This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
autodoc.pl: Add h flag for hidden documentation
[perl5.git] / handy.h
CommitLineData
a0d0e21e 1/* handy.h
a687059c 2 *
1129b882 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
da5d8dbb 4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
a687059c 5 *
6e21c824
LW
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.
8d063cd8 8 *
8d063cd8
LW
9 */
10
4650c663
KW
11/* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
13
6a5bc5ac
KW
14#ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
15#define PERL_HANDY_H_
9d745869 16
24792b8d
NC
17#ifndef PERL_CORE
18# define Null(type) ((type)NULL)
954c1994
GS
19
20/*
ccfc67b7 21=head1 Handy Values
954c1994 22
78342678 23=for apidoc AmnU||Nullch
72d33970
FC
24Null character pointer. (No longer available when C<PERL_CORE> is
25defined.)
2307c6d0 26
78342678 27=for apidoc AmnU||Nullsv
72d33970 28Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
954c1994
GS
29
30=cut
31*/
32
24792b8d
NC
33# define Nullch Null(char*)
34# define Nullfp Null(PerlIO*)
35# define Nullsv Null(SV*)
36#endif
8d063cd8 37
641d3f0b
PP
38#ifdef TRUE
39#undef TRUE
40#endif
41#ifdef FALSE
42#undef FALSE
43#endif
44#define TRUE (1)
45#define FALSE (0)
46
cf3f0ffb
DM
47/* The MUTABLE_*() macros cast pointers to the types shown, in such a way
48 * (compiler permitting) that casting away const-ness will give a warning;
49 * e.g.:
50 *
51 * const SV *sv = ...;
52 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
53 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
54 */
55
b1bc3f34 56#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
f065df04 57# define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
b1bc3f34
NC
58#else
59# define MUTABLE_PTR(p) ((void *) (p))
60#endif
61
a062e10d 62#define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
ea726b52 63#define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
159b6efe 64#define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
dbebbdb4 65#define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
a45c7426 66#define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
b1bc3f34 67#define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
27d4fb96 68
f789f6a4 69#if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
bd31be4b
NC
70# include <stdbool.h>
71# ifndef HAS_BOOL
72# define HAS_BOOL 1
73# endif
74#endif
75
8e84507e 76/* bool is built-in for g++-2.6.3 and later, which might be used
c1d22f6b
GS
77 for extensions. <_G_config.h> defines _G_HAVE_BOOL, but we can't
78 be sure _G_config.h will be included before this file. _G_config.h
8e84507e 79 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
c1d22f6b
GS
80 actually has bool. Hence, _G_HAVE_BOOL is pretty useless for us.
81 g++ can be identified by __GNUG__.
82 Andy Dougherty February 2000
5d94fbed 83*/
3609ea0d 84#ifdef __GNUG__ /* GNU g++ has bool built-in */
f789f6a4 85# ifndef PERL_BOOL_AS_CHAR
5d94fbed 86# ifndef HAS_BOOL
c1d22f6b 87# define HAS_BOOL 1
5d94fbed 88# endif
f789f6a4 89# endif
5d94fbed
AD
90#endif
91
92#ifndef HAS_BOOL
f789f6a4
FC
93# ifdef bool
94# undef bool
95# endif
70d5cb32 96# define bool char
c1d22f6b 97# define HAS_BOOL 1
a687059c 98#endif
0d3e774c 99
25ba28ce
KW
100/*
101=for apidoc Am|bool|cBOOL|bool expr
102
103Cast-to-bool. A simple S<C<(bool) I<expr>>> cast may not do the right thing:
104if C<bool> is defined as C<char>, for example, then the cast from C<int> is
105implementation-defined.
106
107C<(bool)!!(cbool)> in a ternary triggers a bug in xlc on AIX
108
109=cut
110*/
18f5643b 111#define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
f2338a2e 112
46c6c7e2 113/* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
e352bcff
JH
114 * XXX Should really be a Configure probe, with HAS__FUNCTION__
115 * and FUNCTION__ as results.
116 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
46c6c7e2
JH
117#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
118# define FUNCTION__ __func__
07798b17
AC
119#elif (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
120 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
121# define FUNCTION__ ""
46c6c7e2 122#else
07798b17 123# define FUNCTION__ __FUNCTION__ /* Common extension. */
46c6c7e2
JH
124#endif
125
27d4fb96
PP
126/* XXX A note on the perl source internal type system. The
127 original intent was that I32 be *exactly* 32 bits.
128
129 Currently, we only guarantee that I32 is *at least* 32 bits.
130 Specifically, if int is 64 bits, then so is I32. (This is the case
131 for the Cray.) This has the advantage of meshing nicely with
132 standard library calls (where we pass an I32 and the library is
133 expecting an int), but the disadvantage that an I32 is not 32 bits.
134 Andy Dougherty August 1996
24fef2a7 135
dc45a647
MB
136 There is no guarantee that there is *any* integral type with
137 exactly 32 bits. It is perfectly legal for a system to have
138 sizeof(short) == sizeof(int) == sizeof(long) == 8.
693762b4 139
dc45a647
MB
140 Similarly, there is no guarantee that I16 and U16 have exactly 16
141 bits.
693762b4 142
8e84507e
NIS
143 For dealing with issues that may arise from various 32/64-bit
144 systems, we will ask Configure to check out
8175356b 145
3609ea0d
JH
146 SHORTSIZE == sizeof(short)
147 INTSIZE == sizeof(int)
148 LONGSIZE == sizeof(long)
dc45a647 149 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
3609ea0d 150 PTRSIZE == sizeof(void *)
dc45a647
MB
151 DOUBLESIZE == sizeof(double)
152 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
8175356b 153
27d4fb96
PP
154*/
155
69512466
JH
156#ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
157# include <inttypes.h>
dd0eed91
JH
158# ifdef INT32_MIN_BROKEN
159# undef INT32_MIN
160# define INT32_MIN (-2147483647-1)
161# endif
162# ifdef INT64_MIN_BROKEN
163# undef INT64_MIN
164# define INT64_MIN (-9223372036854775807LL-1)
165# endif
69512466
JH
166#endif
167
8175356b
JH
168typedef I8TYPE I8;
169typedef U8TYPE U8;
170typedef I16TYPE I16;
171typedef U16TYPE U16;
172typedef I32TYPE I32;
173typedef U32TYPE U32;
16d89be8 174
74b807c7 175#ifdef QUADKIND
8175356b
JH
176typedef I64TYPE I64;
177typedef U64TYPE U64;
16d89be8 178#endif
8175356b 179
d8668976 180#if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
5ff3f7a4 181
5ff3f7a4
GS
182/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
183 Please search CHAR_MAX in perl.h for further details. */
184#define U8_MAX UINT8_MAX
185#define U8_MIN UINT8_MIN
186
5ff3f7a4
GS
187#define I16_MAX INT16_MAX
188#define I16_MIN INT16_MIN
189#define U16_MAX UINT16_MAX
190#define U16_MIN UINT16_MIN
191
5ff3f7a4
GS
192#define I32_MAX INT32_MAX
193#define I32_MIN INT32_MIN
0e983133
GS
194#ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
195# define U32_MAX UINT32_MAX
196#else
197# define U32_MAX 4294967295U
198#endif
5ff3f7a4
GS
199#define U32_MIN UINT32_MIN
200
201#else
202
5c9fa16e
KA
203/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
204 Please search CHAR_MAX in perl.h for further details. */
27d4fb96
PP
205#define U8_MAX PERL_UCHAR_MAX
206#define U8_MIN PERL_UCHAR_MIN
79072805 207
27d4fb96
PP
208#define I16_MAX PERL_SHORT_MAX
209#define I16_MIN PERL_SHORT_MIN
210#define U16_MAX PERL_USHORT_MAX
211#define U16_MIN PERL_USHORT_MIN
79072805 212
c4f23d77 213#if LONGSIZE > 4
27d4fb96
PP
214# define I32_MAX PERL_INT_MAX
215# define I32_MIN PERL_INT_MIN
216# define U32_MAX PERL_UINT_MAX
217# define U32_MIN PERL_UINT_MIN
79072805 218#else
27d4fb96
PP
219# define I32_MAX PERL_LONG_MAX
220# define I32_MIN PERL_LONG_MIN
221# define U32_MAX PERL_ULONG_MAX
222# define U32_MIN PERL_ULONG_MIN
79072805
LW
223#endif
224
5ff3f7a4
GS
225#endif
226
247cee9f
KW
227/* These C99 typedefs are useful sometimes for, say, loop variables whose
228 * maximum values are small, but for which speed trumps size. If we have a C99
229 * compiler, use that. Otherwise, a plain 'int' should be good enough.
