at least C<len> bytes.
642 =for apidoc Am|bool|memNEs|char* s1|STRLEN l1|"s2"
643 Like L</memNE>, but the second string is a literal enclosed in double quotes,
644 C<l1> gives the number of bytes in C<s1>.
645 Returns true or false.
647 =for apidoc Am|bool|memCHRs|"list"|char c
648 Returns the position of the first occurrence of the byte C<c> in the literal
649 string C<"list">, or NULL if C<c> doesn't appear in C<"list">. All bytes are
650 treated as unsigned char. Thus this macro can be used to determine if C<c> is
651 in a set of particular characters. Unlike L<strchr(3)>, it works even if C<c>
652 is C<NUL> (and the set doesn't include C<NUL>).
656 New macros should use the following conventions for their names (which are
657 based on the underlying C library functions):
659 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
661 Each has two main parameters, string-like operands that are compared
662 against each other, as specified by the macro name. Some macros may
663 additionally have one or potentially even two length parameters. If a length
664 parameter applies to both string parameters, it will be positioned third;
665 otherwise any length parameter immediately follows the string parameter it
668 If the prefix to the name is 'str', the string parameter is a pointer to a C
669 language string. Such a string does not contain embedded NUL bytes; its
670 length may be unknown, but can be calculated by C<strlen()>, since it is
671 terminated by a NUL, which isn't included in its length.
673 The optional 'n' following 'str' means that there is a third parameter,
674 giving the maximum number of bytes to look at in each string. Even if both
675 strings are longer than the length parameter, those extra bytes will be
678 The 's' suffix means that the 2nd byte string parameter is a literal C
679 double-quoted string. Its length will automatically be calculated by the
680 macro, so no length parameter will ever be needed for it.
682 If the prefix is 'mem', the string parameters don't have to be C strings;
683 they may contain embedded NUL bytes, do not necessarily have a terminating
684 NUL, and their lengths can be known only through other means, which in
685 practice are additional parameter(s) passed to the function. All 'mem'
686 functions have at least one length parameter. Barring any 'l' or 's' suffix,
687 there is a single length parameter, in position 3, which applies to both
688 string parameters. The 's' suffix means, as described above, that the 2nd
689 string is a literal double-quoted C string (hence its length is calculated by
690 the macro, and the length parameter to the function applies just to the first
691 string parameter, and hence is positioned just after it). An 'l' suffix
692 means that the 2nd string parameter has its own length parameter, and the
693 signature will look like memFOOl(s1, l1, s2, l2).
695 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
696 substring of the 1st string. 'P' if present indicates that the substring
697 must be a "proper" one in tha mathematical sense that the first one must be
698 strictly larger than the 2nd.
703 #define strNE(s1,s2) (strcmp(s1,s2) != 0)
704 #define strEQ(s1,s2) (strcmp(s1,s2) == 0)
705 #define strLT(s1,s2) (strcmp(s1,s2) < 0)
706 #define strLE(s1,s2) (strcmp(s1,s2) <= 0)
707 #define strGT(s1,s2) (strcmp(s1,s2) > 0)
708 #define strGE(s1,s2) (strcmp(s1,s2) >= 0)
710 #define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
711 #define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
713 #define memEQ(s1,s2,l) (memcmp(((const void *) (s1)), ((const void *) (s2)), l) == 0)
714 #define memNE(s1,s2,l) (! memEQ(s1,s2,l))
716 /* memEQ and memNE where second comparand is a string constant */
717 #define memEQs(s1, l, s2) \
718 (((sizeof(s2)-1) == (l)) && memEQ((s1), ASSERT_IS_LITERAL(s2), (sizeof(s2)-1)))
719 #define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
721 /* Keep these private until we decide it was a good idea */
722 #if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
724 #define strBEGINs(s1,s2) (strncmp(s1,ASSERT_IS_LITERAL(s2), sizeof(s2)-1) == 0)
726 #define memBEGINs(s1, l, s2) \
727 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
728 && memEQ(s1, ASSERT_IS_LITERAL(s2), sizeof(s2)-1))
729 #define memBEGINPs(s1, l, s2) \
730 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) - 1 \
731 && memEQ(s1, ASSERT_IS_LITERAL(s2), sizeof(s2)-1))
732 #define memENDs(s1, l, s2) \
733 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
734 && memEQ(s1 + (l) - (sizeof(s2) - 1), ASSERT_IS_LITERAL(s2), sizeof(s2)-1))
735 #define memENDPs(s1, l, s2) \
736 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) \
737 && memEQ(s1 + (l) - (sizeof(s2) - 1), ASSERT_IS_LITERAL(s2), sizeof(s2)-1))
738 #endif /* End of making macros private */
740 #define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
741 #define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
742 #define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
743 #define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
745 #define memCHRs(s1,c) ((const char *) memchr(ASSERT_IS_LITERAL(s1) , c, sizeof(s1)-1))
750 * Unfortunately, the introduction of locales means that we
751 * can't trust isupper(), etc. to tell the truth. And when
752 * it comes to /\w+/ with tainting enabled, we *must* be able
753 * to trust our character classes.
755 * Therefore, the default tests in the text of Perl will be independent of
756 * locale. Any code that wants to depend on the current locale will use the
757 * macros that contain _LC in their names
760 #ifdef USE_LOCALE_CTYPE
768 =head1 Character classification
769 This section is about functions (really macros) that classify characters
770 into types, such as punctuation versus alphabetic, etc. Most of these are
771 analogous to regular expression character classes. (See
772 L<perlrecharclass/POSIX Character Classes>.) There are several variants for
773 each class. (Not all macros have all variants; each item below lists the
774 ones valid for it.) None are affected by C<use bytes>, and only the ones
775 with C<LC> in the name are affected by the current locale.
777 The base function, e.g., C<isALPHA()>, takes any signed or unsigned value,
778 treating it as a code point, and returns a boolean as to whether or not the
779 character represented by it is (or on non-ASCII platforms, corresponds to) an
780 ASCII character in the named class based on platform, Unicode, and Perl rules.
781 If the input is a number that doesn't fit in an octet, FALSE is returned.
783 Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
784 with no suffix C<"_A">. This variant is used to emphasize by its name that
785 only ASCII-range characters can return TRUE.
787 Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
788 onto the platform. That is, the code points that are ASCII are unaffected,
789 since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
790 as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
791 true when called with the code point 0xDF, which is a word character in both
792 ASCII and EBCDIC (though it represents different characters in each).
793 If the input is a number that doesn't fit in an octet, FALSE is returned.
794 (Perl's documentation uses a colloquial definition of Latin-1, to include all
795 code points below 256.)
797 Variant C<isI<FOO>_uvchr> is exactly like the C<isI<FOO>_L1> variant, for
798 inputs below 256, but if the code point is larger than 255, Unicode rules are
799 used to determine if it is in the character class. For example,
800 C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
801 WITH MACRON in Unicode, and is a word character.
803 Variants C<isI<FOO>_utf8> and C<isI<FOO>_utf8_safe> are like C<isI<FOO>_uvchr>,
804 but are used for UTF-8 encoded strings. The two forms are different names for
805 the same thing. Each call to one of these classifies the first character of
806 the string starting at C<p>. The second parameter, C<e>, points to anywhere in
807 the string beyond the first character, up to one byte past the end of the
808 entire string. Although both variants are identical, the suffix C<_safe> in
809 one name emphasizes that it will not attempt to read beyond S<C<e - 1>>,
810 provided that the constraint S<C<s E<lt> e>> is true (this is asserted for in
811 C<-DDEBUGGING> builds). If the UTF-8 for the input character is malformed in
812 some way, the program may croak, or the function may return FALSE, at the
813 discretion of the implementation, and subject to change in future releases.
815 Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants,
816 but the result is based on the current locale, which is what C<LC> in the name
817 stands for. If Perl can determine that the current locale is a UTF-8 locale,
818 it uses the published Unicode rules; otherwise, it uses the C library function
819 that gives the named classification. For example, C<isDIGIT_LC()> when not in
820 a UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
821 returned if the input won't fit into an octet. On some platforms where the C
822 library function is known to be defective, Perl changes its result to follow
823 the POSIX standard's rules.
825 Variant C<isI<FOO>_LC_uvchr> acts exactly like C<isI<FOO>_LC> for inputs less
826 than 256, but for larger ones it returns the Unicode classification of the code
829 Variants C<isI<FOO>_LC_utf8> and C<isI<FOO>_LC_utf8_safe> are like
830 C<isI<FOO>_LC_uvchr>, but are used for UTF-8 encoded strings. The two forms
831 are different names for the same thing. Each call to one of these classifies
832 the first character of the string starting at C<p>. The second parameter,
833 C<e>, points to anywhere in the string beyond the first character, up to one
834 byte past the end of the entire string. Although both variants are identical,
835 the suffix C<_safe> in one name emphasizes that it will not attempt to read
836 beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
837 is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
838 character is malformed in some way, the program may croak, or the function may
839 return FALSE, at the discretion of the implementation, and subject to change in
842 =for apidoc Am|bool|isALPHA|UV ch
843 =for apidoc_item ||isALPHA_A|UV ch
844 =for apidoc_item ||isALPHA_LC|UV ch
845 =for apidoc_item ||isALPHA_LC_utf8_safe|U8 * s| U8 *end
846 =for apidoc_item ||isALPHA_LC_uvchr|UV ch
847 =for apidoc_item ||isALPHA_L1|UV ch
848 =for apidoc_item ||isALPHA_utf8|U8 * s|U8 * end
849 =for apidoc_item ||isALPHA_utf8_safe|U8 * s|U8 * end
850 =for apidoc_item ||isALPHA_uvchr|UV ch
851 Returns a boolean indicating whether the specified input is one of C<[A-Za-z]>,
852 analogous to C<m/[[:alpha:]]/>.
853 See the L<top of this section|/Character classification> for an explanation of
858 Here and below, we add the prototypes of these macros for downstream programs
859 that would be interested in them, such as Devel::PPPort
861 =for apidoc Am|bool|isALPHANUMERIC|UV ch
862 =for apidoc_item ||isALPHANUMERIC_A|UV ch
863 =for apidoc_item ||isALPHANUMERIC_LC|UV ch
864 =for apidoc_item ||isALPHANUMERIC_LC_utf8_safe|U8 * s| U8 *end
865 =for apidoc_item ||isALPHANUMERIC_LC_uvchr|UV ch
866 =for apidoc_item ||isALPHANUMERIC_L1|UV ch
867 =for apidoc_item ||isALPHANUMERIC_utf8|U8 * s|U8 * end
868 =for apidoc_item ||isALPHANUMERIC_utf8_safe|U8 * s|U8 * end
869 =for apidoc_item ||isALPHANUMERIC_uvchr|UV ch
870 Returns a boolean indicating whether the specified character is one of
871 C<[A-Za-z0-9]>, analogous to C<m/[[:alnum:]]/>.
872 See the L<top of this section|/Character classification> for an explanation of
875 =for apidoc Am|bool|isALNUMC|UV ch
876 =for apidoc_item ||isALNUMC_A|UV ch
877 =for apidoc_item ||isALNUMC_LC|UV ch
878 =for apidoc_item ||isALNUMC_LC_uvchr|UV ch
879 =for apidoc_item ||isALNUMC_L1|UV ch
880 These are discouraged, backward compatibility macros for L</C<isALPHANUMERIC>>.
881 That is, each returns a boolean indicating whether the specified character is
882 one of C<[A-Za-z0-9]>, analogous to C<m/[[:alnum:]]/>.
884 The C<C> suffix in the names was meant to indicate that they correspond to the
885 C language L<C<isalnum(3)>>.
887 =for apidoc Am|bool|isASCII|UV ch
888 =for apidoc_item ||isASCII_A|UV ch
889 =for apidoc_item ||isASCII_LC|UV ch
890 =for apidoc_item ||isASCII_LC_utf8_safe|U8 * s| U8 *end
891 =for apidoc_item ||isASCII_LC_uvchr|UV ch
892 =for apidoc_item ||isASCII_L1|UV ch
893 =for apidoc_item ||isASCII_utf8|U8 * s|U8 * end
894 =for apidoc_item ||isASCII_utf8_safe|U8 * s|U8 * end
895 =for apidoc_item ||isASCII_uvchr|UV ch
896 Returns a boolean indicating whether the specified character is one of the 128
897 characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
898 On non-ASCII platforms, it returns TRUE iff this
899 character corresponds to an ASCII character. Variants C<isASCII_A()> and
900 C<isASCII_L1()> are identical to C<isASCII()>.
901 See the L<top of this section|/Character classification> for an explanation of
903 Note, however, that some platforms do not have the C library routine
904 C<isascii()>. In these cases, the variants whose names contain C<LC> are the
905 same as the corresponding ones without.
907 Also note, that because all ASCII characters are UTF-8 invariant (meaning they
908 have the exact same representation (always a single byte) whether encoded in
909 UTF-8 or not), C<isASCII> will give the correct results when called with any
910 byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8> and
911 C<isASCII_utf8_safe> will work properly on any string encoded or not in UTF-8.
913 =for apidoc Am|bool|isBLANK|UV ch
914 =for apidoc_item ||isBLANK_A|UV ch
915 =for apidoc_item ||isBLANK_LC|UV ch
916 =for apidoc_item ||isBLANK_LC_utf8_safe|U8 * s| U8 *end
917 =for apidoc_item ||isBLANK_LC_uvchr|UV ch
918 =for apidoc_item ||isBLANK_L1|UV ch
919 =for apidoc_item ||isBLANK_utf8|U8 * s|U8 * end
920 =for apidoc_item ||isBLANK_utf8_safe|U8 * s|U8 * end
921 =for apidoc_item ||isBLANK_uvchr|UV ch
922 Returns a boolean indicating whether the specified character is a
923 character considered to be a blank, analogous to C<m/[[:blank:]]/>.
924 See the L<top of this section|/Character classification> for an explanation of
927 however, that some platforms do not have the C library routine
928 C<isblank()>. In these cases, the variants whose names contain C<LC> are
929 the same as the corresponding ones without.
931 =for apidoc Am|bool|isCNTRL|UV ch
932 =for apidoc_item ||isCNTRL_A|UV ch
933 =for apidoc_item ||isCNTRL_LC|UV ch
934 =for apidoc_item ||isCNTRL_LC_utf8_safe|U8 * s| U8 *end
935 =for apidoc_item ||isCNTRL_LC_uvchr|UV ch
936 =for apidoc_item ||isCNTRL_L1|UV ch
937 =for apidoc_item ||isCNTRL_utf8|U8 * s|U8 * end
938 =for apidoc_item ||isCNTRL_utf8_safe|U8 * s|U8 * end
939 =for apidoc_item ||isCNTRL_uvchr|UV ch
941 Returns a boolean indicating whether the specified character is a
942 control character, analogous to C<m/[[:cntrl:]]/>.