230 *
231 * Restrict these to core for now until we are more certain this is a good
232 * idea. */
233#if defined(PERL_CORE) || defined(PERL_EXT)
234# ifdef I_STDINT
235 typedef int_fast8_t PERL_INT_FAST8_T;
236 typedef uint_fast8_t PERL_UINT_FAST8_T;
237 typedef int_fast16_t PERL_INT_FAST16_T;
238 typedef uint_fast16_t PERL_UINT_FAST16_T;
239# else
240 typedef int PERL_INT_FAST8_T;
241 typedef unsigned int PERL_UINT_FAST8_T;
242 typedef int PERL_INT_FAST16_T;
243 typedef unsigned int PERL_UINT_FAST16_T;
244# endif
245#endif
246
464decb6
KW
247/* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
248 * anyone is grepping for it */
249#define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
fc36a67e
PP
250#define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
251#define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
252
88794300 253/* Unused by core; should be deprecated */
ff68c719 254#define Ctl(ch) ((ch) & 037)
8d063cd8 255
98fce2a4
KW
256#if defined(PERL_CORE) || defined(PERL_EXT)
257# ifndef MIN
258# define MIN(a,b) ((a) < (b) ? (a) : (b))
259# endif
260# ifndef MAX
261# define MAX(a,b) ((a) > (b) ? (a) : (b))
262# endif
263#endif
264
84ff4fa9
KW
265/* Returns a boolean as to whether the input unsigned number is a power of 2
266 * (2**0, 2**1, etc). In other words if it has just a single bit set.
267 * If not, subtracting 1 would leave the uppermost bit set, so the & would
268 * yield non-zero */
269#if defined(PERL_CORE) || defined(PERL_EXT)
011b1419 270# define isPOWER_OF_2(n) ((n) && ((n) & ((n)-1)) == 0)
84ff4fa9
KW
271#endif
272
8d9433eb
KW
273/*
274=for apidoc Am|void|__ASSERT_|bool expr
275
276This is a helper macro to avoid preprocessor issues, replaced by nothing
277unless under DEBUGGING, where it expands to an assert of its argument,
278followed by a comma (hence the comma operator). If we just used a straight
279assert(), we would get a comma with nothing before it when not DEBUGGING.
280
281=cut
282
283We also use empty definition under Coverity since the __ASSERT__
284checks often check for things that Really Cannot Happen, and Coverity
285detects that and gets all excited. */
3e94db23
JH
286
287#if defined(DEBUGGING) && !defined(__COVERITY__)
0f092d08
KW
288# define __ASSERT_(statement) assert(statement),
289#else
290# define __ASSERT_(statement)
291#endif
292
3fe05580 293/*
a4ee4fb5 294=head1 SV Manipulation Functions
3fe05580 295
1568d13a
Z
296=for apidoc Ama|SV*|newSVpvs|"literal string" s
297Like C<newSVpvn>, but takes a literal string instead of a
30a15352 298string/length pair.
3fe05580 299
1568d13a
Z
300=for apidoc Ama|SV*|newSVpvs_flags|"literal string" s|U32 flags
301Like C<newSVpvn_flags>, but takes a literal string instead of
30a15352 302a string/length pair.
84bafc02 303
1568d13a
Z
304=for apidoc Ama|SV*|newSVpvs_share|"literal string" s
305Like C<newSVpvn_share>, but takes a literal string instead of
30a15352 306a string/length pair and omits the hash parameter.
3fe05580 307
1568d13a
Z
308=for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string" s|I32 flags
309Like C<sv_catpvn_flags>, but takes a literal string instead
30a15352 310of a string/length pair.
9dcc53ea 311
1568d13a
Z
312=for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string" s
313Like C<sv_catpvn_nomg>, but takes a literal string instead of
0c395ea5 314a string/length pair.
9dcc53ea 315
1568d13a
Z
316=for apidoc Am|void|sv_catpvs|SV* sv|"literal string" s
317Like C<sv_catpvn>, but takes a literal string instead of a
0c395ea5 318string/length pair.
3fe05580 319
1568d13a
Z
320=for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string" s
321Like C<sv_catpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
322string/length pair.
323
1568d13a
Z
324=for apidoc Am|void|sv_setpvs|SV* sv|"literal string" s
325Like C<sv_setpvn>, but takes a literal string instead of a
0c395ea5 326string/length pair.
3fe05580 327
1568d13a
Z
328=for apidoc Am|void|sv_setpvs_mg|SV* sv|"literal string" s
329Like C<sv_setpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
330string/length pair.
331
78342678 332=for apidoc Am|SV *|sv_setref_pvs|SV *const rv|const char *const classname|"literal string" s
1568d13a 333Like C<sv_setref_pvn>, but takes a literal string instead of
0c395ea5 334a string/length pair.
9dcc53ea 335
3fe05580
MHM
336=head1 Memory Management
337
1568d13a
Z
338=for apidoc Ama|char*|savepvs|"literal string" s
339Like C<savepvn>, but takes a literal string instead of a
30a15352 340string/length pair.
3fe05580 341
1568d13a 342=for apidoc Ama|char*|savesharedpvs|"literal string" s
9dcc53ea
Z
343A version of C<savepvs()> which allocates the duplicate string in memory
344which is shared between threads.
345
3fe05580
MHM
346=head1 GV Functions
347
1568d13a
Z
348=for apidoc Am|HV*|gv_stashpvs|"literal string" name|I32 create
349Like C<gv_stashpvn>, but takes a literal string instead of a
0c395ea5 350string/length pair.
3fe05580
MHM
351
352=head1 Hash Manipulation Functions
353
1568d13a
Z
354=for apidoc Am|SV**|hv_fetchs|HV* tb|"literal string" key|I32 lval
355Like C<hv_fetch>, but takes a literal string instead of a
0c395ea5 356string/length pair.
3fe05580 357
d106cb46 358=for apidoc Am|SV**|hv_stores|HV* tb|"literal string" key|SV* val
1568d13a 359Like C<hv_store>, but takes a literal string instead of a
0c395ea5 360string/length pair
3fe05580
MHM
361and omits the hash parameter.
362
510966aa
Z
363=head1 Lexer interface
364
1568d13a 365=for apidoc Amx|void|lex_stuff_pvs|"literal string" pv|U32 flags
510966aa 366
1568d13a 367Like L</lex_stuff_pvn>, but takes a literal string instead of
0c395ea5 368a string/length pair.
510966aa 369
3fe05580
MHM
370=cut
371*/
372
926bb54c
KW
373/* concatenating with "" ensures that only literal strings are accepted as
374 * argument */
26bfd924 375#define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
2efa8cc7 376
926bb54c
KW
377/* note that STR_WITH_LEN() can't be used as argument to macros or functions
378 * that under some configurations might be macros, which means that it requires
379 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
ba3a79e7
GA
380 */
381
382/* STR_WITH_LEN() shortcuts */
383#define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
84bafc02
NC
384#define newSVpvs_flags(str,flags) \
385 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
ba3a79e7 386#define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
9dcc53ea
Z
387#define sv_catpvs_flags(sv, str, flags) \
388 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
389#define sv_catpvs_nomg(sv, str) \
390 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
391#define sv_catpvs(sv, str) \
392 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
393#define sv_catpvs_mg(sv, str) \
394 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
3fe05580 395#define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
9dcc53ea
Z
396#define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
397#define sv_setref_pvs(rv, classname, str) \
398 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
ba3a79e7 399#define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
9dcc53ea
Z
400#define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
401#define gv_stashpvs(str, create) \
402 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
403#define gv_fetchpvs(namebeg, add, sv_type) \
404 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
405#define gv_fetchpvn(namebeg, len, add, sv_type) \
406 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
407#define sv_catxmlpvs(dsv, str, utf8) \
408 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
4ac46235 409
ba3a79e7 410
510966aa
Z
411#define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
412
b96d8cd9
NC
413#define get_cvs(str, flags) \
414 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
5c1737d1 415
954c1994 416/*
ccfc67b7
JH
417=head1 Miscellaneous Functions
418
954c1994 419=for apidoc Am|bool|strNE|char* s1|char* s2
dc6b0978
KW
420Test two C<NUL>-terminated strings to see if they are different. Returns true
421or false.
954c1994
GS
422
423=for apidoc Am|bool|strEQ|char* s1|char* s2
dc6b0978
KW
424Test two C<NUL>-terminated strings to see if they are equal. Returns true or
425false.
954c1994
GS
426
427=for apidoc Am|bool|strLT|char* s1|char* s2
dc6b0978
KW
428Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
429second, C<s2>. Returns true or false.
954c1994
GS
430
431=for apidoc Am|bool|strLE|char* s1|char* s2
dc6b0978
KW
432Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
433equal to the second, C<s2>. Returns true or false.
954c1994
GS
434
435=for apidoc Am|bool|strGT|char* s1|char* s2
dc6b0978
KW
436Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
437the second, C<s2>. Returns true or false.
954c1994
GS
438
439=for apidoc Am|bool|strGE|char* s1|char* s2
dc6b0978
KW
440Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
441or equal to the second, C<s2>. Returns true or false.
954c1994
GS
442
443=for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
dc6b0978
KW
444Test two C<NUL>-terminated strings to see if they are different. The C<len>
445parameter indicates the number of bytes to compare. Returns true or false. (A
954c1994
GS
446wrapper for C<strncmp>).
447
448=for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
dc6b0978
KW
449Test two C<NUL>-terminated strings to see if they are equal. The C<len>
450parameter indicates the number of bytes to compare. Returns true or false. (A
451wrapper for C<strncmp>).
954c1994 452
bd18bd40
KW
453=for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
454Test two buffers (which may contain embedded C<NUL> characters, to see if they
455are equal. The C<len> parameter indicates the number of bytes to compare.
456Returns zero if equal, or non-zero if non-equal.
457
2d8eeddb
KW
458=for apidoc Am|bool|memEQs|char* s1|STRLEN l1|"literal string" s2
459Like L</memEQ>, but the second string is a literal enclosed in double quotes,
460C<l1> gives the number of bytes in C<s1>.
461Returns zero if equal, or non-zero if non-equal.
462
bd18bd40
KW
463=for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
464Test two buffers (which may contain embedded C<NUL> characters, to see if they
465are not equal. The C<len> parameter indicates the number of bytes to compare.
466Returns zero if non-equal, or non-zero if equal.
467
2d8eeddb
KW
468=for apidoc Am|bool|memNEs|char* s1|STRLEN l1|"literal string" s2
469Like L</memNE>, but the second string is a literal enclosed in double quotes,
470C<l1> gives the number of bytes in C<s1>.
471Returns zero if non-equal, or zero if non-equal.
472
954c1994 473=cut
fc169e00
KW
474
475New macros should use the following conventions for their names (which are
476based on the underlying C library functions):
477
478 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
479
480 Each has two main parameters, string-like operands that are compared
481 against each other, as specified by the macro name. Some macros may
482 additionally have one or potentially even two length parameters. If a length
483 parameter applies to both string parameters, it will be positioned third;
484 otherwise any length parameter immediately follows the string parameter it
485 applies to.
486
487 If the prefix to the name is 'str', the string parameter is a pointer to a C
488 language string. Such a string does not contain embedded NUL bytes; its
489 length may be unknown, but can be calculated by C<strlen()>, since it is
490 terminated by a NUL, which isn't included in its length.
491
492 The optional 'n' following 'str' means that that there is a third parameter,
493 giving the maximum number of bytes to look at in each string. Even if both
494 strings are longer than the length parameter, those extra bytes will be
495 unexamined.
496
497 The 's' suffix means that the 2nd byte string parameter is a literal C
498 double-quoted string. Its length will automatically be calculated by the
499 macro, so no length parameter will ever be needed for it.
500
501 If the prefix is 'mem', the string parameters don't have to be C strings;
502 they may contain embedded NUL bytes, do not necessarily have a terminating
503 NUL, and their lengths can be known only through other means, which in
504 practice are additional parameter(s) passed to the function. All 'mem'
505 functions have at least one length parameter. Barring any 'l' or 's' suffix,
506 there is a single length parameter, in position 3, which applies to both
507 string parameters. The 's' suffix means, as described above, that the 2nd
508 string is a literal double-quoted C string (hence its length is calculated by
509 the macro, and the length parameter to the function applies just to the first
510 string parameter, and hence is positioned just after it). An 'l' suffix
511 means that the 2nd string parameter has its own length parameter, and the
512 signature will look like memFOOl(s1, l1, s2, l2).
513
514 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
515 substring of the 1st string. 'P' if present indicates that the substring
516 must be a "proper" one in tha mathematical sense that the first one must be
517 strictly larger than the 2nd.
518
954c1994
GS
519*/
520
62946e08 521
75400963
KW
522#define strNE(s1,s2) (strcmp(s1,s2) != 0)
523#define strEQ(s1,s2) (strcmp(s1,s2) == 0)
8d063cd8
LW
524#define strLT(s1,s2) (strcmp(s1,s2) < 0)
525#define strLE(s1,s2) (strcmp(s1,s2) <= 0)
526#define strGT(s1,s2) (strcmp(s1,s2) > 0)
527#define strGE(s1,s2) (strcmp(s1,s2) >= 0)
62946e08 528
75400963
KW
529#define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
530#define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
378cc40b 531
9d3980bc
KW
532#define memEQ(s1,s2,l) (memcmp(((const void *) (s1)), ((const void *) (s2)), l) == 0)
533#define memNE(s1,s2,l) (! memEQ(s1,s2,l))
36477c24 534
085b7534 535/* memEQ and memNE where second comparand is a string constant */
568a785a 536#define memEQs(s1, l, s2) \
777fa2cb 537 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
5f50c6c9 538#define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
568a785a 539
fdbb9a7c
KW
540/* Keep these private until we decide it was a good idea */
541#if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
542
543#define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
544
bdb7e3f0 545#define memBEGINs(s1, l, s2) \
30a6480c 546 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
bdb7e3f0 547 && memEQ(s1, "" s2 "", sizeof(s2)-1))
de627158 548#define memBEGINPs(s1, l, s2) \
30a6480c 549 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) - 1 \
de627158 550 && memEQ(s1, "" s2 "", sizeof(s2)-1))
bdb7e3f0 551#define memENDs(s1, l, s2) \
30a6480c 552 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
bdb7e3f0 553 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
b80f8424 554#define memENDPs(s1, l, s2) \
30a6480c 555 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) \
b80f8424 556 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
fdbb9a7c 557#endif /* End of making macros private */
bdb7e3f0 558
062b6850
KW
559#define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
560#define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
561#define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
562#define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
563
bbce6d69
PP
564/*
565 * Character classes.
566 *
567 * Unfortunately, the introduction of locales means that we
568 * can't trust isupper(), etc. to tell the truth. And when
569 * it comes to /\w+/ with tainting enabled, we *must* be able
570 * to trust our character classes.
571 *
572 * Therefore, the default tests in the text of Perl will be
573 * independent of locale. Any code that wants to depend on
574 * the current locale will use the tests that begin with "lc".
575 */
576
2304df62
AD
577#ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
578# ifndef CTYPE256
579# define CTYPE256
580# endif
581#endif
582
954c1994 583/*
ccfc67b7 584
dcccc8ff 585=head1 Character classification
243effed
KW
586This section is about functions (really macros) that classify characters
587into types, such as punctuation versus alphabetic, etc. Most of these are
588analogous to regular expression character classes. (See
589L<perlrecharclass/POSIX Character Classes>.) There are several variants for
590each class. (Not all macros have all variants; each item below lists the
591ones valid for it.) None are affected by C<use bytes>, and only the ones
592with C<LC> in the name are affected by the current locale.
593
d713f9d9
KW
594The base function, e.g., C<isALPHA()>, takes any signed or unsigned value,
595treating it as a code point, and returns a boolean as to whether or not the
596character represented by it is (or on non-ASCII platforms, corresponds to) an
6aff1f14
KW
597ASCII character in the named class based on platform, Unicode, and Perl rules.
598If the input is a number that doesn't fit in an octet, FALSE is returned.
243effed 599
c98722a4 600Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
550da823
KW
601with no suffix C<"_A">. This variant is used to emphasize by its name that
602only ASCII-range characters can return TRUE.
4b9734bf 603
d60679e1 604Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
4b9734bf
KW
605onto the platform. That is, the code points that are ASCII are unaffected,
606since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
607as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
608true when called with the code point 0xDF, which is a word character in both
4650c663 609ASCII and EBCDIC (though it represents different characters in each).
d713f9d9
KW
610If the input is a number that doesn't fit in an octet, FALSE is returned.
611(Perl's documentation uses a colloquial definition of Latin-1, to include all
612code points below 256.)
243effed 613
d713f9d9
KW
614Variant C<isI<FOO>_uvchr> is exactly like the C<isI<FOO>_L1> variant, for
615inputs below 256, but if the code point is larger than 255, Unicode rules are
616used to determine if it is in the character class. For example,
d0da05db 617C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
6aff1f14 618WITH MACRON in Unicode, and is a word character.
243effed 619
c98722a4 620Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
d713f9d9
KW
621encoded strings. Each call classifies the first character of the string. This
622variant takes two parameters. The first, C<p>, is a
da8c1a98
KW
623pointer to the first byte of the character to be classified. (Recall that it
624may take more than one byte to represent a character in UTF-8 strings.) The
625second parameter, C<e>, points to anywhere in the string beyond the first
626character, up to one byte past the end of the entire string. The suffix
627C<_safe> in the function's name indicates that it will not attempt to read
628beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
629is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
630character is malformed in some way, the program may croak, or the function may
631return FALSE, at the discretion of the implementation, and subject to change in
632future releases.
633
c98722a4 634Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
da8c1a98 635parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 636in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
8b4b30c5 637beyond the end of the string. Starting in Perl v5.32, it will take a second
c98722a4 638parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
34aeb2e9 639program that uses it will have to be changed to successfully compile. In the
c98722a4 640meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
34aeb2e9 641program will raise a deprecation warning, enabled by default. You can convert
c98722a4 642your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 643extra measure of protection, or you can wait until v5.32, when you'll be forced
34aeb2e9 644to add the C<e> parameter.
243effed 645
d713f9d9
KW
646Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants,
647but the result is based on the current locale, which is what C<LC> in the name
648stands for. If Perl can determine that the current locale is a UTF-8 locale,
649it uses the published Unicode rules; otherwise, it uses the C library function
650that gives the named classification. For example, C<isDIGIT_LC()> when not in
651a UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
1a83413c
KW
652returned if the input won't fit into an octet. On some platforms where the C
653library function is known to be defective, Perl changes its result to follow
654the POSIX standard's rules.