943 See the L<top of this section|/Character classification> for an explanation of
945 On EBCDIC platforms, you almost always want to use the C<isCNTRL_L1> variant.
947 =for apidoc Am|bool|isDIGIT|UV ch
948 =for apidoc_item ||isDIGIT_A|UV ch
949 =for apidoc_item ||isDIGIT_LC|UV ch
950 =for apidoc_item ||isDIGIT_LC_utf8_safe|U8 * s| U8 *end
951 =for apidoc_item ||isDIGIT_LC_uvchr|UV ch
952 =for apidoc_item ||isDIGIT_L1|UV ch
953 =for apidoc_item ||isDIGIT_utf8|U8 * s|U8 * end
954 =for apidoc_item ||isDIGIT_utf8_safe|U8 * s|U8 * end
955 =for apidoc_item ||isDIGIT_uvchr|UV ch
957 Returns a boolean indicating whether the specified character is a
958 digit, analogous to C<m/[[:digit:]]/>.
959 Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
960 See the L<top of this section|/Character classification> for an explanation of
963 =for apidoc Am|bool|isGRAPH|UV ch
964 =for apidoc_item ||isGRAPH_A|UV ch
965 =for apidoc_item ||isGRAPH_LC|UV ch
966 =for apidoc_item ||isGRAPH_LC_utf8_safe|U8 * s| U8 *end
967 =for apidoc_item ||isGRAPH_LC_uvchr|UV ch
968 =for apidoc_item ||isGRAPH_L1|UV ch
969 =for apidoc_item ||isGRAPH_utf8|U8 * s|U8 * end
970 =for apidoc_item ||isGRAPH_utf8_safe|U8 * s|U8 * end
971 =for apidoc_item ||isGRAPH_uvchr|UV ch
972 Returns a boolean indicating whether the specified character is a
973 graphic character, analogous to C<m/[[:graph:]]/>.
974 See the L<top of this section|/Character classification> for an explanation of
977 =for apidoc Am|bool|isLOWER|UV ch
978 =for apidoc_item ||isLOWER_A|UV ch
979 =for apidoc_item ||isLOWER_LC|UV ch
980 =for apidoc_item ||isLOWER_LC_utf8_safe|U8 * s| U8 *end
981 =for apidoc_item ||isLOWER_LC_uvchr|UV ch
982 =for apidoc_item ||isLOWER_L1|UV ch
983 =for apidoc_item ||isLOWER_utf8|U8 * s|U8 * end
984 =for apidoc_item ||isLOWER_utf8_safe|U8 * s|U8 * end
985 =for apidoc_item ||isLOWER_uvchr|UV ch
986 Returns a boolean indicating whether the specified character is a
987 lowercase character, analogous to C<m/[[:lower:]]/>.
988 See the L<top of this section|/Character classification> for an explanation of
991 =for apidoc Am|bool|isOCTAL|UV ch
992 =for apidoc_item ||isOCTAL_A|UV ch
993 =for apidoc_item ||isOCTAL_L1|UV ch
994 Returns a boolean indicating whether the specified character is an
996 The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
999 =for apidoc Am|bool|isPUNCT|UV ch
1000 =for apidoc_item ||isPUNCT_A|UV ch
1001 =for apidoc_item ||isPUNCT_LC|UV ch
1002 =for apidoc_item ||isPUNCT_LC_utf8_safe|U8 * s| U8 *end
1003 =for apidoc_item ||isPUNCT_LC_uvchr|UV ch
1004 =for apidoc_item ||isPUNCT_L1|UV ch
1005 =for apidoc_item ||isPUNCT_utf8|U8 * s|U8 * end
1006 =for apidoc_item ||isPUNCT_utf8_safe|U8 * s|U8 * end
1007 =for apidoc_item ||isPUNCT_uvchr|UV ch
1008 Returns a boolean indicating whether the specified character is a
1009 punctuation character, analogous to C<m/[[:punct:]]/>.
1010 Note that the definition of what is punctuation isn't as
1011 straightforward as one might desire. See L<perlrecharclass/POSIX Character
1012 Classes> for details.
1013 See the L<top of this section|/Character classification> for an explanation of
1016 =for apidoc Am|bool|isSPACE|UV ch
1017 =for apidoc_item ||isSPACE_A|UV ch
1018 =for apidoc_item ||isSPACE_LC|UV ch
1019 =for apidoc_item ||isSPACE_LC_utf8_safe|U8 * s| U8 *end
1020 =for apidoc_item ||isSPACE_LC_uvchr|UV ch
1021 =for apidoc_item ||isSPACE_L1|UV ch
1022 =for apidoc_item ||isSPACE_utf8|U8 * s|U8 * end
1023 =for apidoc_item ||isSPACE_utf8_safe|U8 * s|U8 * end
1024 =for apidoc_item ||isSPACE_uvchr|UV ch
1025 Returns a boolean indicating whether the specified character is a
1026 whitespace character. This is analogous
1027 to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
1028 this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
1029 locale forms of this macro (the ones with C<LC> in their names) matched
1030 precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
1031 in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
1032 (See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
1033 See the L<top of this section|/Character classification> for an explanation of
1036 =for apidoc Am|bool|isPSXSPC|UV ch
1037 =for apidoc_item ||isPSXSPC_A|UV ch
1038 =for apidoc_item ||isPSXSPC_LC|UV ch
1039 =for apidoc_item ||isPSXSPC_LC_utf8_safe|U8 * s| U8 *end
1040 =for apidoc_item ||isPSXSPC_LC_uvchr|UV ch
1041 =for apidoc_item ||isPSXSPC_L1|UV ch
1042 =for apidoc_item ||isPSXSPC_utf8|U8 * s|U8 * end
1043 =for apidoc_item ||isPSXSPC_utf8_safe|U8 * s|U8 * end
1044 =for apidoc_item ||isPSXSPC_uvchr|UV ch
1045 (short for Posix Space)
1046 Starting in 5.18, this is identical in all its forms to the
1047 corresponding C<isSPACE()> macros.
1048 The locale forms of this macro are identical to their corresponding
1049 C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
1050 non-locale forms differ from their C<isSPACE()> forms only in that the
1051 C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
1052 Otherwise they are identical. Thus this macro is analogous to what
1053 C<m/[[:space:]]/> matches in a regular expression.
1054 See the L<top of this section|/Character classification> for an explanation of
1057 =for apidoc Am|bool|isUPPER|UV ch
1058 =for apidoc_item ||isUPPER_A|UV ch
1059 =for apidoc_item ||isUPPER_LC|UV ch
1060 =for apidoc_item ||isUPPER_LC_utf8_safe|U8 * s| U8 *end
1061 =for apidoc_item ||isUPPER_LC_uvchr|UV ch
1062 =for apidoc_item ||isUPPER_L1|UV ch
1063 =for apidoc_item ||isUPPER_utf8|U8 * s|U8 * end
1064 =for apidoc_item ||isUPPER_utf8_safe|U8 * s|U8 * end
1065 =for apidoc_item ||isUPPER_uvchr|UV ch
1066 Returns a boolean indicating whether the specified character is an
1067 uppercase character, analogous to C<m/[[:upper:]]/>.
1068 See the L<top of this section|/Character classification> for an explanation of
1071 =for apidoc Am|bool|isPRINT|UV ch
1072 =for apidoc_item ||isPRINT_A|UV ch
1073 =for apidoc_item ||isPRINT_LC|UV ch
1074 =for apidoc_item ||isPRINT_LC_utf8_safe|U8 * s| U8 *end
1075 =for apidoc_item ||isPRINT_LC_uvchr|UV ch
1076 =for apidoc_item ||isPRINT_L1|UV ch
1077 =for apidoc_item ||isPRINT_utf8|U8 * s|U8 * end
1078 =for apidoc_item ||isPRINT_utf8_safe|U8 * s|U8 * end
1079 =for apidoc_item ||isPRINT_uvchr|UV ch
1080 Returns a boolean indicating whether the specified character is a
1081 printable character, analogous to C<m/[[:print:]]/>.
1082 See the L<top of this section|/Character classification> for an explanation of
1085 =for apidoc Am|bool|isWORDCHAR|UV ch
1086 =for apidoc_item ||isWORDCHAR_A|UV ch
1087 =for apidoc_item ||isWORDCHAR_LC|UV ch
1088 =for apidoc_item ||isWORDCHAR_LC_utf8_safe|U8 * s| U8 *end
1089 =for apidoc_item ||isWORDCHAR_LC_uvchr|UV ch
1090 =for apidoc_item ||isWORDCHAR_L1|UV ch
1091 =for apidoc_item ||isWORDCHAR_utf8|U8 * s|U8 * end
1092 =for apidoc_item ||isWORDCHAR_utf8_safe|U8 * s|U8 * end
1093 =for apidoc_item ||isWORDCHAR_uvchr|UV ch
1094 Returns a boolean indicating whether the specified character is a character
1095 that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
1096 in a regular expression. A word character is an alphabetic character, a
1097 decimal digit, a connecting punctuation character (such as an underscore), or
1098 a "mark" character that attaches to one of those (like some sort of accent).
1100 See the L<top of this section|/Character classification> for an explanation of
1103 C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>,
1104 C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, C<isWORDCHAR_LC_utf8>, and
1105 C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
1106 include the platform's native underscore.
1108 =for apidoc Am|bool|isALNUM |UV ch
1109 =for apidoc_item ||isALNUM_A |UV ch
1110 =for apidoc_item ||isALNUM_LC |UV ch
1111 =for apidoc_item ||isALNUM_LC_uvchr|UV ch
1112 These are each a synonym for their respectively named L</C<isWORDCHAR>>
1115 They are provided for backward compatibility, even though a word character
1116 includes more than the standard C language meaning of alphanumeric.
1117 To get the C language definition, use the corresponding L</C<isALPHANUMERIC>>
1120 =for apidoc Am|bool|isXDIGIT|UV ch
1121 =for apidoc_item ||isXDIGIT_A|UV ch
1122 =for apidoc_item ||isXDIGIT_LC|UV ch
1123 =for apidoc_item ||isXDIGIT_LC_utf8_safe|U8 * s| U8 *end
1124 =for apidoc_item ||isXDIGIT_LC_uvchr|UV ch
1125 =for apidoc_item ||isXDIGIT_L1|UV ch
1126 =for apidoc_item ||isXDIGIT_utf8|U8 * s|U8 * end
1127 =for apidoc_item ||isXDIGIT_utf8_safe|U8 * s|U8 * end
1128 =for apidoc_item ||isXDIGIT_uvchr|UV ch
1129 Returns a boolean indicating whether the specified character is a hexadecimal
1130 digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
1131 and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
1132 See the L<top of this section|/Character classification> for an explanation of
1135 =for apidoc Am|bool|isIDFIRST|UV ch
1136 =for apidoc_item ||isIDFIRST_A|UV ch
1137 =for apidoc_item ||isIDFIRST_LC|UV ch
1138 =for apidoc_item ||isIDFIRST_LC_utf8_safe|U8 * s| U8 *end
1139 =for apidoc_item ||isIDFIRST_LC_uvchr|UV ch
1140 =for apidoc_item ||isIDFIRST_L1|UV ch
1141 =for apidoc_item ||isIDFIRST_utf8|U8 * s|U8 * end
1142 =for apidoc_item ||isIDFIRST_utf8_safe|U8 * s|U8 * end
1143 =for apidoc_item ||isIDFIRST_uvchr|UV ch
1144 Returns a boolean indicating whether the specified character can be the first
1145 character of an identifier. This is very close to, but not quite the same as
1146 the official Unicode property C<XID_Start>. The difference is that this
1147 returns true only if the input character also matches L</isWORDCHAR>.
1148 See the L<top of this section|/Character classification> for an explanation of
1151 =for apidoc Am|bool|isIDCONT|UV ch
1152 =for apidoc_item ||isIDCONT_A|UV ch
1153 =for apidoc_item ||isIDCONT_LC|UV ch
1154 =for apidoc_item ||isIDCONT_LC_utf8_safe|U8 * s| U8 *end
1155 =for apidoc_item ||isIDCONT_LC_uvchr|UV ch
1156 =for apidoc_item ||isIDCONT_L1|UV ch
1157 =for apidoc_item ||isIDCONT_utf8|U8 * s|U8 * end
1158 =for apidoc_item ||isIDCONT_utf8_safe|U8 * s|U8 * end
1159 =for apidoc_item ||isIDCONT_uvchr|UV ch
1160 Returns a boolean indicating whether the specified character can be the
1161 second or succeeding character of an identifier. This is very close to, but
1162 not quite the same as the official Unicode property C<XID_Continue>. The
1163 difference is that this returns true only if the input character also matches
1164 L</isWORDCHAR>. See the L<top of this section|/Character classification> for
1165 an explanation of the variants.
1167 =for apidoc_section $numeric
1169 =for apidoc Am|U8|READ_XDIGIT|char str*
1170 Returns the value of an ASCII-range hex digit and advances the string pointer.
1171 Behaviour is only well defined when isXDIGIT(*str) is true.
1173 =head1 Character case changing
1174 Perl uses "full" Unicode case mappings. This means that converting a single
1175 character to another case may result in a sequence of more than one character.
1176 For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
1177 character sequence C<SS>. This presents some complications The lowercase of
1178 all characters in the range 0..255 is a single character, and thus
1179 C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
1180 return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
1181 an API that does allow every possible legal result to be returned.) Likewise
1182 no other function that is crippled by not being able to give the correct
1183 results for the full range of possible inputs has been implemented here.
1185 =for apidoc Am|UV|toUPPER|UV cp
1186 =for apidoc_item |UV|toUPPER_A|UV cp
1187 =for apidoc_item |UV|toUPPER_utf8|U8* p|U8* e|U8* s|STRLEN* lenp
1188 =for apidoc_item |UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1189 =for apidoc_item |UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
1191 These all return the uppercase of a character. The differences are what domain
1192 they operate on, and whether the input is specified as a code point (those
1193 forms with a C<cp> parameter) or as a UTF-8 string (the others). In the latter
1194 case, the code point to use is the first one in the buffer of UTF-8 encoded
1195 code points, delineated by the arguments S<C<p .. e - 1>>.
1197 C<toUPPER> and C<toUPPER_A> are synonyms of each other. They return the
1198 uppercase of any lowercase ASCII-range code point. All other inputs are
1199 returned unchanged. Since these are macros, the input type may be any integral
1200 one, and the output will occupy the same number of bits as the input.
1202 There is no C<toUPPER_L1> nor C<toUPPER_LATIN1> as the uppercase of some code
1203 points in the 0..255 range is above that range or consists of multiple
1204 characters. Instead use C<toUPPER_uvchr>.
1206 C<toUPPER_uvchr> returns the uppercase of any Unicode code point. The return
1207 value is identical to that of C<toUPPER_A> for input code points in the ASCII
1208 range. The uppercase of the vast majority of Unicode code points is the same
1209 as the code point itself. For these, and for code points above the legal
1210 Unicode maximum, this returns the input code point unchanged. It additionally
1211 stores the UTF-8 of the result into the buffer beginning at C<s>, and its
1212 length in bytes into C<*lenp>. The caller must have made C<s> large enough to
1213 contain at least C<UTF8_MAXBYTES_CASE+1> bytes to avoid possible overflow.