243effed 655
d713f9d9
KW
656Variant C<isI<FOO>_LC_uvchr> acts exactly like C<isI<FOO>_LC> for inputs less
657than 256, but for larger ones it returns the Unicode classification of the code
658point.
243effed 659
c98722a4 660Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
d713f9d9
KW
661encoded strings. Each call classifies the first character of the string. This
662variant takes two parameters. The first, C<p>, is a pointer to the first byte
663of the character to be classified. (Recall that it may take more than one byte
664to represent a character in UTF-8 strings.) The second parameter, C<e>,
665points to anywhere in the string beyond the first character, up to one byte
666past the end of the entire string. The suffix C<_safe> in the function's name
667indicates that it will not attempt to read beyond S<C<e - 1>>, provided that
668the constraint S<C<s E<lt> e>> is true (this is asserted for in C<-DDEBUGGING>
669builds). If the UTF-8 for the input character is malformed in some way, the
670program may croak, or the function may return FALSE, at the discretion of the
671implementation, and subject to change in future releases.
da8c1a98 672
c98722a4 673Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
da8c1a98 674parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 675in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
8b4b30c5 676beyond the end of the string. Starting in Perl v5.32, it will take a second
c98722a4 677parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
34aeb2e9 678program that uses it will have to be changed to successfully compile. In the
c98722a4 679meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
34aeb2e9 680the program will raise a deprecation warning, enabled by default. You can
c98722a4 681convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
8b4b30c5 682and get an extra measure of protection, or you can wait until v5.32, when
34aeb2e9 683you'll be forced to add the C<e> parameter.
ccfc67b7 684
d713f9d9
KW
685=for apidoc Am|bool|isALPHA|int ch
686Returns a boolean indicating whether the specified input is one of C<[A-Za-z]>,
687analogous to C<m/[[:alpha:]]/>.
dcccc8ff
KW
688See the L<top of this section|/Character classification> for an explanation of
689variants
da8c1a98
KW
690C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
691C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
8a58bdcf 692
d713f9d9
KW
693=for apidoc Am|bool|isALPHANUMERIC|int ch
694Returns a boolean indicating whether the specified character is one of
695C<[A-Za-z0-9]>, analogous to C<m/[[:alnum:]]/>.
dcccc8ff
KW
696See the L<top of this section|/Character classification> for an explanation of
697variants
d0da05db 698C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
da8c1a98
KW
699C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
700and C<isALPHANUMERIC_LC_utf8_safe>.
15861f94 701
255b632a
KW
702A (discouraged from use) synonym is C<isALNUMC> (where the C<C> suffix means
703this corresponds to the C language alphanumeric definition). Also
704there are the variants
705C<isALNUMC_A>, C<isALNUMC_L1>
706C<isALNUMC_LC>, and C<isALNUMC_LC_uvchr>.
707
d713f9d9 708=for apidoc Am|bool|isASCII|int ch
8a58bdcf 709Returns a boolean indicating whether the specified character is one of the 128
243effed 710characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
e5ad6aba 711On non-ASCII platforms, it returns TRUE iff this
8a58bdcf
KW
712character corresponds to an ASCII character. Variants C<isASCII_A()> and
713C<isASCII_L1()> are identical to C<isASCII()>.
dcccc8ff
KW
714See the L<top of this section|/Character classification> for an explanation of
715variants
da8c1a98
KW
716C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
717C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
243effed
KW
718library routine C<isascii()>. In these cases, the variants whose names contain
719C<LC> are the same as the corresponding ones without.
720
d98532ea
KW
721Also note, that because all ASCII characters are UTF-8 invariant (meaning they
722have the exact same representation (always a single byte) whether encoded in
723UTF-8 or not), C<isASCII> will give the correct results when called with any
da8c1a98
KW
724byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
725will work properly on any string encoded or not in UTF-8.
d98532ea 726
243effed
KW
727=for apidoc Am|bool|isBLANK|char ch
728Returns a boolean indicating whether the specified character is a
6aff1f14 729character considered to be a blank, analogous to C<m/[[:blank:]]/>.
dcccc8ff
KW
730See the L<top of this section|/Character classification> for an explanation of
731variants
da8c1a98
KW
732C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
733C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
734however, that some platforms do not have the C library routine
735C<isblank()>. In these cases, the variants whose names contain C<LC> are
736the same as the corresponding ones without.
243effed
KW
737
738=for apidoc Am|bool|isCNTRL|char ch
739Returns a boolean indicating whether the specified character is a
6aff1f14 740control character, analogous to C<m/[[:cntrl:]]/>.
dcccc8ff
KW
741See the L<top of this section|/Character classification> for an explanation of
742variants
da8c1a98
KW
743C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
744C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
745platforms, you almost always want to use the C<isCNTRL_L1> variant.
954c1994 746
954c1994 747=for apidoc Am|bool|isDIGIT|char ch
2787a470 748Returns a boolean indicating whether the specified character is a
6aff1f14 749digit, analogous to C<m/[[:digit:]]/>.
8a58bdcf 750Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
dcccc8ff
KW
751See the L<top of this section|/Character classification> for an explanation of
752variants
da8c1a98
KW
753C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
754C<isDIGIT_LC_utf8_safe>.
243effed
KW
755
756=for apidoc Am|bool|isGRAPH|char ch
757Returns a boolean indicating whether the specified character is a
6aff1f14 758graphic character, analogous to C<m/[[:graph:]]/>.
dcccc8ff 759See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
760variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
761C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
954c1994 762
0c82b6df 763=for apidoc Am|bool|isLOWER|char ch
2787a470 764Returns a boolean indicating whether the specified character is a
6aff1f14 765lowercase character, analogous to C<m/[[:lower:]]/>.
dcccc8ff
KW
766See the L<top of this section|/Character classification> for an explanation of
767variants
da8c1a98
KW
768C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
769C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
0c82b6df 770
c99e91e9 771=for apidoc Am|bool|isOCTAL|char ch
2787a470 772Returns a boolean indicating whether the specified character is an
6aff1f14 773octal digit, [0-7].
243effed
KW
774The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
775C<isOCTAL>.
776
777=for apidoc Am|bool|isPUNCT|char ch
778Returns a boolean indicating whether the specified character is a
6aff1f14
KW
779punctuation character, analogous to C<m/[[:punct:]]/>.
780Note that the definition of what is punctuation isn't as
243effed
KW
781straightforward as one might desire. See L<perlrecharclass/POSIX Character
782Classes> for details.
dcccc8ff 783See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
784variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
785C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
c99e91e9 786
0c82b6df 787=for apidoc Am|bool|isSPACE|char ch
2787a470 788Returns a boolean indicating whether the specified character is a
6aff1f14 789whitespace character. This is analogous
398d098a 790to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
779cf272 791this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
398d098a
KW
792locale forms of this macro (the ones with C<LC> in their names) matched
793precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
794in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
795(See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
dcccc8ff
KW
796See the L<top of this section|/Character classification> for an explanation of
797variants
da8c1a98
KW
798C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
799C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
0c82b6df 800
398d098a
KW
801=for apidoc Am|bool|isPSXSPC|char ch
802(short for Posix Space)
779cf272
KW
803Starting in 5.18, this is identical in all its forms to the
804corresponding C<isSPACE()> macros.
398d098a
KW
805The locale forms of this macro are identical to their corresponding
806C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
807non-locale forms differ from their C<isSPACE()> forms only in that the
808C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
809Otherwise they are identical. Thus this macro is analogous to what
810C<m/[[:space:]]/> matches in a regular expression.
dcccc8ff 811See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
812variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
813C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
398d098a 814
954c1994 815=for apidoc Am|bool|isUPPER|char ch
2787a470 816Returns a boolean indicating whether the specified character is an
6aff1f14 817uppercase character, analogous to C<m/[[:upper:]]/>.
dcccc8ff 818See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
819variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
820C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
954c1994 821
243effed 822=for apidoc Am|bool|isPRINT|char ch
8eea39dd 823Returns a boolean indicating whether the specified character is a
6aff1f14 824printable character, analogous to C<m/[[:print:]]/>.
dcccc8ff
KW
825See the L<top of this section|/Character classification> for an explanation of
826variants
da8c1a98
KW
827C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
828C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
243effed
KW
829
830=for apidoc Am|bool|isWORDCHAR|char ch
831Returns a boolean indicating whether the specified character is a character
832that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
833in a regular expression. A word character is an alphabetic character, a
834decimal digit, a connecting punctuation character (such as an underscore), or
835a "mark" character that attaches to one of those (like some sort of accent).
836C<isALNUM()> is a synonym provided for backward compatibility, even though a
837word character includes more than the standard C language meaning of
838alphanumeric.
dcccc8ff 839See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
840variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
841C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
842C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
843include the platform's native underscore.
8a58bdcf
KW
844
845=for apidoc Am|bool|isXDIGIT|char ch
846Returns a boolean indicating whether the specified character is a hexadecimal
243effed
KW
847digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
848and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
dcccc8ff
KW
849See the L<top of this section|/Character classification> for an explanation of
850variants
da8c1a98
KW
851C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
852and C<isXDIGIT_LC_utf8_safe>.
243effed 853
3c3ecf18
KW
854=for apidoc Am|bool|isIDFIRST|char ch
855Returns a boolean indicating whether the specified character can be the first
856character of an identifier. This is very close to, but not quite the same as
857the official Unicode property C<XID_Start>. The difference is that this
858returns true only if the input character also matches L</isWORDCHAR>.
dcccc8ff
KW
859See the L<top of this section|/Character classification> for an explanation of
860variants
da8c1a98
KW
861C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
862C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
3c3ecf18
KW
863
864=for apidoc Am|bool|isIDCONT|char ch
865Returns a boolean indicating whether the specified character can be the
866second or succeeding character of an identifier. This is very close to, but
867not quite the same as the official Unicode property C<XID_Continue>. The
868difference is that this returns true only if the input character also matches
dcccc8ff
KW
869L</isWORDCHAR>. See the L<top of this section|/Character classification> for
870an
d0da05db 871explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
da8c1a98
KW
872C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
873C<isIDCONT_LC_utf8_safe>.
3c3ecf18 874
243effed 875=head1 Miscellaneous Functions
8eea39dd 876
95a59cab 877=for apidoc Am|U8|READ_XDIGIT|char str*
243effed 878Returns the value of an ASCII-range hex digit and advances the string pointer.
95a59cab
YO
879Behaviour is only well defined when isXDIGIT(*str) is true.
880
e7c1e6c1 881=head1 Character case changing
21da7284
KW
882Perl uses "full" Unicode case mappings. This means that converting a single
883character to another case may result in a sequence of more than one character.
884For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
885character sequence C<SS>. This presents some complications The lowercase of
886all characters in the range 0..255 is a single character, and thus
887C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
888return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
889an API that does allow every possible legal result to be returned.) Likewise
890no other function that is crippled by not being able to give the correct
891results for the full range of possible inputs has been implemented here.
e7c1e6c1 892
d713f9d9 893=for apidoc Am|U8|toUPPER|int ch
1f607577
KW
894Converts the specified character to uppercase. If the input is anything but an
895ASCII lowercase character, that input character itself is returned. Variant
c753c8d3 896C<toUPPER_A> is equivalent.
954c1994 897
d0da05db
KW
898=for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
899Converts the code point C<cp> to its uppercase version, and
900stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
901point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
902that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
903bytes since the uppercase version may be longer than the original character.
904
905The first code point of the uppercased version is returned
21da7284
KW
906(but note, as explained at L<the top of this section|/Character case
907changing>, that there may be more.)
1f607577 908
a239b1e2
KW
909=for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
910Converts the first UTF-8 encoded character in the sequence starting at C<p> and
911extending no further than S<C<e - 1>> to its uppercase version, and
1f607577
KW
912stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
913that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
914bytes since the uppercase version may be longer than the original character.
915
916The first code point of the uppercased version is returned
21da7284
KW
917(but note, as explained at L<the top of this section|/Character case
918changing>, that there may be more).
1f607577 919
a239b1e2
KW
920The suffix C<_safe> in the function's name indicates that it will not attempt
921to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
922true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
923input character is malformed in some way, the program may croak, or the
924function may return the REPLACEMENT CHARACTER, at the discretion of the
925implementation, and subject to change in future releases.
926
927=for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
928This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
929parameter The function therefore can't check if it is reading
8b4b30c5 930beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
931parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
932program that uses it will have to be changed to successfully compile. In the
933meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
934program will raise a deprecation warning, enabled by default. You can convert
935your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 936extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 937to add the C<e> parameter.
1f607577 938
25200305
KW
939=for apidoc Am|U8|toFOLD|U8 ch
940Converts the specified character to foldcase. If the input is anything but an
941ASCII uppercase character, that input character itself is returned. Variant
942C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
d0da05db 943Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
25200305 944
d0da05db
KW
945=for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
946Converts the code point C<cp> to its foldcase version, and
947stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
948point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
949that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
950bytes since the foldcase version may be longer than the original character.
951
952The first code point of the foldcased version is returned
21da7284
KW
953(but note, as explained at L<the top of this section|/Character case
954changing>, that there may be more).
1f607577 955
a239b1e2
KW
956=for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
957Converts the first UTF-8 encoded character in the sequence starting at C<p> and
958extending no further than S<C<e - 1>> to its foldcase version, and
1f607577
KW
959stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
960that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
961bytes since the foldcase version may be longer than the original character.
962
963The first code point of the foldcased version is returned
21da7284
KW
964(but note, as explained at L<the top of this section|/Character case
965changing>, that there may be more).
1f607577 966
a239b1e2
KW
967The suffix C<_safe> in the function's name indicates that it will not attempt
968to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
969true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
970input character is malformed in some way, the program may croak, or the
971function may return the REPLACEMENT CHARACTER, at the discretion of the
972implementation, and subject to change in future releases.
973
974=for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
975This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
976parameter The function therefore can't check if it is reading
8b4b30c5 977beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
978parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
979program that uses it will have to be changed to successfully compile. In the
980meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
981program will raise a deprecation warning, enabled by default. You can convert
982your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 983extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 984to add the C<e> parameter.
1f607577
KW
985
986=for apidoc Am|U8|toLOWER|U8 ch
987Converts the specified character to lowercase. If the input is anything but an
988ASCII uppercase character, that input character itself is returned. Variant
c753c8d3 989C<toLOWER_A> is equivalent.
954c1994 990
1f607577 991=for apidoc Am|U8|toLOWER_L1|U8 ch
b7d90381
KW
992Converts the specified Latin1 character to lowercase. The results are
993undefined if the input doesn't fit in a byte.
1f607577
KW
994
995=for apidoc Am|U8|toLOWER_LC|U8 ch
996Converts the specified character to lowercase using the current locale's rules,
997if possible; otherwise returns the input character itself.
998
d0da05db
KW
999=for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
1000Converts the code point C<cp> to its lowercase version, and
1001stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1002point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1003that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1004bytes since the lowercase version may be longer than the original character.
1005
1006The first code point of the lowercased version is returned
21da7284
KW
1007(but note, as explained at L<the top of this section|/Character case
1008changing>, that there may be more).
1f607577 1009
a239b1e2
KW
1010
1011=for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1012Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1013extending no further than S<C<e - 1>> to its lowercase version, and
1f607577
KW
1014stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1015that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1016bytes since the lowercase version may be longer than the original character.
1017
1018The first code point of the lowercased version is returned
21da7284
KW
1019(but note, as explained at L<the top of this section|/Character case
1020changing>, that there may be more).
1f607577 1021
a239b1e2
KW
1022The suffix C<_safe> in the function's name indicates that it will not attempt
1023to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1024true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1025input character is malformed in some way, the program may croak, or the
1026function may return the REPLACEMENT CHARACTER, at the discretion of the
1027implementation, and subject to change in future releases.
1028
1029=for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
1030This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1031parameter The function therefore can't check if it is reading
8b4b30c5 1032beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
1033parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
1034program that uses it will have to be changed to successfully compile. In the
1035meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
1036program will raise a deprecation warning, enabled by default. You can convert
1037your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 1038extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 1039to add the C<e> parameter.
1f607577 1040
25200305
KW
1041=for apidoc Am|U8|toTITLE|U8 ch
1042Converts the specified character to titlecase. If the input is anything but an
1043ASCII lowercase character, that input character itself is returned. Variant
b7d90381 1044C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
d0da05db 1045range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
b7d90381 1046not a concept used in locale handling, so there is no functionality for that.)
25200305 1047
d0da05db
KW
1048=for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1049Converts the code point C<cp> to its titlecase version, and
1050stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1051point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1052that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1053bytes since the titlecase version may be longer than the original character.
1054
1055The first code point of the titlecased version is returned
21da7284
KW
1056(but note, as explained at L<the top of this section|/Character case
1057changing>, that there may be more).
1f607577 1058
a239b1e2
KW
1059=for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1060Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1061extending no further than S<C<e - 1>> to its titlecase version, and
1f607577
KW
1062stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1063that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1064bytes since the titlecase version may be longer than the original character.
1065
1066The first code point of the titlecased version is returned
21da7284
KW
1067(but note, as explained at L<the top of this section|/Character case
1068changing>, that there may be more).
1f607577 1069
a239b1e2
KW
1070The suffix C<_safe> in the function's name indicates that it will not attempt
1071to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1072true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1073input character is malformed in some way, the program may croak, or the
1074function may return the REPLACEMENT CHARACTER, at the discretion of the
1075implementation, and subject to change in future releases.
1076
1077=for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1078This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1079parameter The function therefore can't check if it is reading
8b4b30c5 1080beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
1081parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1082program that uses it will have to be changed to successfully compile. In the
1083meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1084program will raise a deprecation warning, enabled by default. You can convert
1085your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 1086extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 1087to add the C<e> parameter.
1f607577 1088
954c1994 1089=cut
353c9b6f 1090
d0da05db 1091XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1e222e4f
KW
1092really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1093and aren't general purpose as they don't work on U+DF, and assert against that.