1215 NOTE: the uppercase of a code point may be more than one code point. The
1216 return value of this function is only the first of these. The entire uppercase
1217 is returned in C<s>. To determine if the result is more than a single code
1218 point, you can do something like this:
1220 uc = toUPPER_uvchr(cp, s, &len);
1221 if (len > UTF8SKIP(s)) { is multiple code points }
1222 else { is a single code point }
1224 C<toUPPER_utf8> and C<toUPPER_utf8_safe> are synonyms of each other. The only
1225 difference between these and C<toUPPER_uvchr> is that the source for these is
1226 encoded in UTF-8, instead of being a code point. It is passed as a buffer
1227 starting at C<p>, with C<e> pointing to one byte beyond its end. The C<p>
1228 buffer may certainly contain more than one code point; but only the first one
1229 (up through S<C<e - 1>>) is examined. If the UTF-8 for the input character is
1230 malformed in some way, the program may croak, or the function may return the
1231 REPLACEMENT CHARACTER, at the discretion of the implementation, and subject to
1232 change in future releases.
1234 =for apidoc Am|UV|toFOLD|UV cp
1235 =for apidoc_item |UV|toFOLD_A|UV cp
1236 =for apidoc_item |UV|toFOLD_utf8|U8* p|U8* e|U8* s|STRLEN* lenp
1237 =for apidoc_item |UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1238 =for apidoc_item |UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
1240 These all return the foldcase of a character. "foldcase" is an internal case
1241 for C</i> pattern matching. If the foldcase of character A and the foldcase of
1242 character B are the same, they match caselessly; otherwise they don't.
1244 The differences in the forms are what domain they operate on, and whether the
1245 input is specified as a code point (those forms with a C<cp> parameter) or as a
1246 UTF-8 string (the others). In the latter case, the code point to use is the
1247 first one in the buffer of UTF-8 encoded code points, delineated by the
1248 arguments S<C<p .. e - 1>>.
1250 C<toFOLD> and C<toFOLD_A> are synonyms of each other. They return the
1251 foldcase of any ASCII-range code point. In this range, the foldcase is
1252 identical to the lowercase. All other inputs are returned unchanged. Since
1253 these are macros, the input type may be any integral one, and the output will
1254 occupy the same number of bits as the input.
1256 There is no C<toFOLD_L1> nor C<toFOLD_LATIN1> as the foldcase of some code
1257 points in the 0..255 range is above that range or consists of multiple
1258 characters. Instead use C<toFOLD_uvchr>.
1260 C<toFOLD_uvchr> returns the foldcase of any Unicode code point. The return
1261 value is identical to that of C<toFOLD_A> for input code points in the ASCII
1262 range. The foldcase of the vast majority of Unicode code points is the same
1263 as the code point itself. For these, and for code points above the legal
1264 Unicode maximum, this returns the input code point unchanged. It additionally
1265 stores the UTF-8 of the result into the buffer beginning at C<s>, and its
1266 length in bytes into C<*lenp>. The caller must have made C<s> large enough to
1267 contain at least C<UTF8_MAXBYTES_CASE+1> bytes to avoid possible overflow.
1269 NOTE: the foldcase of a code point may be more than one code point. The
1270 return value of this function is only the first of these. The entire foldcase
1271 is returned in C<s>. To determine if the result is more than a single code
1272 point, you can do something like this:
1274 uc = toFOLD_uvchr(cp, s, &len);
1275 if (len > UTF8SKIP(s)) { is multiple code points }
1276 else { is a single code point }
1278 C<toFOLD_utf8> and C<toFOLD_utf8_safe> are synonyms of each other. The only
1279 difference between these and C<toFOLD_uvchr> is that the source for these is
1280 encoded in UTF-8, instead of being a code point. It is passed as a buffer
1281 starting at C<p>, with C<e> pointing to one byte beyond its end. The C<p>
1282 buffer may certainly contain more than one code point; but only the first one
1283 (up through S<C<e - 1>>) is examined. If the UTF-8 for the input character is
1284 malformed in some way, the program may croak, or the function may return the
1285 REPLACEMENT CHARACTER, at the discretion of the implementation, and subject to
1286 change in future releases.
1288 =for apidoc Am|UV|toLOWER|UV cp
1289 =for apidoc_item |UV|toLOWER_A|UV cp
1290 =for apidoc_item |UV|toLOWER_LATIN1|UV cp
1291 =for apidoc_item |UV|toLOWER_LC|UV cp
1292 =for apidoc_item |UV|toLOWER_L1|UV cp
1293 =for apidoc_item |UV|toLOWER_utf8|U8* p|U8* e|U8* s|STRLEN* lenp
1294 =for apidoc_item |UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1295 =for apidoc_item |UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
1297 These all return the lowercase of a character. The differences are what domain
1298 they operate on, and whether the input is specified as a code point (those
1299 forms with a C<cp> parameter) or as a UTF-8 string (the others). In the latter
1300 case, the code point to use is the first one in the buffer of UTF-8 encoded
1301 code points, delineated by the arguments S<C<p .. e - 1>>.
1303 C<toLOWER> and C<toLOWER_A> are synonyms of each other. They return the
1304 lowercase of any uppercase ASCII-range code point. All other inputs are
1305 returned unchanged. Since these are macros, the input type may be any integral
1306 one, and the output will occupy the same number of bits as the input.
1308 C<toLOWER_L1> and C<toLOWER_LATIN1> are synonyms of each other. They behave
1309 identically as C<toLOWER> for ASCII-range input. But additionally will return
1310 the lowercase of any uppercase code point in the entire 0..255 range, assuming
1311 a Latin-1 encoding (or the EBCDIC equivalent on such platforms).
1313 C<toLOWER_LC> returns the lowercase of the input code point according to the
1314 rules of the current POSIX locale. Input code points outside the range 0..255
1315 are returned unchanged.
1317 C<toLOWER_uvchr> returns the lowercase of any Unicode code point. The return
1318 value is identical to that of C<toLOWER_L1> for input code points in the 0..255
1319 range. The lowercase of the vast majority of Unicode code points is the same
1320 as the code point itself. For these, and for code points above the legal
1321 Unicode maximum, this returns the input code point unchanged. It additionally
1322 stores the UTF-8 of the result into the buffer beginning at C<s>, and its
1323 length in bytes into C<*lenp>. The caller must have made C<s> large enough to
1324 contain at least C<UTF8_MAXBYTES_CASE+1> bytes to avoid possible overflow.
1326 NOTE: the lowercase of a code point may be more than one code point. The
1327 return value of this function is only the first of these. The entire lowercase
1328 is returned in C<s>. To determine if the result is more than a single code
1329 point, you can do something like this:
1331 uc = toLOWER_uvchr(cp, s, &len);
1332 if (len > UTF8SKIP(s)) { is multiple code points }
1333 else { is a single code point }
1335 C<toLOWER_utf8> and C<toLOWER_utf8_safe> are synonyms of each other. The only
1336 difference between these and C<toLOWER_uvchr> is that the source for these is
1337 encoded in UTF-8, instead of being a code point. It is passed as a buffer
1338 starting at C<p>, with C<e> pointing to one byte beyond its end. The C<p>
1339 buffer may certainly contain more than one code point; but only the first one
1340 (up through S<C<e - 1>>) is examined. If the UTF-8 for the input character is
1341 malformed in some way, the program may croak, or the function may return the
1342 REPLACEMENT CHARACTER, at the discretion of the implementation, and subject to
1343 change in future releases.
1345 =for apidoc Am|UV|toTITLE|UV cp
1346 =for apidoc_item |UV|toTITLE_A|UV cp
1347 =for apidoc_item |UV|toTITLE_utf8|U8* p|U8* e|U8* s|STRLEN* lenp
1348 =for apidoc_item |UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1349 =for apidoc_item |UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1351 These all return the titlecase of a character. The differences are what domain
1352 they operate on, and whether the input is specified as a code point (those
1353 forms with a C<cp> parameter) or as a UTF-8 string (the others). In the latter
1354 case, the code point to use is the first one in the buffer of UTF-8 encoded
1355 code points, delineated by the arguments S<C<p .. e - 1>>.
1357 C<toTITLE> and C<toTITLE_A> are synonyms of each other. They return the
1358 titlecase of any lowercase ASCII-range code point. In this range, the
1359 titlecase is identical to the uppercase. All other inputs are returned
1360 unchanged. Since these are macros, the input type may be any integral one, and
1361 the output will occupy the same number of bits as the input.
1363 There is no C<toTITLE_L1> nor C<toTITLE_LATIN1> as the titlecase of some code
1364 points in the 0..255 range is above that range or consists of multiple
1365 characters. Instead use C<toTITLE_uvchr>.
1367 C<toTITLE_uvchr> returns the titlecase of any Unicode code point. The return
1368 value is identical to that of C<toTITLE_A> for input code points in the ASCII
1369 range. The titlecase of the vast majority of Unicode code points is the same
1370 as the code point itself. For these, and for code points above the legal
1371 Unicode maximum, this returns the input code point unchanged. It additionally
1372 stores the UTF-8 of the result into the buffer beginning at C<s>, and its
1373 length in bytes into C<*lenp>. The caller must have made C<s> large enough to
1374 contain at least C<UTF8_MAXBYTES_CASE+1> bytes to avoid possible overflow.
1376 NOTE: the titlecase of a code point may be more than one code point. The
1377 return value of this function is only the first of these. The entire titlecase
1378 is returned in C<s>. To determine if the result is more than a single code
1379 point, you can do something like this:
1381 uc = toTITLE_uvchr(cp, s, &len);
1382 if (len > UTF8SKIP(s)) { is multiple code points }
1383 else { is a single code point }
1385 C<toTITLE_utf8> and C<toTITLE_utf8_safe> are synonyms of each other. The only
1386 difference between these and C<toTITLE_uvchr> is that the source for these is
1387 encoded in UTF-8, instead of being a code point. It is passed as a buffer
1388 starting at C<p>, with C<e> pointing to one byte beyond its end. The C<p>
1389 buffer may certainly contain more than one code point; but only the first one
1390 (up through S<C<e - 1>>) is examined. If the UTF-8 for the input character is
1391 malformed in some way, the program may croak, or the function may return the
1392 REPLACEMENT CHARACTER, at the discretion of the implementation, and subject to
1393 change in future releases.
1397 XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1398 really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1399 and aren't general purpose as they don't work on U+DF, and assert against that.
1400 and isCASED_LC, as it really is more of an internal thing.
1402 Note that these macros are repeated in Devel::PPPort, so should also be
1403 patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1408 void below because that's the best fit, and works for Devel::PPPort
1409 =for apidoc_section $integer
1410 =for apidoc AyT||WIDEST_UTYPE
1412 Yields the widest unsigned integer type on the platform, currently either
1413 C<U32> or C<U64>. This can be used in declarations such as
1419 my_uv = (WIDEST_UTYPE) val;
1424 #define WIDEST_UTYPE PERL_UINTMAX_T
1426 /* Where there could be some confusion, use this as a static assert in macros
1427 * to make sure that a parameter isn't a pointer. But some compilers can't
1428 * handle this. The only one known so far that doesn't is gcc 3.3.6; the check
1429 * below isn't thorough for such an old compiler, so may have to be revised if
1430 * experience so dictates. */
1431 #if ! PERL_IS_GCC || PERL_GCC_VERSION_GT(3,3,6)
1432 # define ASSERT_NOT_PTR(x) ((x) | 0)
1434 # define ASSERT_NOT_PTR(x) (x)
1437 /* Likewise, this is effectively a static assert to be used to guarantee the
1438 * parameter is a pointer
1440 * NOT suitable for void*
1442 #define ASSERT_IS_PTR(x) (__ASSERT_(sizeof(*(x))) (x))
1444 /* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1445 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1446 * bits of information are lost even on a 64-bit machine, but to get the
1447 * compiler to optimize it out if possible. This is because Configure makes
1448 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1449 * sizeof() guarantees that this evaluates to a constant true at compile time.
1451 * For Coverity, be always true, because otherwise Coverity thinks
1452 * it finds several expressions that are always true, independent
1453 * of operands. Well, they are, but that is kind of the point.
1455 #ifndef __COVERITY__
1456 /* The '| 0' part in ASSERT_NOT_PTR ensures a compiler error if c is not
1457 * integer (like e.g., a pointer) */
1458 # define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
1459 || (((WIDEST_UTYPE) ASSERT_NOT_PTR(c)) >> 8) == 0)
1461 # define FITS_IN_8_BITS(c) (1)
1464 /* Returns true if l <= c <= (l + n), where 'l' and 'n' are non-negative
1465 * Written this way so that after optimization, only one conditional test is
1466 * needed. (The NV casts stop any warnings about comparison always being true
1467 * if called with an unsigned. The cast preserves the sign, which is all we
1469 #define withinCOUNT(c, l, n) (__ASSERT_((NV) (l) >= 0) \
1470 __ASSERT_((NV) (n) >= 0) \
1471 withinCOUNT_KNOWN_VALID_((c), (l), (n)))
1473 /* For internal use only, this can be used in places where it is known that the
1474 * parameters to withinCOUNT() are valid, to avoid the asserts. For example,
1475 * inRANGE() below, calls this several times, but does all the necessary
1476 * asserts itself, once. The reason that this is necessary is that the
1477 * duplicate asserts were exceeding the internal limits of some compilers */
1478 #define withinCOUNT_KNOWN_VALID_(c, l, n) \
1479 ((((WIDEST_UTYPE) (c)) - ASSERT_NOT_PTR(l)) \
1480 <= ((WIDEST_UTYPE) ASSERT_NOT_PTR(n)))
1482 /* Returns true if c is in the range l..u, where 'l' is non-negative
1483 * Written this way so that after optimization, only one conditional test is
1485 #define inRANGE(c, l, u) (__ASSERT_((NV) (l) >= 0) __ASSERT_((u) >= (l)) \
1486 ( (sizeof(c) == sizeof(U8)) ? inRANGE_helper_(U8, (c), (l), ((u))) \
1487 : (sizeof(c) == sizeof(U16)) ? inRANGE_helper_(U16,(c), (l), ((u))) \
1488 : (sizeof(c) == sizeof(U32)) ? inRANGE_helper_(U32,(c), (l), ((u))) \
1489 : (__ASSERT_(sizeof(c) == sizeof(WIDEST_UTYPE)) \
1490 inRANGE_helper_(WIDEST_UTYPE,(c), (l), ((u))))))
1492 /* For internal use, this is used by machine-generated code which generates
1493 * known valid calls, with a known sizeof(). This avoids the extra code and
1494 * asserts that were exceeding internal limits of some compilers. */
1495 #define inRANGE_helper_(cast, c, l, u) \
1496 withinCOUNT_KNOWN_VALID_(((cast) (c)), (l), ((u) - (l)))
1499 # ifndef _ALL_SOURCE
1500 /* The native libc isascii() et.al. functions return the wrong results
1501 * on at least z/OS unless this is defined. */
1502 # error _ALL_SOURCE should probably be defined
1505 /* There is a simple definition of ASCII for ASCII platforms. But the
1506 * EBCDIC one isn't so simple, so is defined using table look-up like the
1507 * other macros below.