243effed 1094
8a58bdcf 1095Note that these macros are repeated in Devel::PPPort, so should also be
62fa66b6
KW
1096patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1097
954c1994
GS
1098*/
1099
de9e2639
KW
1100/* Specify the widest unsigned type on the platform. */
1101#ifdef QUADKIND
1102# define WIDEST_UTYPE U64
7c062697
KW
1103#else
1104# define WIDEST_UTYPE U32
1105#endif
1106
3912bc88
KW
1107/* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1108 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1109 * bits of information are lost even on a 64-bit machine, but to get the
1110 * compiler to optimize it out if possible. This is because Configure makes
1111 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1112 * sizeof() guarantees that this evaluates to a constant true at compile time.
7e75d1a1
JH
1113 *
1114 * For Coverity, be always true, because otherwise Coverity thinks
1115 * it finds several expressions that are always true, independent
1116 * of operands. Well, they are, but that is kind of the point.
220c71bf 1117 */
7e75d1a1 1118#ifndef __COVERITY__
6c5b02ac
KW
1119 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1120 * pointer) */
1121#define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
ace3ad0f 1122 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
7e75d1a1
JH
1123#else
1124#define FITS_IN_8_BITS(c) (1)
1125#endif
cf301eb7 1126
94250c4f
KW
1127/* Returns true if c is in the range l..u, where 'l' is non-negative
1128 * Written this way so that after optimization, only one conditional test is
1129 * needed.
1130 *
1131 * This isn't fully general, except for the special cased 'signed char' (which
1132 * should be resolved at compile time): It won't work if 'c' is negative, and
1133 * 'l' is larger than the max for that signed type. Thus if 'c' is a negative
1134 * int, and 'l' is larger than INT_MAX, it will fail. To protect agains this
1135 * happening, there is an assert that will generate a warning if c is larger
1136 * than e.g. INT_MAX if it is an 'unsigned int'. This could be a false
1137 * positive, but khw couldn't figure out a way to make it better. It's good
1138 * enough so far */
1139#define inRANGE(c, l, u) (__ASSERT_((l) >= 0) __ASSERT_((u) >= (l)) \
1140 ((sizeof(c) == 1) \
1141 ? (((WIDEST_UTYPE) ((((U8) (c))|0) - (l))) <= ((WIDEST_UTYPE) ((u) - (l)))) \
1142 : (__ASSERT_( (((WIDEST_UTYPE) 1) << (CHARBITS * sizeof(c) - 1) & (c)) \
1143 /* sign bit of c is 0 */ == 0 \
1144 || (((~ ((WIDEST_UTYPE) 1) << ((CHARBITS * sizeof(c) - 1) - 1))\
1145 /* l not larger than largest value in c's signed type */ \
1146 & ~ ((WIDEST_UTYPE) 0)) & (l)) == 0) \
1147 ((WIDEST_UTYPE) (((c) - (l)) | 0) <= ((WIDEST_UTYPE) ((u) - (l)))))))
305fe86e 1148
41f43cc2 1149#ifdef EBCDIC
b6340bd0 1150# ifndef _ALL_SOURCE
0852beac
KW
1151 /* The native libc isascii() et.al. functions return the wrong results
1152 * on at least z/OS unless this is defined. */
b6340bd0
KW
1153# error _ALL_SOURCE should probably be defined
1154# endif
41f43cc2 1155#else
0852beac
KW
1156 /* There is a simple definition of ASCII for ASCII platforms. But the
1157 * EBCDIC one isn't so simple, so is defined using table look-up like the
9c903d59 1158 * other macros below.
3f3c579d
KW
1159 *
1160 * The cast here is used instead of '(c) >= 0', because some compilers emit
1161 * a warning that that test is always true when the parameter is an
1162 * unsigned type. khw supposes that it could be written as
1163 * && ((c) == '\0' || (c) > 0)
1164 * to avoid the message, but the cast will likely avoid extra branches even
1165 * with stupid compilers.
1166 *
1167 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1168 * a pointer) */
9c903d59 1169# define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
41f43cc2
KW
1170#endif
1171
38694112
KW
1172/* Take the eight possible bit patterns of the lower 3 bits and you get the
1173 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1174 * can be ignored. If the rest match '0', we have an octal */
1175#define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
c2da0b36 1176
9fb1bf9d 1177#ifdef H_PERL /* If have access to perl.h, lookup in its table */
f4cdb42c 1178
a500dc72
KW
1179/* Character class numbers. For internal core Perl use only. The ones less
1180 * than 32 are used in PL_charclass[] and the ones up through the one that
1181 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1182 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1183 * their actual definitions are here. If that file has a name not used here,
1184 * it won't compile.
1709d539
KW
1185 *
1186 * The first group of these is ordered in what I (khw) estimate to be the
31c7f561 1187 * frequency of their use. This gives a slight edge to exiting a loop earlier
58a3ba2c
KW
1188 * (in reginclass() in regexec.c). Except \v should be last, as it isn't a
1189 * real Posix character class, and some (small) inefficiencies in regular
1190 * expression handling would be introduced by putting it in the middle of those
1191 * that are. Also, cntrl and ascii come after the others as it may be useful
1192 * to group these which have no members that match above Latin1, (or above
1193 * ASCII in the latter case) */
1194
1709d539
KW
1195# define _CC_WORDCHAR 0 /* \w and [:word:] */
1196# define _CC_DIGIT 1 /* \d and [:digit:] */
1197# define _CC_ALPHA 2 /* [:alpha:] */
1198# define _CC_LOWER 3 /* [:lower:] */
1199# define _CC_UPPER 4 /* [:upper:] */
1200# define _CC_PUNCT 5 /* [:punct:] */
1201# define _CC_PRINT 6 /* [:print:] */
15861f94 1202# define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1709d539 1203# define _CC_GRAPH 8 /* [:graph:] */
359b005e 1204# define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
779cf272 1205# define _CC_SPACE 10 /* \s, [:space:] */
34aeb2e9
KW
1206# define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1207 when the deprecated isFOO_utf8()
1208 functions are removed */
b0d691b2
KW
1209# define _CC_BLANK 11 /* [:blank:] */
1210# define _CC_XDIGIT 12 /* [:xdigit:] */
779cf272
KW
1211# define _CC_CNTRL 13 /* [:cntrl:] */
1212# define _CC_ASCII 14 /* [:ascii:] */
1213# define _CC_VERTSPACE 15 /* \v */
1709d539 1214
a0947d7b
KW
1215# define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1216
1709d539 1217/* The members of the third group below do not need to be coordinated with data
3ffc8c70 1218 * structures in regcomp.[ch] and regexec.c. */
779cf272
KW
1219# define _CC_IDFIRST 16
1220# define _CC_CHARNAME_CONT 17
1221# define _CC_NONLATIN1_FOLD 18
1222# define _CC_NONLATIN1_SIMPLE_FOLD 19
1223# define _CC_QUOTEMETA 20
1224# define _CC_NON_FINAL_FOLD 21
1225# define _CC_IS_IN_SOME_FOLD 22
1226# define _CC_MNEMONIC_CNTRL 23
073c22b3 1227
34aeb2e9
KW
1228# define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1229 isFOO_utf8() functions are removed */
1230
073c22b3
KW
1231/* This next group is only used on EBCDIC platforms, so theoretically could be
1232 * shared with something entirely different that's only on ASCII platforms */
5d5376e2
KW
1233# define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1234# define _CC_UTF8_IS_START 29
1235# define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1236# define _CC_UTF8_IS_CONTINUATION 31
1237/* Unused: 24-27
f4cdb42c
KW
1238 * If more bits are needed, one could add a second word for non-64bit
1239 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
37ede926
KW
1240 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1241 * is used only for optimization (as of this writing), and differs in the
1242 * Latin1 range from the ALPHA bit only in two relatively unimportant
a500dc72 1243 * characters: the masculine and feminine ordinal indicators, so removing it
073c22b3
KW
1244 * would just cause /i regexes which match them to run less efficiently.
1245 * Similarly the EBCDIC-only bits are used just for speed, and could be
1246 * replaced by other means */
96ac0975 1247
3a371f2f
KW
1248#if defined(PERL_CORE) || defined(PERL_EXT)
1249/* An enum version of the character class numbers, to help compilers
1250 * optimize */
1251typedef enum {
3a371f2f 1252 _CC_ENUM_ALPHA = _CC_ALPHA,
e8d596e0
KW
1253 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1254 _CC_ENUM_ASCII = _CC_ASCII,
1255 _CC_ENUM_BLANK = _CC_BLANK,
b0d691b2 1256 _CC_ENUM_CASED = _CC_CASED,
e8d596e0 1257 _CC_ENUM_CNTRL = _CC_CNTRL,
3a371f2f
KW
1258 _CC_ENUM_DIGIT = _CC_DIGIT,
1259 _CC_ENUM_GRAPH = _CC_GRAPH,
1260 _CC_ENUM_LOWER = _CC_LOWER,
1261 _CC_ENUM_PRINT = _CC_PRINT,
1262 _CC_ENUM_PUNCT = _CC_PUNCT,
e8d596e0 1263 _CC_ENUM_SPACE = _CC_SPACE,
3a371f2f 1264 _CC_ENUM_UPPER = _CC_UPPER,
e8d596e0 1265 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
3a371f2f 1266 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
e8d596e0 1267 _CC_ENUM_XDIGIT = _CC_XDIGIT
3a371f2f
KW
1268} _char_class_number;
1269#endif
1270
86f72d56 1271#define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
63c61c3f 1272
6635f04f 1273START_EXTERN_C
96ac0975
NC
1274# ifdef DOINIT
1275EXTCONST U32 PL_charclass[] = {
1276# include "l1_char_class_tab.h"
1277};
1278
1279# else /* ! DOINIT */
1280EXTCONST U32 PL_charclass[];
1281# endif
6635f04f 1282END_EXTERN_C
96ac0975 1283
265c1f46 1284 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
430b7c70 1285# define _CC_mask(classnum) (1U << (classnum))
4650c663
KW
1286
1287 /* For internal core Perl use only: the base macro for defining macros like
1288 * isALPHA */
ff7ecfc3 1289# define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
f4cd282c 1290 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
4eeeb416 1291
f4cdb42c
KW
1292 /* The mask for the _A versions of the macros; it just adds in the bit for
1293 * ASCII. */
1294# define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1295
4650c663
KW
1296 /* For internal core Perl use only: the base macro for defining macros like
1297 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1298 * the ASCII bit are present */
b7d90381
KW
1299# define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1300 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1301 == _CC_mask_A(classnum)))
f4cdb42c 1302
26c1d9d8
KW
1303/* On ASCII platforms certain classes form a single range. It's faster to
1304 * special case these. isDIGIT is a single range on all platforms */
b877c1ff
KW
1305# ifdef EBCDIC
1306# define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
1307# define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1308# define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1309# define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
1310# define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1311# else
26c1d9d8 1312 /* By folding the upper and lowercase, we can use a single range */
b877c1ff 1313# define isALPHA_A(c) inRANGE((~('A' ^ 'a') & (c)), 'A', 'Z')
26c1d9d8 1314# define isGRAPH_A(c) inRANGE(c, ' ' + 1, 0x7e)
b877c1ff
KW
1315# define isLOWER_A(c) inRANGE(c, 'a', 'z')
1316# define isPRINT_A(c) inRANGE(c, ' ', 0x7e)
1317# define isUPPER_A(c) inRANGE(c, 'A', 'Z')
1318# endif
15861f94 1319# define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
f4cdb42c
KW
1320# define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1321# define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
b877c1ff 1322# define isDIGIT_A(c) inRANGE(c, '0', '9')
f4cdb42c
KW
1323# define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1324# define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
f4cdb42c 1325# define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
b7d90381
KW
1326# define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1327 */
d95f8b6a 1328# define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
3ded5eb0
KW
1329# define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1330# define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1331# define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1332
1333 /* continuation character for legal NAME in \N{NAME} */
1334# define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1335
1336# define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1337# define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1338# define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1339# define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
b7d90381 1340# define isPSXSPC_L1(c) isSPACE_L1(c)
3ded5eb0
KW
1341# define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1342# define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1343# define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1344# define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1345# define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
f4cdb42c 1346
0852beac
KW
1347# ifdef EBCDIC
1348# define isASCII(c) _generic_isCC(c, _CC_ASCII)
1349# endif
1350
f12c0118
KW
1351 /* Participates in a single-character fold with a character above 255 */
1352# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) ((! cBOOL(FITS_IN_8_BITS(c))) || (PL_charclass[(U8) (c)] & _CC_mask(_CC_NONLATIN1_SIMPLE_FOLD)))
1353
1354 /* Like the above, but also can be part of a multi-char fold */
f4cd282c 1355# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) ((! cBOOL(FITS_IN_8_BITS(c))) || (PL_charclass[(U8) (c)] & _CC_mask(_CC_NONLATIN1_FOLD)))
430b7c70 1356
4eeeb416 1357# define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
26faadbd 1358# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1359 _generic_isCC(c, _CC_NON_FINAL_FOLD)
37ede926 1360# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1361 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
8e35b056
KW
1362# define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1363 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
687c8d01 1364#else /* else we don't have perl.h H_PERL */
3ded5eb0
KW
1365
1366 /* If we don't have perl.h, we are compiling a utility program. Below we
1367 * hard-code various macro definitions that wouldn't otherwise be available
fc273927 1368 * to it. Most are coded based on first principles. These are written to
74665a89 1369 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
182c4ace 1370# define isDIGIT_A(c) inRANGE(c, '0', '9')
0852beac 1371# define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
74665a89
KW
1372# define isSPACE_A(c) (isBLANK_A(c) \
1373 || (c) == '\n' \
1374 || (c) == '\r' \
1375 || (c) == '\v' \
0852beac 1376 || (c) == '\f')
74665a89
KW
1377 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1378 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1379 * performance (if optimization isn't on), and allow the same code to be
1380 * used for both platform types */
182c4ace
KW
1381# define isLOWER_A(c) inRANGE((c), 'a', 'i') \
1382 || inRANGE((c), 'j', 'r') \
1383 || inRANGE((c), 's', 'z')
1384# define isUPPER_A(c) inRANGE((c), 'A', 'I') \
1385 || inRANGE((c), 'J', 'R') \
1386 || inRANGE((c), 'S', 'Z')
a4d7a999
KW
1387# define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1388# define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
3ded5eb0 1389# define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
0852beac 1390# define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
182c4ace
KW
1391# define isXDIGIT_A(c) ( isDIGIT_A(c) \
1392 || inRANGE((c), 'a', 'f') \
1393 || inRANGE((c), 'A', 'F')
74665a89
KW
1394# define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1395 || (c) == '#' || (c) == '$' || (c) == '%' \
1396 || (c) == '&' || (c) == '\'' || (c) == '(' \
1397 || (c) == ')' || (c) == '*' || (c) == '+' \
1398 || (c) == ',' || (c) == '.' || (c) == '/' \
1399 || (c) == ':' || (c) == ';' || (c) == '<' \
1400 || (c) == '=' || (c) == '>' || (c) == '?' \
1401 || (c) == '@' || (c) == '[' || (c) == '\\' \
1402 || (c) == ']' || (c) == '^' || (c) == '_' \
1403 || (c) == '`' || (c) == '{' || (c) == '|' \
1404 || (c) == '}' || (c) == '~')
1405# define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1406# define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
3ded5eb0 1407
0852beac 1408# ifdef EBCDIC
74665a89
KW
1409 /* The below is accurate for the 3 EBCDIC code pages traditionally
1410 * supported by perl. The only difference between them in the controls
1411 * is the position of \n, and that is represented symbolically below */
1412# define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1413 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1414 || (c) == '\t' || (c) == '\v' \
182c4ace 1415 || inRANGE((c), 1, 3) /* SOH, STX, ETX */ \
74665a89 1416 || (c) == 7 /* U+7F DEL */ \
182c4ace
KW
1417 || inRANGE((c), 0x0E, 0x13) /* SO SI DLE \
1418 DC[1-3] */ \
74665a89
KW
1419 || (c) == 0x18 /* U+18 CAN */ \
1420 || (c) == 0x19 /* U+19 EOM */ \
182c4ace 1421 || inRANGE((c), 0x1C, 0x1F) /* [FGRU]S */ \
74665a89
KW
1422 || (c) == 0x26 /* U+17 ETB */ \
1423 || (c) == 0x27 /* U+1B ESC */ \
1424 || (c) == 0x2D /* U+05 ENQ */ \
1425 || (c) == 0x2E /* U+06 ACK */ \
1426 || (c) == 0x32 /* U+16 SYN */ \
1427 || (c) == 0x37 /* U+04 EOT */ \
1428 || (c) == 0x3C /* U+14 DC4 */ \
1429 || (c) == 0x3D /* U+15 NAK */ \
1430 || (c) == 0x3F)/* U+1A SUB */
0852beac 1431# define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
74665a89
KW
1432# else /* isASCII is already defined for ASCII platforms, so can use that to
1433 define isCNTRL */
1434# define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
0852beac
KW
1435# endif
1436
3ffc8c70 1437 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
caa94d35
KW
1438 * during debugging, and it seems best to keep them. We may be called
1439 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1440 * do anything anyway, so make it not a problem */
1441# if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1442# define NATIVE_TO_LATIN1(ch) (ch)
1443# endif
3ded5eb0
KW
1444# define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1445# define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1446# define isBLANK_L1(c) (isBLANK_A(c) \
1447 || (FITS_IN_8_BITS(c) \
1448 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1449# define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1450# define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1451# define isLOWER_L1(c) (isLOWER_A(c) \
1452 || (FITS_IN_8_BITS(c) \
ae683a5f 1453 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
3ded5eb0