1509 * The cast here is used instead of '(c) >= 0', because some compilers emit
1510 * a warning that that test is always true when the parameter is an
1511 * unsigned type. khw supposes that it could be written as
1512 * && ((c) == '\0' || (c) > 0)
1513 * to avoid the message, but the cast will likely avoid extra branches even
1514 * with stupid compilers. */
1515 # define isASCII(c) (((WIDEST_UTYPE) ASSERT_NOT_PTR(c)) < 128)
1518 /* Take the eight possible bit patterns of the lower 3 bits and you get the
1519 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1520 * can be ignored. If the rest match '0', we have an octal */
1521 #define isOCTAL_A(c) ((((WIDEST_UTYPE) ASSERT_NOT_PTR(c)) & ~7) == '0')
1523 #ifdef H_PERL /* If have access to perl.h, lookup in its table */
1525 /* Character class numbers. For internal core Perl use only. The ones less
1526 * than 32 are used in PL_charclass[] and the ones up through the one that
1527 * corresponds to <HIGHEST_REGCOMP_DOT_H_SYNC_> are used by regcomp.h and
1528 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1529 * their actual definitions are here. If that file has a name not used here,
1532 * The first group of these is ordered in what I (khw) estimate to be the
1533 * frequency of their use. This gives a slight edge to exiting a loop earlier
1534 * (in reginclass() in regexec.c). Except \v should be last, as it isn't a
1535 * real Posix character class, and some (small) inefficiencies in regular
1536 * expression handling would be introduced by putting it in the middle of those
1537 * that are. Also, cntrl and ascii come after the others as it may be useful
1538 * to group these which have no members that match above Latin1, (or above
1539 * ASCII in the latter case) */
1541 # define CC_WORDCHAR_ 0 /* \w and [:word:] */
1542 # define CC_DIGIT_ 1 /* \d and [:digit:] */
1543 # define CC_ALPHA_ 2 /* [:alpha:] */
1544 # define CC_LOWER_ 3 /* [:lower:] */
1545 # define CC_UPPER_ 4 /* [:upper:] */
1546 # define CC_PUNCT_ 5 /* [:punct:] */
1547 # define CC_PRINT_ 6 /* [:print:] */
1548 # define CC_ALPHANUMERIC_ 7 /* [:alnum:] */
1549 # define CC_GRAPH_ 8 /* [:graph:] */
1550 # define CC_CASED_ 9 /* [:lower:] or [:upper:] under /i */
1551 # define CC_SPACE_ 10 /* \s, [:space:] */
1552 # define CC_BLANK_ 11 /* [:blank:] */
1553 # define CC_XDIGIT_ 12 /* [:xdigit:] */
1554 # define CC_CNTRL_ 13 /* [:cntrl:] */
1555 # define CC_ASCII_ 14 /* [:ascii:] */
1556 # define CC_VERTSPACE_ 15 /* \v */
1558 # define HIGHEST_REGCOMP_DOT_H_SYNC_ CC_VERTSPACE_
1560 /* The members of the third group below do not need to be coordinated with data
1561 * structures in regcomp.[ch] and regexec.c. */
1562 # define CC_IDFIRST_ 16
1563 # define CC_CHARNAME_CONT_ 17
1564 # define CC_NONLATIN1_FOLD_ 18
1565 # define CC_NONLATIN1_SIMPLE_FOLD_ 19
1566 # define CC_QUOTEMETA_ 20
1567 # define CC_NON_FINAL_FOLD_ 21
1568 # define CC_IS_IN_SOME_FOLD_ 22
1569 # define CC_BINDIGIT_ 23
1570 # define CC_OCTDIGIT_ 24
1571 # define CC_MNEMONIC_CNTRL_ 25
1574 * If more bits are needed, one could add a second word for non-64bit
1575 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
1576 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1577 * is used only for optimization (as of this writing), and differs in the
1578 * Latin1 range from the ALPHA bit only in two relatively unimportant
1579 * characters: the masculine and feminine ordinal indicators, so removing it
1580 * would just cause /i regexes which match them to run less efficiently.
1581 * Similarly the EBCDIC-only bits are used just for speed, and could be
1582 * replaced by other means */
1584 #if defined(PERL_CORE) || defined(PERL_EXT)
1585 /* An enum version of the character class numbers, to help compilers
1588 CC_ENUM_ALPHA_ = CC_ALPHA_,
1589 CC_ENUM_ALPHANUMERIC_ = CC_ALPHANUMERIC_,
1590 CC_ENUM_ASCII_ = CC_ASCII_,
1591 CC_ENUM_BLANK_ = CC_BLANK_,
1592 CC_ENUM_CASED_ = CC_CASED_,
1593 CC_ENUM_CNTRL_ = CC_CNTRL_,
1594 CC_ENUM_DIGIT_ = CC_DIGIT_,
1595 CC_ENUM_GRAPH_ = CC_GRAPH_,
1596 CC_ENUM_LOWER_ = CC_LOWER_,
1597 CC_ENUM_PRINT_ = CC_PRINT_,
1598 CC_ENUM_PUNCT_ = CC_PUNCT_,
1599 CC_ENUM_SPACE_ = CC_SPACE_,
1600 CC_ENUM_UPPER_ = CC_UPPER_,
1601 CC_ENUM_VERTSPACE_ = CC_VERTSPACE_,
1602 CC_ENUM_WORDCHAR_ = CC_WORDCHAR_,
1603 CC_ENUM_XDIGIT_ = CC_XDIGIT_
1604 } char_class_number_;
1607 #define POSIX_CC_COUNT (HIGHEST_REGCOMP_DOT_H_SYNC_ + 1)
1611 EXTCONST U32 PL_charclass[] = {
1612 # include "l1_char_class_tab.h"
1615 # else /* ! DOINIT */
1616 EXTCONST U32 PL_charclass[];
1620 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
1621 # define CC_mask_(classnum) (1U << (classnum))
1623 /* For internal core Perl use only: the base macro for defining macros like
1625 # define generic_isCC_(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
1626 && (PL_charclass[(U8) (c)] & CC_mask_(classnum)))
1628 /* The mask for the _A versions of the macros; it just adds in the bit for
1630 # define CC_mask_A_(classnum) (CC_mask_(classnum) | CC_mask_(CC_ASCII_))
1632 /* For internal core Perl use only: the base macro for defining macros like
1633 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1634 * the ASCII bit are present */
1635 # define generic_isCC_A_(c, classnum) (FITS_IN_8_BITS(c) \
1636 && ((PL_charclass[(U8) (c)] & CC_mask_A_(classnum)) \
1637 == CC_mask_A_(classnum)))
1639 /* On ASCII platforms certain classes form a single range. It's faster to
1640 * special case these. isDIGIT is a single range on all platforms */
1642 # define isALPHA_A(c) generic_isCC_A_(c, CC_ALPHA_)
1643 # define isGRAPH_A(c) generic_isCC_A_(c, CC_GRAPH_)
1644 # define isLOWER_A(c) generic_isCC_A_(c, CC_LOWER_)
1645 # define isPRINT_A(c) generic_isCC_A_(c, CC_PRINT_)
1646 # define isUPPER_A(c) generic_isCC_A_(c, CC_UPPER_)
1648 /* By folding the upper and lowercase, we can use a single range */
1649 # define isALPHA_A(c) inRANGE((~('A' ^ 'a') & (c)), 'A', 'Z')
1650 # define isGRAPH_A(c) inRANGE(c, ' ' + 1, 0x7e)
1651 # define isLOWER_A(c) inRANGE(c, 'a', 'z')
1652 # define isPRINT_A(c) inRANGE(c, ' ', 0x7e)
1653 # define isUPPER_A(c) inRANGE(c, 'A', 'Z')
1655 # define isALPHANUMERIC_A(c) generic_isCC_A_(c, CC_ALPHANUMERIC_)
1656 # define isBLANK_A(c) generic_isCC_A_(c, CC_BLANK_)
1657 # define isCNTRL_A(c) generic_isCC_A_(c, CC_CNTRL_)
1658 # define isDIGIT_A(c) inRANGE(c, '0', '9')
1659 # define isPUNCT_A(c) generic_isCC_A_(c, CC_PUNCT_)
1660 # define isSPACE_A(c) generic_isCC_A_(c, CC_SPACE_)
1661 # define isWORDCHAR_A(c) generic_isCC_A_(c, CC_WORDCHAR_)
1662 # define isXDIGIT_A(c) generic_isCC_(c, CC_XDIGIT_) /* No non-ASCII xdigits */
1663 # define isIDFIRST_A(c) generic_isCC_A_(c, CC_IDFIRST_)
1664 # define isALPHA_L1(c) generic_isCC_(c, CC_ALPHA_)
1665 # define isALPHANUMERIC_L1(c) generic_isCC_(c, CC_ALPHANUMERIC_)
1666 # define isBLANK_L1(c) generic_isCC_(c, CC_BLANK_)
1668 /* continuation character for legal NAME in \N{NAME} */
1669 # define isCHARNAME_CONT(c) generic_isCC_(c, CC_CHARNAME_CONT_)
1671 # define isCNTRL_L1(c) generic_isCC_(c, CC_CNTRL_)
1672 # define isGRAPH_L1(c) generic_isCC_(c, CC_GRAPH_)
1673 # define isLOWER_L1(c) generic_isCC_(c, CC_LOWER_)
1674 # define isPRINT_L1(c) generic_isCC_(c, CC_PRINT_)
1675 # define isPSXSPC_L1(c) isSPACE_L1(c)
1676 # define isPUNCT_L1(c) generic_isCC_(c, CC_PUNCT_)
1677 # define isSPACE_L1(c) generic_isCC_(c, CC_SPACE_)
1678 # define isUPPER_L1(c) generic_isCC_(c, CC_UPPER_)
1679 # define isWORDCHAR_L1(c) generic_isCC_(c, CC_WORDCHAR_)
1680 # define isIDFIRST_L1(c) generic_isCC_(c, CC_IDFIRST_)
1683 # define isASCII(c) generic_isCC_(c, CC_ASCII_)
1686 /* Participates in a single-character fold with a character above 255 */
1687 # if defined(PERL_IN_REGCOMP_ANY) || defined(PERL_IN_REGEXEC_C)
1688 # define HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(c) \
1689 (( ! cBOOL(FITS_IN_8_BITS(c))) \
1690 || (PL_charclass[(U8) (c)] & CC_mask_(CC_NONLATIN1_SIMPLE_FOLD_)))
1692 # define IS_NON_FINAL_FOLD(c) generic_isCC_(c, CC_NON_FINAL_FOLD_)
1693 # define IS_IN_SOME_FOLD_L1(c) generic_isCC_(c, CC_IS_IN_SOME_FOLD_)
1696 /* Like the above, but also can be part of a multi-char fold */
1697 # define HAS_NONLATIN1_FOLD_CLOSURE(c) \
1698 ( (! cBOOL(FITS_IN_8_BITS(c))) \
1699 || (PL_charclass[(U8) (c)] & CC_mask_(CC_NONLATIN1_FOLD_)))
1701 # define _isQUOTEMETA(c) generic_isCC_(c, CC_QUOTEMETA_)
1703 /* is c a control character for which we have a mnemonic? */
1704 # if defined(PERL_CORE) || defined(PERL_EXT)
1705 # define isMNEMONIC_CNTRL(c) generic_isCC_(c, CC_MNEMONIC_CNTRL_)
1707 #else /* else we don't have perl.h H_PERL */
1709 /* If we don't have perl.h, we are compiling a utility program. Below we
1710 * hard-code various macro definitions that wouldn't otherwise be available
1711 * to it. Most are coded based on first principles. These are written to
1712 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
1713 # define isDIGIT_A(c) inRANGE(c, '0', '9')
1714 # define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
1715 # define isSPACE_A(c) (isBLANK_A(c) \
1720 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1721 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1722 * performance (if optimization isn't on), and allow the same code to be
1723 * used for both platform types */
1724 # define isLOWER_A(c) inRANGE((c), 'a', 'i') \
1725 || inRANGE((c), 'j', 'r') \
1726 || inRANGE((c), 's', 'z')
1727 # define isUPPER_A(c) inRANGE((c), 'A', 'I') \
1728 || inRANGE((c), 'J', 'R') \
1729 || inRANGE((c), 'S', 'Z')
1730 # define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1731 # define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
1732 # define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
1733 # define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
1734 # define isXDIGIT_A(c) ( isDIGIT_A(c) \
1735 || inRANGE((c), 'a', 'f') \
1736 || inRANGE((c), 'A', 'F')
1737 # define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1738 || (c) == '#' || (c) == '$' || (c) == '%' \
1739 || (c) == '&' || (c) == '\'' || (c) == '(' \
1740 || (c) == ')' || (c) == '*' || (c) == '+' \
1741 || (c) == ',' || (c) == '.' || (c) == '/' \
1742 || (c) == ':' || (c) == ';' || (c) == '<' \
1743 || (c) == '=' || (c) == '>' || (c) == '?' \
1744 || (c) == '@' || (c) == '[' || (c) == '\\' \
1745 || (c) == ']' || (c) == '^' || (c) == '_' \
1746 || (c) == '`' || (c) == '{' || (c) == '|' \
1747 || (c) == '}' || (c) == '~')
1748 # define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1749 # define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
1752 /* The below is accurate for the 3 EBCDIC code pages traditionally
1753 * supported by perl. The only difference between them in the controls
1754 * is the position of \n, and that is represented symbolically below */
1755 # define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1756 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1757 || (c) == '\t' || (c) == '\v' \
1758 || inRANGE((c), 1, 3) /* SOH, STX, ETX */ \
1759 || (c) == 7F /* U+7F DEL */ \
1760 || inRANGE((c), 0x0E, 0x13) /* SO SI DLE \
1762 || (c) == 0x18 /* U+18 CAN */ \
1763 || (c) == 0x19 /* U+19 EOM */ \
1764 || inRANGE((c), 0x1C, 0x1F) /* [FGRU]S */ \
1765 || (c) == 0x26 /* U+17 ETB */ \
1766 || (c) == 0x27 /* U+1B ESC */ \
1767 || (c) == 0x2D /* U+05 ENQ */ \
1768 || (c) == 0x2E /* U+06 ACK */ \
1769 || (c) == 0x32 /* U+16 SYN */ \
1770 || (c) == 0x37 /* U+04 EOT */ \
1771 || (c) == 0x3C /* U+14 DC4 */ \
1772 || (c) == 0x3D /* U+15 NAK */ \
1773 || (c) == 0x3F)/* U+1A SUB */
1774 # define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
1775 # else /* isASCII is already defined for ASCII platforms, so can use that to
1777 # define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
1780 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
1781 * during debugging, and it seems best to keep them. We may be called
1782 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1783 * do anything anyway, so make it not a problem */
1784 # if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1785 # define NATIVE_TO_LATIN1(ch) (ch)
1787 # define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1788 # define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1789 # define isBLANK_L1(c) (isBLANK_A(c) \
1790 || (FITS_IN_8_BITS(c) \
1791 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1792 # define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1793 # define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1794 # define isLOWER_L1(c) (isLOWER_A(c) \
1795 || (FITS_IN_8_BITS(c) \
1796 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
1797 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1798 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1799 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1800 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1801 # define isPRINT_L1(c) (isPRINT_A(c) \
1802 || (FITS_IN_8_BITS(c) \
1803 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
1804 # define isPUNCT_L1(c) (isPUNCT_A(c) \
1805 || (FITS_IN_8_BITS(c) \
1806 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
1807 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1808 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1809 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1810 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1811 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1812 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1813 # define isSPACE_L1(c) (isSPACE_A(c) \
1814 || (FITS_IN_8_BITS(c) \
1815 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
1816 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1817 # define isUPPER_L1(c) (isUPPER_A(c) \
1818 || (FITS_IN_8_BITS(c) \
1819 && ( IN_RANGE(NATIVE_TO_LATIN1((U8) c), \
1821 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1822 # define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1823 # define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1824 # define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1829 /* The following are not fully accurate in the above-ASCII range. I (khw)
1830 * don't think it's necessary to be so for the purposes where this gets
1832 # define isQUOTEMETA_(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1834 /* Many of the macros later in this file are defined in terms of these. By
1835 * implementing them with a function, which converts the class number into
1836 * a call to the desired macro, all of the later ones work. However, that
1837 * function won't be actually defined when building a utility program (no
1838 * perl.h), and so a compiler error will be generated if one is attempted
1839 * to be used. And the above-Latin1 code points require Unicode tables to
1840 * be present, something unlikely to be the case when bootstrapping */
1841 # define generic_isCC_(c, classnum) \
1842 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1843 # define generic_isCC_A_(c, classnum) \
1844 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
1845 #endif /* End of no perl.h H_PERL */
1847 #define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1848 #define isALPHA(c) isALPHA_A(c)
1849 #define isASCII_A(c) isASCII(c)
1850 #define isASCII_L1(c) isASCII(c)
1851 #define isBLANK(c) isBLANK_A(c)
1852 #define isCNTRL(c) isCNTRL_A(c)
1853 #define isDIGIT(c) isDIGIT_A(c)
1854 #define isGRAPH(c) isGRAPH_A(c)
1855 #define isIDFIRST(c) isIDFIRST_A(c)
1856 #define isLOWER(c) isLOWER_A(c)
1857 #define isPRINT(c) isPRINT_A(c)
1858 #define isPSXSPC_A(c) isSPACE_A(c)
1859 #define isPSXSPC(c) isPSXSPC_A(c)
1860 #define isPSXSPC_L1(c) isSPACE_L1(c)
1861 #define isPUNCT(c) isPUNCT_A(c)
1862 #define isSPACE(c) isSPACE_A(c)
1863 #define isUPPER(c) isUPPER_A(c)
1864 #define isWORDCHAR(c) isWORDCHAR_A(c)
1865 #define isXDIGIT(c) isXDIGIT_A(c)
1867 /* ASCII casing. These could also be written as
1868 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1869 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1870 which uses table lookup and mask instead of subtraction. (This would
1871 work because the _MOD does not apply in the ASCII range).
1873 These actually are UTF-8 invariant casing, not just ASCII, as any non-ASCII
1874 UTF-8 invariants are neither upper nor lower. (Only on EBCDIC platforms are
1875 there non-ASCII invariants, and all of them are controls.) */
1876 #define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1877 #define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
1879 /* In the ASCII range, these are equivalent to what they're here defined to be.
1880 * But by creating these definitions, other code doesn't have to be aware of
1881 * this detail. Actually this works for all UTF-8 invariants, not just the
1882 * ASCII range. (EBCDIC platforms can have non-ASCII invariants.) */
1883 #define toFOLD(c) toLOWER(c)
1884 #define toTITLE(c) toUPPER(c)
1886 #define toLOWER_A(c) toLOWER(c)
1887 #define toUPPER_A(c) toUPPER(c)
1888 #define toFOLD_A(c) toFOLD(c)
1889 #define toTITLE_A(c) toTITLE(c)
1891 /* Use table lookup for speed; returns the input itself if is out-of-range */
1892 #define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
1894 : PL_latin1_lc[ (U8) (c) ])
1895 #define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1897 /* Modified uc. Is correct uc except for three non-ascii chars which are
1898 * all mapped to one of them, and these need special handling; returns the
1899 * input itself if is out-of-range */
1900 #define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
1902 : PL_mod_latin1_uc[ (U8) (c) ])
1903 #ifdef USE_LOCALE_CTYPE
1904 # define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
1905 # define IN_UTF8_TURKIC_LOCALE PL_in_utf8_turkic_locale
1907 # define IN_UTF8_CTYPE_LOCALE false
1908 # define IN_UTF8_TURKIC_LOCALE false
1911 /* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1913 /* For internal core Perl use only: the base macro for defining macros like
1914 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
1915 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1916 * isFOO_L1(); 'classnum' is something like CC_UPPER_, which gives the class
1917 * number for doing this. For non-UTF-8 locales, the code to actually do the
1918 * test this is passed in 'non_utf8'. If 'c' is above 255, 0 is returned. For
1919 * accessing the full range of possible code points under locale rules, use the
1920 * macros based on generic_LC_uvchr_ instead of this. */
1921 #define generic_LC_base_(c, classnum, non_utf8_func) \
1922 (! FITS_IN_8_BITS(c) \
1924 : IN_UTF8_CTYPE_LOCALE \
1925 ? cBOOL(PL_charclass[(U8) (c)] & CC_mask_(classnum)) \
1926 : cBOOL(non_utf8_func(c)))
1928 /* A helper macro for defining macros like isALPHA_LC. On systems without
1929 * proper locales, these reduce to, e.g., isALPHA_A */
1931 # define generic_LC_(c, classnum, non_utf8_func) \
1932 generic_LC_base_(c, classnum, non_utf8_func)
1934 # define generic_LC_(c, classnum, non_utf8_func) \
1935 generic_isCC_A_(c, classnum)
1938 /* Below are the definitions for the locale-sensitive character classification
1939 * macros whose input domain is a byte, and the locale isn't UTF-8. These are
1940 * as close as possible to the bare versions on the platform and still yield
1941 * POSIX Standard-compliant results.
1943 * There is currently only one place these definitions should be used, in
1944 * certain function calls like Perl_iswordchar_() in inline.h.
1946 * Most likely you want to use the macros a ways below with names like
1947 * isALPHA_LC(). Rarely, you may want isU8_ALPHA_LC(), somewhat below.
1949 * The first two aren't in C89, so the fallback is to use the non-locale
1950 * sensitive versions; these are the same for all platforms */
1951 #if defined(HAS_ISASCII)
1952 # define is_posix_ASCII(c) isascii((U8) (c))
1954 # define is_posix_ASCII(c) isASCII(c)
1957 #if defined(HAS_ISBLANK)
1958 # define is_posix_BLANK(c) isblank((U8) (c))
1960 # define is_posix_BLANK(c) isBLANK(c)
1963 /* The next few are the same in all platforms. */
1964 #define is_posix_CNTRL(c) iscntrl((U8) (c))
1965 #define is_posix_IDFIRST(c) (UNLIKELY((c) == '_') || is_posix_ALPHA(c))
1966 #define is_posix_SPACE(c) isspace((U8) (c))
1967 #define is_posix_WORDCHAR(c) (UNLIKELY((c) == '_') || is_posix_ALPHANUMERIC(c))
1969 /* The base-level case changing macros are also the same in all platforms */
1970 #define to_posix_LOWER(c) tolower((U8) (c))
1971 #define to_posix_UPPER(c) toupper((U8) (c))
1972 #define to_posix_FOLD(c) to_posix_LOWER(c)
1976 /* The Windows functions don't bother to follow the POSIX standard, which for
1977 * example says that something can't both be a printable and a control. But
1978 * Windows treats \t as both a control and a printable, and does such things as
1979 * making superscripts into both digits and punctuation. These #defines tame
1980 * these flaws by assuming that the definitions of controls (and the other few
1981 * ones defined above) are correct, and then making sure that other definitions
1982 * don't have weirdnesses, by adding a check that \w and its subsets aren't
1983 * ispunct(), and things that are \W, like ispunct(), arent't controls. Not
1984 * all possible weirdnesses are checked for, just ones that were detected on
1985 * actual Microsoft code pages */
1986 # define is_posix_ALPHA(c) \
1987 (isalpha((U8) (c)) && ! is_posix_PUNCT(c))
1988 # define is_posix_ALPHANUMERIC(c) \
1989 (isalnum((U8) (c)) && ! is_posix_PUNCT(c))
1990 # define is_posix_CASED(c) \
1991 ((isupper((U8) (c)) || islower((U8) (c))) && ! is_posix_PUNCT(c))
1992 # define is_posix_DIGIT(c) \
1993 (isdigit((U8) (c)) && ! is_posix_PUNCT(c))
1994 # define is_posix_GRAPH(c) \
1995 (isgraph((U8) (c)) && ! is_posix_CNTRL(c))
1996 # define is_posix_LOWER(c) \
1997 (islower((U8) (c)) && ! is_posix_PUNCT(c))
1998 # define is_posix_PRINT(c) \
1999 (isprint((U8) (c)) && ! is_posix_CNTRL(c))
2000 # define is_posix_PUNCT(c) \
2001 (ispunct((U8) (c)) && ! is_posix_CNTRL(c))
2002 # define is_posix_UPPER(c) \
2003 (isupper((U8) (c)) && ! is_posix_PUNCT(c))
2004 # define is_posix_XDIGIT(c) \
2005 (isxdigit((U8) (c)) && ! is_posix_PUNCT(c))
2008 /* For all other platforms, as far as we know, isdigit(), etc. work sanely
2010 # define is_posix_ALPHA(c) isalpha((U8) (c))
2011 # define is_posix_ALPHANUMERIC(c) isalnum((U8) (c))
2012 # define is_posix_CASED(c) (islower((U8) (c)) || isupper((U8) (c)))
2013 # define is_posix_DIGIT(c) isdigit((U8) (c))
2015 /* ... But it seems that IBM products treat NBSP as both a space and a
2016 * graphic; these are the two platforms that we have active test beds for.
2018 # if defined(OS390) || defined(_AIX)
2019 # define is_posix_GRAPH(c) (isgraph((U8) (c)) && ! isspace((U8) (c)))
2021 # define is_posix_GRAPH(c) isgraph((U8) (c))
2023 # define is_posix_LOWER(c) islower((U8) (c))
2024 # define is_posix_PRINT(c) isprint((U8) (c))
2025 # define is_posix_PUNCT(c) ispunct((U8) (c))
2026 # define is_posix_UPPER(c) isupper((U8) (c))
2027 # define is_posix_XDIGIT(c) isxdigit((U8) (c))
2030 /* Below is the next level up, which currently expands to nothing more
2031 * than the previous layer. These are the macros to use if you really need
2032 * something whose input domain is a byte, and the locale isn't UTF-8; that is,
2033 * where you normally would have to use things like bare isalnum().
2035 * But most likely you should instead use the layer defined further below which
2036 * has names like isALPHA_LC. They deal with larger-than-byte inputs, and
2039 * (Note, proper general operation of the bare libc functions requires you to
2040 * cast to U8. These do that for you automatically.) */
2042 # define WRAP_U8_LC_(c, classnum, posix) posix(c)
2044 #define isU8_ALPHANUMERIC_LC(c) \
2045 WRAP_U8_LC_((c), CC_ALPHANUMERIC_, is_posix_ALPHANUMERIC)
2046 #define isU8_ALPHA_LC(c) WRAP_U8_LC_((c), CC_ALPHA_, is_posix_ALPHA)
2047 #define isU8_ASCII_LC(c) WRAP_U8_LC_((c), CC_ASCII_, is_posix_ASCII)
2048 #define isU8_BLANK_LC(c) WRAP_U8_LC_((c), CC_BLANK_, is_posix_BLANK)
2049 #define isU8_CASED_LC(c) WRAP_U8_LC_((c), CC_CASED_, is_posix_CASED)
2050 #define isU8_CNTRL_LC(c) WRAP_U8_LC_((c), CC_CNTRL_, is_posix_CNTRL)
2051 #define isU8_DIGIT_LC(c) WRAP_U8_LC_((c), CC_DIGIT_, is_posix_DIGIT)
2052 #define isU8_GRAPH_LC(c) WRAP_U8_LC_((c), CC_GRAPH_, is_posix_GRAPH)
2053 #define isU8_IDFIRST_LC(c) WRAP_U8_LC_((c), CC_IDFIRST_, is_posix_IDFIRST)
2054 #define isU8_LOWER_LC(c) WRAP_U8_LC_((c), CC_LOWER_, is_posix_LOWER)
2055 #define isU8_PRINT_LC(c) WRAP_U8_LC_((c), CC_PRINT_, is_posix_PRINT)
2056 #define isU8_PUNCT_LC(c) WRAP_U8_LC_((c), CC_PUNCT_, is_posix_PUNCT)
2057 #define isU8_SPACE_LC(c) WRAP_U8_LC_((c), CC_SPACE_, is_posix_SPACE)
2058 #define isU8_UPPER_LC(c) WRAP_U8_LC_((c), CC_UPPER_, is_posix_UPPER)
2059 #define isU8_WORDCHAR_LC(c) WRAP_U8_LC_((c), CC_WORDCHAR_, is_posix_WORDCHAR)
2060 #define isU8_XDIGIT_LC(c) WRAP_U8_LC_((c), CC_XDIGIT_, is_posix_XDIGIT)
2062 #define toU8_LOWER_LC(c) WRAP_U8_LC_((c), CC_TOLOWER_, to_posix_LOWER)
2063 #define toU8_UPPER_LC(c) WRAP_U8_LC_((c), CC_TOUPPER_, to_posix_UPPER)
2064 #define toU8_FOLD_LC(c) toU8_LOWER_LC(c)
2066 /* The definitions below use the ones above to create versions in which the
2067 * input domain isn't restricted to bytes (though always returning false if the
2068 * input doesn't fit in a byte), and to behave properly should the locale be
2069 * UTF-8. These are the documented ones, suitable for general use (though
2070 * toUPPER_LC and toFOLD_LC aren't documented because they need special
2071 * handling to deal with SHARP S expanding to two characters). */
2073 #define isASCII_LC(c) (FITS_IN_8_BITS(c) && isU8_ASCII_LC(c))
2074 #define isALPHA_LC(c) generic_LC_(c, CC_ALPHA_, isU8_ALPHA_LC)
2075 #define isALPHANUMERIC_LC(c) \
2076 generic_LC_(c, CC_ALPHANUMERIC_, isU8_ALPHANUMERIC_LC)
2077 #define isBLANK_LC(c) generic_LC_(c, CC_BLANK_, isU8_BLANK_LC)
2078 #define isCASED_LC(c) generic_LC_(c, CC_CASED_, isU8_CASED_LC)
2079 #define isCNTRL_LC(c) generic_LC_(c, CC_CNTRL_, isU8_CNTRL_LC)
2080 #define isDIGIT_LC(c) generic_LC_(c, CC_DIGIT_, isU8_DIGIT_LC)
2081 #define isGRAPH_LC(c) generic_LC_(c, CC_GRAPH_, isU8_GRAPH_LC)
2082 #define isIDFIRST_LC(c) generic_LC_(c, CC_IDFIRST_, isU8_IDFIRST_LC)
2083 #define isLOWER_LC(c) generic_LC_(c, CC_LOWER_, isU8_LOWER_LC)
2084 #define isPRINT_LC(c) generic_LC_(c, CC_PRINT_, isU8_PRINT_LC)
2085 #define isPUNCT_LC(c) generic_LC_(c, CC_PUNCT_, isU8_PUNCT_LC)
2086 #define isSPACE_LC(c) generic_LC_(c, CC_SPACE_, isU8_SPACE_LC)
2087 #define isUPPER_LC(c) generic_LC_(c, CC_UPPER_, isU8_UPPER_LC)
2088 #define isWORDCHAR_LC(c) generic_LC_(c, CC_WORDCHAR_, isU8_WORDCHAR_LC)
2089 #define isXDIGIT_LC(c) generic_LC_(c, CC_XDIGIT_, isU8_XDIGIT_LC)
2092 # define toLOWER_LC(c) toLOWER_A(c)
2093 # define toUPPER_LC(c) toUPPER_A(c)
2094 # define toFOLD_LC(c) toFOLD_A(c)
2097 /* In the next three macros, the reason for using the PL_latin arrays is in
2098 * case the system function is defective; it ensures uniform results that
2099 * conform to the Unicode standard. */
2101 /* This does not handle the anomalies in UTF-8 Turkic locales. */
2102 # define toLOWER_LC(c) ((! FITS_IN_8_BITS(c)) \
2104 : ((IN_UTF8_CTYPE_LOCALE) \
2105 ? PL_latin1_lc[ (U8) (c) ] \
2106 : ((U8) toU8_LOWER_LC(c))))
2108 /* In this macro, note that the result can be larger than a byte in a UTF-8
2109 * locale. It returns a single value, so can't adequately return the upper
2110 * case of LATIN SMALL LETTER SHARP S in a UTF-8 locale (which should be a
2111 * string of two values "SS"); instead it asserts against that under
2112 * DEBUGGING, and otherwise returns its input. It does not handle the
2113 * anomalies in UTF-8 Turkic locales. */
2114 # define toUPPER_LC(c) \
2115 ((! FITS_IN_8_BITS(c)) \
2117 : ((! IN_UTF8_CTYPE_LOCALE) \
2118 ? ((U8) toU8_UPPER_LC(c)) \
2119 : (UNLIKELY(((U8)(c)) == MICRO_SIGN) \
2120 ? GREEK_CAPITAL_LETTER_MU \
2121 : ((UNLIKELY(((U8) (c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
2122 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
2123 : (UNLIKELY(((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
2124 ? (__ASSERT_(0) (c)) /* Fail on Sharp S in DEBUGGING */ \
2125 : PL_mod_latin1_uc[ (U8) (c) ]))))))
2127 /* In this macro, note that the result can be larger than a byte in a UTF-8
2128 * locale. It returns a single value, so can't adequately return the fold case
2129 * of LATIN SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of
2130 * two values "ss"); instead it asserts against that under DEBUGGING, and
2131 * otherwise returns its input. It does not handle the anomalies in UTF-8
2133 # define toFOLD_LC(c) \
2134 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
2135 ? GREEK_SMALL_LETTER_MU \
2136 : (__ASSERT_( ! IN_UTF8_CTYPE_LOCALE \
2137 || LIKELY((c) != LATIN_SMALL_LETTER_SHARP_S)) \
2141 #define isIDCONT(c) isWORDCHAR(c)
2142 #define isIDCONT_A(c) isWORDCHAR_A(c)
2143 #define isIDCONT_L1(c) isWORDCHAR_L1(c)
2144 #define isIDCONT_LC(c) isWORDCHAR_LC(c)
2145 #define isPSXSPC_LC(c) isSPACE_LC(c)
2147 /* For internal core Perl use only: the base macros for defining macros like
2148 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
2149 * number defined earlier in this file. generic_uvchr_() is used for POSIX
2150 * classes where there is a macro or function 'above_latin1' that takes the
2151 * single argument 'c' and returns the desired value. These exist for those
2152 * classes which have simple definitions, avoiding the overhead of an inversion
2153 * list binary search. generic_invlist_uvchr_() can be used
2154 * for classes where that overhead is faster than a direct lookup.
2155 * generic_uvchr_() won't compile if 'c' isn't unsigned, as it won't match the
2156 * 'above_latin1' prototype. generic_isCC_() macro does bounds checking, so
2157 * have duplicate checks here, so could create versions of the macros that
2158 * don't, but experiments show that gcc optimizes them out anyway. */
2160 /* Note that all ignore 'use bytes' */
2161 #define generic_uvchr_(classnum, above_latin1, c) ((c) < 256 \
2162 ? generic_isCC_(c, classnum) \
2164 #define generic_invlist_uvchr_(classnum, c) ((c) < 256 \
2165 ? generic_isCC_(c, classnum) \
2166 : _is_uni_FOO(classnum, c))
2167 #define isALPHA_uvchr(c) generic_invlist_uvchr_(CC_ALPHA_, c)
2168 #define isALPHANUMERIC_uvchr(c) generic_invlist_uvchr_(CC_ALPHANUMERIC_, c)
2169 #define isASCII_uvchr(c) isASCII(c)
2170 #define isBLANK_uvchr(c) generic_uvchr_(CC_BLANK_, is_HORIZWS_cp_high, c)
2171 #define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
2172 #define isDIGIT_uvchr(c) generic_invlist_uvchr_(CC_DIGIT_, c)
2173 #define isGRAPH_uvchr(c) generic_invlist_uvchr_(CC_GRAPH_, c)
2174 #define isIDCONT_uvchr(c) \
2175 generic_uvchr_(CC_WORDCHAR_, _is_uni_perl_idcont, c)
2176 #define isIDFIRST_uvchr(c) \
2177 generic_uvchr_(CC_IDFIRST_, _is_uni_perl_idstart, c)
2178 #define isLOWER_uvchr(c) generic_invlist_uvchr_(CC_LOWER_, c)
2179 #define isPRINT_uvchr(c) generic_invlist_uvchr_(CC_PRINT_, c)
2181 #define isPUNCT_uvchr(c) generic_invlist_uvchr_(CC_PUNCT_, c)
2182 #define isSPACE_uvchr(c) generic_uvchr_(CC_SPACE_, is_XPERLSPACE_cp_high, c)
2183 #define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
2185 #define isUPPER_uvchr(c) generic_invlist_uvchr_(CC_UPPER_, c)
2186 #define isVERTWS_uvchr(c) generic_uvchr_(CC_VERTSPACE_, is_VERTWS_cp_high, c)
2187 #define isWORDCHAR_uvchr(c) generic_invlist_uvchr_(CC_WORDCHAR_, c)
2188 #define isXDIGIT_uvchr(c) generic_uvchr_(CC_XDIGIT_, is_XDIGIT_cp_high, c)
2190 #define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
2191 #define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
2192 #define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
2193 #define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
2195 /* For backwards compatibility, even though '_uni' should mean official Unicode
2196 * code points, in Perl it means native for those below 256 */
2197 #define isALPHA_uni(c) isALPHA_uvchr(c)
2198 #define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
2199 #define isASCII_uni(c) isASCII_uvchr(c)
2200 #define isBLANK_uni(c) isBLANK_uvchr(c)
2201 #define isCNTRL_uni(c) isCNTRL_uvchr(c)
2202 #define isDIGIT_uni(c) isDIGIT_uvchr(c)
2203 #define isGRAPH_uni(c) isGRAPH_uvchr(c)
2204 #define isIDCONT_uni(c) isIDCONT_uvchr(c)
2205 #define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
2206 #define isLOWER_uni(c) isLOWER_uvchr(c)
2207 #define isPRINT_uni(c) isPRINT_uvchr(c)
2208 #define isPUNCT_uni(c) isPUNCT_uvchr(c)
2209 #define isSPACE_uni(c) isSPACE_uvchr(c)
2210 #define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
2211 #define isUPPER_uni(c) isUPPER_uvchr(c)
2212 #define isVERTWS_uni(c) isVERTWS_uvchr(c)
2213 #define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
2214 #define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
2215 #define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
2216 #define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
2217 #define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
2218 #define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
2220 /* For internal core Perl use only: the base macros for defining macros like
2221 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
2222 * point, not just 0-255. Like generic_uvchr_, there are two versions, one for
2223 * simple class definitions; the other for more complex. These are like
2224 * generic_uvchr_, so see it for more info. */
2225 #define generic_LC_uvchr_(latin1, above_latin1, c) \
2226 (c < 256 ? latin1(c) : above_latin1(c))
2227 #define generic_LC_invlist_uvchr_(latin1, classnum, c) \
2228 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
2230 #define isALPHA_LC_uvchr(c) generic_LC_invlist_uvchr_(isALPHA_LC, CC_ALPHA_, c)
2231 #define isALPHANUMERIC_LC_uvchr(c) generic_LC_invlist_uvchr_(isALPHANUMERIC_LC, \
2232 CC_ALPHANUMERIC_, c)
2233 #define isASCII_LC_uvchr(c) isASCII_LC(c)
2234 #define isBLANK_LC_uvchr(c) generic_LC_uvchr_(isBLANK_LC, \
2235 is_HORIZWS_cp_high, c)
2236 #define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
2237 #define isDIGIT_LC_uvchr(c) generic_LC_invlist_uvchr_(isDIGIT_LC, CC_DIGIT_, c)
2238 #define isGRAPH_LC_uvchr(c) generic_LC_invlist_uvchr_(isGRAPH_LC, CC_GRAPH_, c)
2239 #define isIDCONT_LC_uvchr(c) generic_LC_uvchr_(isIDCONT_LC, \
2240 _is_uni_perl_idcont, c)
2241 #define isIDFIRST_LC_uvchr(c) generic_LC_uvchr_(isIDFIRST_LC, \
2242 _is_uni_perl_idstart, c)
2243 #define isLOWER_LC_uvchr(c) generic_LC_invlist_uvchr_(isLOWER_LC, CC_LOWER_, c)
2244 #define isPRINT_LC_uvchr(c) generic_LC_invlist_uvchr_(isPRINT_LC, CC_PRINT_, c)
2245 #define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
2246 #define isPUNCT_LC_uvchr(c) generic_LC_invlist_uvchr_(isPUNCT_LC, CC_PUNCT_, c)
2247 #define isSPACE_LC_uvchr(c) generic_LC_uvchr_(isSPACE_LC, \
2248 is_XPERLSPACE_cp_high, c)
2249 #define isUPPER_LC_uvchr(c) generic_LC_invlist_uvchr_(isUPPER_LC, CC_UPPER_, c)
2250 #define isWORDCHAR_LC_uvchr(c) generic_LC_invlist_uvchr_(isWORDCHAR_LC, \
2252 #define isXDIGIT_LC_uvchr(c) generic_LC_uvchr_(isXDIGIT_LC, \
2253 is_XDIGIT_cp_high, c)
2255 #define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
2257 /* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
2258 * they don't otherwise go out of their way to look for malformed UTF-8. If
2259 * they can return accurate results without knowing if the input is otherwise
2260 * malformed, they do so. For example isASCII is accurate in spite of any
2261 * non-length malformations because it looks only at a single byte. Likewise
2262 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
2263 * variant ones return FALSE. But, if the input has to be well-formed in order
2264 * for the results to be accurate, the macros will test and if malformed will
2265 * call a routine to die
2267 * Except for toke.c, the macros do assume that e > p, asserting that on
2268 * DEBUGGING builds. Much code that calls these depends on this being true,
2269 * for other reasons. toke.c is treated specially as using the regular
2270 * assertion breaks it in many ways. All strings that these operate on there
2271 * are supposed to have an extra NUL character at the end, so that *e = \0. A
2272 * bunch of code in toke.c assumes that this is true, so the assertion allows
2274 #ifdef PERL_IN_TOKE_C
2275 # define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
2277 # define _utf8_safe_assert(p,e) ((e) > (p))
2280 #define generic_utf8_safe_(classnum, p, e, above_latin1) \
2281 ((! _utf8_safe_assert(p, e)) \
2282 ? (_force_out_malformed_utf8_message((U8 *) (p), (U8 *) (e), 0, 1), 0)\
2283 : (UTF8_IS_INVARIANT(*(p))) \
2284 ? generic_isCC_(*(p), classnum) \
2285 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2286 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2287 ? generic_isCC_(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
2289 : (_force_out_malformed_utf8_message( \
2290 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2292 /* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
2293 * 'above_latin1' can be a macro */
2294 #define generic_func_utf8_safe_(classnum, above_latin1, p, e) \
2295 generic_utf8_safe_(classnum, p, e, above_latin1(p, e))
2296 #define generic_non_invlist_utf8_safe_(classnum, above_latin1, p, e) \
2297 generic_utf8_safe_(classnum, p, e, \
2298 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2299 ? (_force_out_malformed_utf8_message( \
2300 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2302 /* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
2303 * 'above_latin1' parameter */
2304 #define generic_invlist_utf8_safe_(classnum, p, e) \
2305 generic_utf8_safe_(classnum, p, e, _is_utf8_FOO(classnum, p, e))
2307 /* Like the above, but should be used only when it is known that there are no
2308 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
2309 * class is TRUE for. Hence it can skip the tests for this range.
2310 * 'above_latin1' should include its arguments */
2311 #define generic_utf8_safe_no_upper_latin1_(classnum, p, e, above_latin1) \
2312 (__ASSERT_(_utf8_safe_assert(p, e)) \
2314 ? generic_isCC_(*(p), classnum) \
2315 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
2316 ? 0 /* Note that doesn't check validity for latin1 */ \
2320 #define isALPHA_utf8(p, e) isALPHA_utf8_safe(p, e)
2321 #define isALPHANUMERIC_utf8(p, e) isALPHANUMERIC_utf8_safe(p, e)
2322 #define isASCII_utf8(p, e) isASCII_utf8_safe(p, e)
2323 #define isBLANK_utf8(p, e) isBLANK_utf8_safe(p, e)
2324 #define isCNTRL_utf8(p, e) isCNTRL_utf8_safe(p, e)
2325 #define isDIGIT_utf8(p, e) isDIGIT_utf8_safe(p, e)
2326 #define isGRAPH_utf8(p, e) isGRAPH_utf8_safe(p, e)
2327 #define isIDCONT_utf8(p, e) isIDCONT_utf8_safe(p, e)
2328 #define isIDFIRST_utf8(p, e) isIDFIRST_utf8_safe(p, e)
2329 #define isLOWER_utf8(p, e) isLOWER_utf8_safe(p, e)
2330 #define isPRINT_utf8(p, e) isPRINT_utf8_safe(p, e)
2331 #define isPSXSPC_utf8(p, e) isPSXSPC_utf8_safe(p, e)
2332 #define isPUNCT_utf8(p, e) isPUNCT_utf8_safe(p, e)
2333 #define isSPACE_utf8(p, e) isSPACE_utf8_safe(p, e)
2334 #define isUPPER_utf8(p, e) isUPPER_utf8_safe(p, e)
2335 #define isVERTWS_utf8(p, e) isVERTWS_utf8_safe(p, e)
2336 #define isWORDCHAR_utf8(p, e) isWORDCHAR_utf8_safe(p, e)
2337 #define isXDIGIT_utf8(p, e) isXDIGIT_utf8_safe(p, e)
2339 #define isALPHA_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_ALPHA_, p, e)
2340 #define isALPHANUMERIC_utf8_safe(p, e) \
2341 generic_invlist_utf8_safe_(CC_ALPHANUMERIC_, p, e)
2342 #define isASCII_utf8_safe(p, e) \
2343 /* Because ASCII is invariant under utf8, the non-utf8 macro \
2345 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
2346 #define isBLANK_utf8_safe(p, e) \
2347 generic_non_invlist_utf8_safe_(CC_BLANK_, is_HORIZWS_high, p, e)
2350 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
2351 * more efficient macro instead of the more general one */
2352 # define isCNTRL_utf8_safe(p, e) \
2353 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
2355 # define isCNTRL_utf8_safe(p, e) generic_utf8_safe_(CC_CNTRL_, p, e, 0)
2358 #define isDIGIT_utf8_safe(p, e) \
2359 generic_utf8_safe_no_upper_latin1_(CC_DIGIT_, p, e, \
2360 _is_utf8_FOO(CC_DIGIT_, p, e))
2361 #define isGRAPH_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_GRAPH_, p, e)
2362 #define isIDCONT_utf8_safe(p, e) generic_func_utf8_safe_(CC_WORDCHAR_, \
2363 _is_utf8_perl_idcont, p, e)
2365 /* To prevent S_scan_word in toke.c from hanging, we have to make sure that
2366 * IDFIRST is an alnum. See
2367 * https://github.com/Perl/perl5/issues/10275 for more detail than you
2368 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
2369 * This used to be not the XID version, but we decided to go with the more
2370 * modern Unicode definition */
2371 #define isIDFIRST_utf8_safe(p, e) \
2372 generic_func_utf8_safe_(CC_IDFIRST_, \
2373 _is_utf8_perl_idstart, (U8 *) (p), (U8 *) (e))
2375 #define isLOWER_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_LOWER_, p, e)
2376 #define isPRINT_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_PRINT_, p, e)
2377 #define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
2378 #define isPUNCT_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_PUNCT_, p, e)
2379 #define isSPACE_utf8_safe(p, e) \
2380 generic_non_invlist_utf8_safe_(CC_SPACE_, is_XPERLSPACE_high, p, e)
2381 #define isUPPER_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_UPPER_, p, e)
2382 #define isVERTWS_utf8_safe(p, e) \
2383 generic_non_invlist_utf8_safe_(CC_VERTSPACE_, is_VERTWS_high, p, e)
2384 #define isWORDCHAR_utf8_safe(p, e) \
2385 generic_invlist_utf8_safe_(CC_WORDCHAR_, p, e)
2386 #define isXDIGIT_utf8_safe(p, e) \
2387 generic_utf8_safe_no_upper_latin1_(CC_XDIGIT_, p, e, \
2388 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2389 ? (_force_out_malformed_utf8_message( \
2390 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2391 : is_XDIGIT_high(p)))
2393 #define toFOLD_utf8(p,e,s,l) toFOLD_utf8_safe(p,e,s,l)
2394 #define toLOWER_utf8(p,e,s,l) toLOWER_utf8_safe(p,e,s,l)
2395 #define toTITLE_utf8(p,e,s,l) toTITLE_utf8_safe(p,e,s,l)
2396 #define toUPPER_utf8(p,e,s,l) toUPPER_utf8_safe(p,e,s,l)
2398 /* For internal core use only, subject to change */
2399 #define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f)
2400 #define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f)
2401 #define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f)
2402 #define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f)
2404 #define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
2405 #define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
2406 #define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
2407 #define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
2409 #define isALPHA_LC_utf8(p, e) isALPHA_LC_utf8_safe(p, e)
2410 #define isALPHANUMERIC_LC_utf8(p, e) isALPHANUMERIC_LC_utf8_safe(p, e)
2411 #define isASCII_LC_utf8(p, e) isASCII_LC_utf8_safe(p, e)
2412 #define isBLANK_LC_utf8(p, e) isBLANK_LC_utf8_safe(p, e)
2413 #define isCNTRL_LC_utf8(p, e) isCNTRL_LC_utf8_safe(p, e)
2414 #define isDIGIT_LC_utf8(p, e) isDIGIT_LC_utf8_safe(p, e)
2415 #define isGRAPH_LC_utf8(p, e) isGRAPH_LC_utf8_safe(p, e)
2416 #define isIDCONT_LC_utf8(p, e) isIDCONT_LC_utf8_safe(p, e)
2417 #define isIDFIRST_LC_utf8(p, e) isIDFIRST_LC_utf8_safe(p, e)
2418 #define isLOWER_LC_utf8(p, e) isLOWER_LC_utf8_safe(p, e)
2419 #define isPRINT_LC_utf8(p, e) isPRINT_LC_utf8_safe(p, e)
2420 #define isPSXSPC_LC_utf8(p, e) isPSXSPC_LC_utf8_safe(p, e)
2421 #define isPUNCT_LC_utf8(p, e) isPUNCT_LC_utf8_safe(p, e)
2422 #define isSPACE_LC_utf8(p, e) isSPACE_LC_utf8_safe(p, e)
2423 #define isUPPER_LC_utf8(p, e) isUPPER_LC_utf8_safe(p, e)
2424 #define isWORDCHAR_LC_utf8(p, e) isWORDCHAR_LC_utf8_safe(p, e)
2425 #define isXDIGIT_LC_utf8(p, e) isXDIGIT_LC_utf8_safe(p, e)
2427 /* For internal core Perl use only: the base macros for defining macros like
2428 * isALPHA_LC_utf8_safe. These are like generic_utf8_, but if the first code
2429 * point in 'p' is within the 0-255 range, it uses locale rules from the
2430 * passed-in 'macro' parameter */
2431 #define generic_LC_utf8_safe_(macro, p, e, above_latin1) \
2432 (__ASSERT_(_utf8_safe_assert(p, e)) \
2433 (UTF8_IS_INVARIANT(*(p))) \
2435 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2436 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2437 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2438 : (_force_out_malformed_utf8_message( \
2439 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2442 #define generic_LC_invlist_utf8_safe_(macro, classnum, p, e) \
2443 generic_LC_utf8_safe_(macro, p, e, \
2444 _is_utf8_FOO(classnum, p, e))
2446 #define generic_LC_func_utf8_safe_(macro, above_latin1, p, e) \
2447 generic_LC_utf8_safe_(macro, p, e, above_latin1(p, e))
2449 #define generic_LC_non_invlist_utf8_safe_(classnum, above_latin1, p, e) \
2450 generic_LC_utf8_safe_(classnum, p, e, \
2451 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2452 ? (_force_out_malformed_utf8_message( \
2453 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2456 #define isALPHANUMERIC_LC_utf8_safe(p, e) \
2457 generic_LC_invlist_utf8_safe_(isALPHANUMERIC_LC, \
2458 CC_ALPHANUMERIC_, p, e)
2459 #define isALPHA_LC_utf8_safe(p, e) \
2460 generic_LC_invlist_utf8_safe_(isALPHA_LC, CC_ALPHA_, p, e)
2461 #define isASCII_LC_utf8_safe(p, e) \
2462 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2463 #define isBLANK_LC_utf8_safe(p, e) \
2464 generic_LC_non_invlist_utf8_safe_(isBLANK_LC, is_HORIZWS_high, p, e)
2465 #define isCNTRL_LC_utf8_safe(p, e) \
2466 generic_LC_utf8_safe_(isCNTRL_LC, p, e, 0)
2467 #define isDIGIT_LC_utf8_safe(p, e) \
2468 generic_LC_invlist_utf8_safe_(isDIGIT_LC, CC_DIGIT_, p, e)
2469 #define isGRAPH_LC_utf8_safe(p, e) \
2470 generic_LC_invlist_utf8_safe_(isGRAPH_LC, CC_GRAPH_, p, e)
2471 #define isIDCONT_LC_utf8_safe(p, e) \
2472 generic_LC_func_utf8_safe_(isIDCONT_LC, \
2473 _is_utf8_perl_idcont, p, e)
2474 #define isIDFIRST_LC_utf8_safe(p, e) \
2475 generic_LC_func_utf8_safe_(isIDFIRST_LC, \
2476 _is_utf8_perl_idstart, p, e)
2477 #define isLOWER_LC_utf8_safe(p, e) \
2478 generic_LC_invlist_utf8_safe_(isLOWER_LC, CC_LOWER_, p, e)
2479 #define isPRINT_LC_utf8_safe(p, e) \
2480 generic_LC_invlist_utf8_safe_(isPRINT_LC, CC_PRINT_, p, e)
2481 #define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2482 #define isPUNCT_LC_utf8_safe(p, e) \
2483 generic_LC_invlist_utf8_safe_(isPUNCT_LC, CC_PUNCT_, p, e)
2484 #define isSPACE_LC_utf8_safe(p, e) \
2485 generic_LC_non_invlist_utf8_safe_(isSPACE_LC, is_XPERLSPACE_high, p, e)
2486 #define isUPPER_LC_utf8_safe(p, e) \
2487 generic_LC_invlist_utf8_safe_(isUPPER_LC, CC_UPPER_, p, e)
2488 #define isWORDCHAR_LC_utf8_safe(p, e) \
2489 generic_LC_invlist_utf8_safe_(isWORDCHAR_LC, CC_WORDCHAR_, p, e)
2490 #define isXDIGIT_LC_utf8_safe(p, e) \
2491 generic_LC_non_invlist_utf8_safe_(isXDIGIT_LC, is_XDIGIT_high, p, e)
2493 /* Macros for backwards compatibility and for completeness when the ASCII and
2494 * Latin1 values are identical */
2495 #define isALPHAU(c) isALPHA_L1(c)
2496 #define isDIGIT_L1(c) isDIGIT_A(c)
2497 #define isOCTAL(c) isOCTAL_A(c)
2498 #define isOCTAL_L1(c) isOCTAL_A(c)
2499 #define isXDIGIT_L1(c) isXDIGIT_A(c)
2500 #define isALNUM(c) isWORDCHAR(c)
2501 #define isALNUM_A(c) isALNUM(c)
2502 #define isALNUMU(c) isWORDCHAR_L1(c)
2503 #define isALNUM_LC(c) isWORDCHAR_LC(c)
2504 #define isALNUM_uni(c) isWORDCHAR_uni(c)
2505 #define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
2506 #define isALNUM_utf8(p,e) isWORDCHAR_utf8(p,e)
2507 #define isALNUM_utf8_safe(p,e) isWORDCHAR_utf8_safe(p,e)
2508 #define isALNUM_LC_utf8(p,e)isWORDCHAR_LC_utf8(p,e)
2509 #define isALNUM_LC_utf8_safe(p,e)isWORDCHAR_LC_utf8_safe(p,e)
2510 #define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2511 #define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2512 #define isALNUMC(c) isALPHANUMERIC(c)
2513 #define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2514 #define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
2515 #define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
2516 #define isALNUMC_utf8(p,e) isALPHANUMERIC_utf8(p,e)
2517 #define isALNUMC_utf8_safe(p,e) isALPHANUMERIC_utf8_safe(p,e)
2518 #define isALNUMC_LC_utf8_safe(p,e) isALPHANUMERIC_LC_utf8_safe(p,e)
2520 /* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2521 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2522 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2523 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2524 * the outlier from the block that contains the other controls, just like
2525 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2526 * block. If it weren't special cased, it would yield a non-control.
2527 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2530 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2532 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2534 ? (UNLIKELY((c) == '?') \
2535 ? QUESTION_MARK_CTRL \
2536 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2537 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2539 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2543 =for apidoc Ay||line_t
2544 The typedef to use to declare variables that are to hold line numbers.
2548 Line numbers are unsigned, 32 bits.
2551 #define LINE_Tf U32uf
2552 #define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
2554 /* Helpful alias for version prescan */
2555 #define is_LAX_VERSION(a,b) \
2556 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2558 #define is_STRICT_VERSION(a,b) \
2559 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2561 #define BADVERSION(a,b,c) \
2567 /* Converts a character KNOWN to represent a hexadecimal digit (0-9, A-F, or
2568 * a-f) to its numeric value without using any branches. The input is
2569 * validated only by an assert() in DEBUGGING builds.
2571 * It works by right shifting and isolating the bit that is 0 for the digits,
2572 * and 1 for at least the alphas A-F, a-f. The bit is shifted to the ones
2573 * position, and then to the eights position. Both are added together to form
2574 * 0 if the input is '0'-'9' and to form 9 if alpha. This is added to the
2575 * final four bits of the input to form the correct value. */
2576 #define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) \
2577 ((NATIVE_TO_LATIN1(c) >> 6) & 1) /* 1 if alpha; 0 if not */ \
2578 + ((NATIVE_TO_LATIN1(c) >> 3) & 8) /* 8 if alpha; 0 if not */ \
2579 + ((c) & 0xF)) /* 0-9 if input valid hex digit */
2581 /* The argument is a string pointer, which is advanced. */
2582 #define READ_XDIGIT(s) ((s)++, XDIGIT_VALUE(*((s) - 1)))
2584 /* Converts a character known to represent an octal digit (0-7) to its numeric
2585 * value. The input is validated only by an assert() in DEBUGGING builds. In
2586 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2587 #define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2589 /* Efficiently returns a boolean as to if two native characters are equivalent
2590 * case-insensitively. At least one of the characters must be one of [A-Za-z];
2591 * the ALPHA in the name is to remind you of that. This is asserted() in
2592 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2593 * works (on valid input) for both non- and UTF-8-encoded bytes.
2595 * When one of the inputs is a compile-time constant and gets folded by the
2596 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2597 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2598 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
2599 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2600 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2601 * just a single 0, in the bit position where the upper- and lowercase differ.
2603 #define isALPHA_FOLD_EQ(c1, c2) \
2604 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2605 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2606 #define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2609 =for apidoc_section $memory
2611 =for apidoc Am|void|Newx|void* ptr|int nitems|type
2612 =for apidoc_item |void*|safemalloc|size_t size
2614 The XSUB-writer's interface to the C C<malloc> function.
2616 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2618 In 5.9.3, Newx() and friends replace the older New() API, and drops
2619 the first parameter, I<x>, a debug aid which allowed callers to identify
2620 themselves. This aid has been superseded by a new build option,
2621 PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
2622 there for use in XS modules supporting older perls.
2624 =for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
2625 The XSUB-writer's interface to the C C<malloc> function, with
2626 cast. See also C<L</Newx>>.
2628 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2630 =for apidoc Am|void|Newxz|void* ptr|int nitems|type
2631 =for apidoc_item |void*|safecalloc|size_t nitems|size_t item_size
2633 The XSUB-writer's interface to the C C<malloc> function. The allocated
2634 memory is zeroed with C<memzero>. See also C<L</Newx>>.
2636 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2638 =for apidoc Am|void|Renew|void* ptr|int nitems|type
2639 =for apidoc_item |void*|saferealloc|void *ptr|size_t size
2641 The XSUB-writer's interface to the C C<realloc> function.
2643 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2645 =for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2646 The XSUB-writer's interface to the C C<realloc> function, with
2649 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2651 =for apidoc Am|void|Safefree|void* ptr
2652 The XSUB-writer's interface to the C C<free> function.
2654 This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
2656 =for apidoc_section $string
2657 =for apidoc Am|void |Move |void* src|void* dest|int nitems|type
2658 =for apidoc_item |void *|MoveD|void* src|void* dest|int nitems|type
2659 The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
2660 source, C<dest> is the destination, C<nitems> is the number of items, and
2661 C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
2663 C<MoveD> is like C<Move> but returns C<dest>. Useful
2664 for encouraging compilers to tail-call
2667 =for apidoc Am|void |Copy |void* src|void* dest|int nitems|type
2668 =for apidoc_item |void *|CopyD|void* src|void* dest|int nitems|type
2669 The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
2670 source, C<dest> is the destination, C<nitems> is the number of items, and
2671 C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
2673 C<CopyD> is like C<Copy> but returns C<dest>. Useful
2674 for encouraging compilers to tail-call
2677 =for apidoc Am|void |NewCopy |void* src|void* dest|int nitems|type
2678 Combines Newx() and Copy() into a single macro. Dest will be allocated
2679 using Newx() and then src will be copied into it.
2681 =for apidoc Am|void |Zero |void* dest|int nitems|type
2682 =for apidoc_item |void *|ZeroD|void* dest|int nitems|type
2684 The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2685 destination, C<nitems> is the number of items, and C<type> is the type.
2687 C<ZeroD> is like C<Zero> but returns C<dest>. Useful
2688 for encouraging compilers to tail-call
2691 =for apidoc_section $utility
2692 =for apidoc Amu|void|StructCopy|type *src|type *dest|type
2693 This is an architecture-independent macro to copy one structure to another.
2695 =for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2697 Fill up memory with a byte pattern (a byte repeated over and over
2698 again) that hopefully catches attempts to access uninitialized memory.
2700 =for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2702 PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2704 =for apidoc Am|void|PoisonFree|void* dest|int nitems|type
2706 PoisonWith(0xEF) for catching access to freed memory.
2708 =for apidoc Am|void|Poison|void* dest|int nitems|type
2710 PoisonWith(0xEF) for catching access to freed memory.
2714 /* Maintained for backwards-compatibility only. Use newSV() instead. */
2716 #define NEWSV(x,len) newSV(len)
2719 #define MEM_SIZE_MAX ((MEM_SIZE)-1)
2721 #define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
2723 #ifdef PERL_MALLOC_WRAP
2725 /* This expression will be constant-folded at compile time. It checks
2726 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2727 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2728 * As well as avoiding the need for a run-time check in some cases, it's
2729 * designed to avoid compiler warnings like:
2730 * comparison is always false due to limited range of data type
2731 * It's mathematically equivalent to
2732 * max(n) * sizeof(t) > MEM_SIZE_MAX
2735 # define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
2736 ( sizeof(MEM_SIZE) < sizeof(n) \
2737 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
2739 /* This is written in a slightly odd way to avoid various spurious
2740 * compiler warnings. We *want* to write the expression as
2741 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2742 * (for some compile-time constant C), but even when the LHS
2743 * constant-folds to false at compile-time, g++ insists on emitting
2744 * warnings about the RHS (e.g. "comparison is always false"), so instead
2747 * (cond ? n : X) > C
2749 * where X is a constant with X > C always false. Choosing a value for X
2750 * is tricky. If 0, some compilers will complain about 0 > C always being
2751 * false; if 1, Coverity complains when n happens to be the constant value
2752 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2753 * for X and hope that nothing else whines.
2756 # define _MEM_WRAP_WILL_WRAP(n,t) \
2757 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2758 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
2760 # define MEM_WRAP_CHECK(n,t) \
2761 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2762 && (croak_memory_wrap(),0))
2764 # define MEM_WRAP_CHECK_1(n,t,a) \
2765 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2766 && (Perl_croak_nocontext("%s",(a)),0))
2768 /* "a" arg must be a string literal */
2769 # define MEM_WRAP_CHECK_s(n,t,a) \
2770 ( (void) (UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2771 && (Perl_croak_nocontext(ASSERT_IS_LITERAL(a)), 0)))
2773 # define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
2775 # define PERL_STRLEN_ROUNDUP(n) ((void)(((n) > MEM_SIZE_MAX - 2 * PERL_STRLEN_ROUNDUP_QUANTUM) ? (croak_memory_wrap(),0) : 0), _PERL_STRLEN_ROUNDUP_UNCHECKED(n))
2778 # define MEM_WRAP_CHECK(n,t)
2779 # define MEM_WRAP_CHECK_1(n,t,a)
2780 # define MEM_WRAP_CHECK_s(n,t,a)
2781 # define MEM_WRAP_CHECK_(n,t)
2783 # define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
2789 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
2790 * go through functions, which are handy for debugging breakpoints, but
2791 * which more importantly get the immediate calling environment (file and
2792 * line number, and C function name if available) passed in. This info can
2793 * then be used for logging the calls, for which one gets a sample
2794 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
2797 * - not all memory allocs get logged, only those
2798 * that go through Newx() and derivatives (while all
2799 * Safefrees do get logged)
2800 * - __FILE__ and __LINE__ do not work everywhere
2801 * - __func__ or __FUNCTION__ even less so
2802 * - I think more goes on after the perlio frees but
2803 * the thing is that STDERR gets closed (as do all
2804 * the file descriptors)
2805 * - no deeper calling stack than the caller of the Newx()
2806 * or the kind, but do I look like a C reflection/introspection
2808 * - the function prototypes for the logging functions
2809 * probably should maybe be somewhere else than handy.h
2810 * - one could consider inlining (macrofying) the logging
2811 * for speed, but I am too lazy
2812 * - one could imagine recording the allocations in a hash,
2813 * (keyed by the allocation address?), and maintain that
2814 * through reallocs and frees, but how to do that without
2815 * any News() happening...?
2816 * - lots of -Ddefines to get useful/controllable output
2817 * - lots of ENV reads
2821 # ifndef PERL_MEM_LOG_NOIMPL
2835 #define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2836 #define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
2837 #define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
2840 #ifndef MEM_LOG_ALLOC
2841 #define MEM_LOG_ALLOC(n,t,a) (a)
2843 #ifndef MEM_LOG_REALLOC
2844 #define MEM_LOG_REALLOC(n,t,v,a) (a)
2846 #ifndef MEM_LOG_FREE
2847 #define MEM_LOG_FREE(a) (a)
2850 #define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2851 #define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2852 #define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
2855 /* pre 5.9.x compatibility */
2856 #define New(x,v,n,t) Newx(v,n,t)
2857 #define Newc(x,v,n,t,c) Newxc(v,n,t,c)
2858 #define Newz(x,v,n,t) Newxz(v,n,t)
2861 #define Renew(v,n,t) \
2862 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2863 #define Renewc(v,n,t,c) \
2864 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2867 #define Safefree(d) \
2868 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
2870 #define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
2873 /* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2874 * as it handles cases like constant string arguments without throwing warnings *
2875 * the cast is required, as is the inequality check, to avoid warnings */
2876 #define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
2879 #define Move(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2880 #define Copy(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2881 #define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2883 /* Like above, but returns a pointer to 'd' */
2884 #define MoveD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2885 #define CopyD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2886 #define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
2888 #define NewCopy(s,d,n,t) STMT_START { \
2893 #define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2894 #define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2895 #define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2896 #define Poison(d,n,t) PoisonFree(d,n,t)
2899 # define PERL_POISON_EXPR(x) x
2901 # define PERL_POISON_EXPR(x)
2905 #define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2908 =for apidoc_section $utility
2910 =for apidoc Am|STRLEN|C_ARRAY_LENGTH|void *a
2912 Returns the number of elements in the input C array (so you want your
2913 zero-based indices to be less than but not equal to).
2915 =for apidoc Am|void *|C_ARRAY_END|void *a
2917 Returns a pointer to one element past the final element of the input C array.
2921 C_ARRAY_END is one past the last: half-open/half-closed range, not
2922 last-inclusive range.
2924 #define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
2925 #define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
2927 #if defined(PERL_CORE) || defined(PERL_EXT_RE_BUILD)
2928 /* strlen() of a literal string constant. Restricting this to core, in part
2929 * because it can generate compiler warnings about comparing unlike signs */
2930 # define STRLENs(s) (sizeof("" s "") - 1)
2935 # define Perl_va_copy(s, d) va_copy(d, s)
2936 # elif defined(__va_copy)
2937 # define Perl_va_copy(s, d) __va_copy(d, s)
2939 # define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2943 /* convenience debug macros */
2945 #define pTHX_FORMAT "Perl interpreter: 0x%p"
2946 #define pTHX__FORMAT ", Perl interpreter: 0x%p"
2947 #define pTHX_VALUE_ (void *)my_perl,
2948 #define pTHX_VALUE (void *)my_perl
2949 #define pTHX__VALUE_ ,(void *)my_perl,
2950 #define pTHX__VALUE ,(void *)my_perl
2953 #define pTHX__FORMAT
2956 #define pTHX__VALUE_
2958 #endif /* USE_ITHREADS */
2960 /* Perl_deprecate was not part of the public API, and did not have a deprecate()
2961 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2962 CPAN::Unpack show any users outside the core. */
2964 # define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2965 "Use of " s " is deprecated")
2966 # define deprecate_disappears_in(when,message) \
2967 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2968 message " is deprecated, and will disappear in Perl " when)
2969 # define deprecate_fatal_in(when,message) \
2970 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2971 message " is deprecated, and will become fatal in Perl " when)
2974 /* Internal macros to deal with gids and uids */
2977 # if Uid_t_size > IVSIZE
2978 # define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2979 # define SvUID(sv) SvNV(sv)
2980 # elif Uid_t_sign <= 0
2981 # define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2982 # define SvUID(sv) SvIV(sv)
2984 # define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2985 # define SvUID(sv) SvUV(sv)
2986 # endif /* Uid_t_size */
2988 # if Gid_t_size > IVSIZE
2989 # define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2990 # define SvGID(sv) SvNV(sv)
2991 # elif Gid_t_sign <= 0
2992 # define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2993 # define SvGID(sv) SvIV(sv)
2995 # define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2996 # define SvGID(sv) SvUV(sv)
2997 # endif /* Gid_t_size */
3001 /* These are simple Marsaglia XOR-SHIFT RNG's for 64 and 32 bits. These
3002 * RNG's are of reasonable quality, very fast, and have the interesting
3003 * property that provided 'x' is non-zero they create a cycle of 2^32-1
3004 * or 2^64-1 "random" like numbers, with the exception of 0. Thus they
3005 * are very useful when you want an integer to "dance" in a random way,
3006 * but you also never want it to become 0 and thus false.
3008 * Obviously they leave x unchanged if it starts out as 0.
3010 * We have two variants just because that can be helpful in certain
3011 * places. There is no advantage to either, they are equally bad as each
3012 * other as far RNG's go. Sufficiently random for many purposes, but
3013 * insufficiently random for serious use as they fail important tests in
3014 * the Test01 BigCrush RNG test suite by L’Ecuyer and Simard. (Note
3015 * that Drand48 also fails BigCrush). The main point is they produce
3016 * different sequences and in places where we want some randomlike
3017 * behavior they are cheap and easy.
3019 * Marsaglia was one of the early researchers into RNG testing and wrote
3020 * the Diehard RNG test suite, which after his death become the
3021 * Dieharder RNG suite, and was generally supplanted by the Test01 suite
3022 * by L'Ecruyer and associates.
3024 * There are dozens of shift parameters that create a pseudo random ring
3025 * of integers 1..2^N-1, if you need a different sequence just read the
3026 * paper and select a set of parameters. In fact, simply reversing the
3027 * shift order from L/R/L to R/L/R should result in another valid
3028 * example, but read the paper before you do that.
3030 * PDF of the original paper:
3031 * https://www.jstatsoft.org/article/download/v008i14/916
3033 * https://en.wikipedia.org/wiki/Xorshift
3035 * https://www.iro.umontreal.ca/~lecuyer/myftp/papers/xorshift.pdf
3037 * http://simul.iro.umontreal.ca/testu01/tu01.html
3039 * https://en.wikipedia.org/wiki/Diehard_tests
3041 * https://webhome.phy.duke.edu/~rgb/General/rand_rate/rand_rate.abs
3045 /* 32 bit version */
3046 #define PERL_XORSHIFT32_A(x) \
3048 (x) ^= ((x) << 13); \
3049 (x) ^= ((x) >> 17); \
3050 (x) ^= ((x) << 5); \
3053 /* 64 bit version */
3054 #define PERL_XORSHIFT64_A(x) \
3056 (x) ^= ((x) << 13); \
3057 (x) ^= ((x) >> 7); \
3058 (x) ^= ((x) << 17); \
3061 /* 32 bit version */
3062 #define PERL_XORSHIFT32_B(x) \
3064 (x) ^= ((x) << 5); \
3065 (x) ^= ((x) >> 27); \
3066 (x) ^= ((x) << 8); \
3069 /* 64 bit version - currently this is unused,
3070 * it is provided here to complement the 32 bit _B
3071 * variant which IS used. */
3072 #define PERL_XORSHIFT64_B(x) \
3074 (x) ^= ((x) << 15); \
3075 (x) ^= ((x) >> 49); \
3076 (x) ^= ((x) << 26); \
3080 #endif /* PERL_HANDY_H_ */
3083 * ex: set ts=8 sts=4 sw=4 et:
https://perl5.git.perl.org / perl5.git / blob