KW
1454 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1455 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1456 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1457 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1458# define isPRINT_L1(c) (isPRINT_A(c) \
1459 || (FITS_IN_8_BITS(c) \
1460 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
3ded5eb0
KW
1461# define isPUNCT_L1(c) (isPUNCT_A(c) \
1462 || (FITS_IN_8_BITS(c) \
ae683a5f 1463 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
3ded5eb0
KW
1464 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1465 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1466 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1467 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1468 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1469 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1470# define isSPACE_L1(c) (isSPACE_A(c) \
1471 || (FITS_IN_8_BITS(c) \
ae683a5f 1472 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
3ded5eb0
KW
1473 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1474# define isUPPER_L1(c) (isUPPER_A(c) \
1475 || (FITS_IN_8_BITS(c) \
182c4ace
KW
1476 && ( IN_RANGE(NATIVE_TO_LATIN1((U8) c), \
1477 0xC0, 0xDE) \
3ded5eb0
KW
1478 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1479# define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1480# define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1481# define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1482 || isBLANK_L1(c) \
1483 || (c) == '-' \
1484 || (c) == '(' \
1485 || (c) == ')')
1486 /* The following are not fully accurate in the above-ASCII range. I (khw)
1487 * don't think it's necessary to be so for the purposes where this gets
1488 * compiled */
1489# define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1490# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1491
1492 /* And these aren't accurate at all. They are useful only for above
1493 * Latin1, which utilities and bootstrapping don't deal with */
1494# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
6838b41e 1495# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
3ded5eb0
KW
1496# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1497
1498 /* Many of the macros later in this file are defined in terms of these. By
1499 * implementing them with a function, which converts the class number into
1500 * a call to the desired macro, all of the later ones work. However, that
1501 * function won't be actually defined when building a utility program (no
1502 * perl.h), and so a compiler error will be generated if one is attempted
1503 * to be used. And the above-Latin1 code points require Unicode tables to
1504 * be present, something unlikely to be the case when bootstrapping */
1505# define _generic_isCC(c, classnum) \
1506 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1507# define _generic_isCC_A(c, classnum) \
1508 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
687c8d01 1509#endif /* End of no perl.h H_PERL */
8a58bdcf 1510
e66b99e9
KW
1511#define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1512#define isALPHA(c) isALPHA_A(c)
0852beac
KW
1513#define isASCII_A(c) isASCII(c)
1514#define isASCII_L1(c) isASCII(c)
e66b99e9
KW
1515#define isBLANK(c) isBLANK_A(c)
1516#define isCNTRL(c) isCNTRL_A(c)
1517#define isDIGIT(c) isDIGIT_A(c)
1518#define isGRAPH(c) isGRAPH_A(c)
1519#define isIDFIRST(c) isIDFIRST_A(c)
1520#define isLOWER(c) isLOWER_A(c)
1521#define isPRINT(c) isPRINT_A(c)
779cf272 1522#define isPSXSPC_A(c) isSPACE_A(c)
e66b99e9 1523#define isPSXSPC(c) isPSXSPC_A(c)
779cf272 1524#define isPSXSPC_L1(c) isSPACE_L1(c)
e66b99e9
KW
1525#define isPUNCT(c) isPUNCT_A(c)
1526#define isSPACE(c) isSPACE_A(c)
1527#define isUPPER(c) isUPPER_A(c)
1528#define isWORDCHAR(c) isWORDCHAR_A(c)
1529#define isXDIGIT(c) isXDIGIT_A(c)
1530
1531/* ASCII casing. These could also be written as
1532 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1533 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1534 which uses table lookup and mask instead of subtraction. (This would
c5e9991e
KW
1535 work because the _MOD does not apply in the ASCII range).
1536
1537 These actually are UTF-8 invariant casing, not just ASCII, as any non-ASCII
1538 UTF-8 invariants are neither upper nor lower. (Only on EBCDIC platforms are
1539 there non-ASCII invariants, and all of them are controls.) */
68067e4e
DM
1540#define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1541#define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
bbce6d69 1542
25200305
KW
1543/* In the ASCII range, these are equivalent to what they're here defined to be.
1544 * But by creating these definitions, other code doesn't have to be aware of
c5e9991e
KW
1545 * this detail. Actually this works for all UTF-8 invariants, not just the
1546 * ASCII range. (EBCDIC platforms can have non-ASCII invariants.) */
25200305 1547#define toFOLD(c) toLOWER(c)
25200305
KW
1548#define toTITLE(c) toUPPER(c)
1549
c753c8d3
KW
1550#define toLOWER_A(c) toLOWER(c)
1551#define toUPPER_A(c) toUPPER(c)
25200305
KW
1552#define toFOLD_A(c) toFOLD(c)
1553#define toTITLE_A(c) toTITLE(c)
1a0901db 1554
4650c663 1555/* Use table lookup for speed; returns the input itself if is out-of-range */
b2bf251f 1556#define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1557 ? (c) \
f4cd282c 1558 : PL_latin1_lc[ (U8) (c) ])
c753c8d3
KW
1559#define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1560
1a0901db 1561/* Modified uc. Is correct uc except for three non-ascii chars which are
4650c663
KW
1562 * all mapped to one of them, and these need special handling; returns the
1563 * input itself if is out-of-range */
b2bf251f 1564#define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1565 ? (c) \
f4cd282c 1566 : PL_mod_latin1_uc[ (U8) (c) ])
31f05a37 1567#define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
84061b6a 1568
beab9ebe
KW
1569/* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1570
1571/* For internal core Perl use only: the base macro for defining macros like
1572 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
31f05a37
KW
1573 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1574 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1575 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1576 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1577 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1578 * code points under locale rules, use the macros based on _generic_LC_uvchr
1579 * instead of this. */
beab9ebe
KW
1580#define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1581 (! FITS_IN_8_BITS(c) \
1582 ? 0 \
31f05a37
KW
1583 : IN_UTF8_CTYPE_LOCALE \
1584 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
beab9ebe
KW
1585 : cBOOL(non_utf8))
1586
1587/* For internal core Perl use only: a helper macro for defining macros like
1588 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1589 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1590 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1591 * _generic_LC_base for more info */
1592#define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1593 _generic_LC_base(c,utf8_locale_classnum, \
1594 non_utf8_func( (_LC_CAST) (c)))
1595
1596/* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1597 * 'c' is the platform's native underscore character */
1598#define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1599 _generic_LC_base(c, utf8_locale_classnum, \
1600 (non_utf8_func( (_LC_CAST) (c)) \
1601 || (char)(c) == '_'))
1602
1603/* These next three are also for internal core Perl use only: case-change
247985d4
KW
1604 * helper macros. The reason for using the PL_latin arrays is in case the
1605 * system function is defective; it ensures uniform results that conform to the
b257a28c 1606 * Unicod standard. It does not handle the anomalies in UTF-8 Turkic locales */
beab9ebe
KW
1607#define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1608 ? (c) \
31f05a37
KW
1609 : (IN_UTF8_CTYPE_LOCALE) \
1610 ? PL_latin1_lc[ (U8) (c) ] \
5a10328c 1611 : (cast)function((cast)(c)))
beab9ebe 1612
31f05a37
KW
1613/* Note that the result can be larger than a byte in a UTF-8 locale. It
1614 * returns a single value, so can't adequately return the upper case of LATIN
1615 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1616 * values "SS"); instead it asserts against that under DEBUGGING, and
b257a28c
KW
1617 * otherwise returns its input. It does not handle the anomalies in UTF-8
1618 * Turkic locales. */
beab9ebe
KW
1619#define _generic_toUPPER_LC(c, function, cast) \
1620 (! FITS_IN_8_BITS(c) \
1621 ? (c) \
31f05a37 1622 : ((! IN_UTF8_CTYPE_LOCALE) \
b7d90381 1623 ? (cast)function((cast)(c)) \
31f05a37
KW
1624 : ((((U8)(c)) == MICRO_SIGN) \
1625 ? GREEK_CAPITAL_LETTER_MU \
1626 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1627 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1628 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1629 ? (__ASSERT_(0) (c)) \
1630 : PL_mod_latin1_uc[ (U8) (c) ])))))
1631
1632/* Note that the result can be larger than a byte in a UTF-8 locale. It
1633 * returns a single value, so can't adequately return the fold case of LATIN
1634 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1635 * values "ss"); instead it asserts against that under DEBUGGING, and
b257a28c
KW
1636 * otherwise returns its input. It does not handle the anomalies in UTF-8
1637 * Turkic locales */
beab9ebe 1638#define _generic_toFOLD_LC(c, function, cast) \
4d7db1b9 1639 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
31f05a37 1640 ? GREEK_SMALL_LETTER_MU \
4d7db1b9
KW
1641 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1642 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1643 _generic_toLOWER_LC(c, function, cast)))
beab9ebe 1644
84061b6a 1645/* Use the libc versions for these if available. */
f05550c0 1646#if defined(HAS_ISASCII)
84061b6a
KW
1647# define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1648#else
1649# define isASCII_LC(c) isASCII(c)
1650#endif
1651
f05550c0 1652#if defined(HAS_ISBLANK)
84061b6a 1653# define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
95f34b6f 1654#else /* Unlike isASCII, varies if in a UTF-8 locale */
278b2b56 1655# define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
84061b6a
KW
1656#endif
1657
f05550c0 1658#define _LC_CAST U8
bbce6d69 1659
f05550c0 1660#ifdef WIN32
375f5f06
KW
1661 /* The Windows functions don't bother to follow the POSIX standard, which
1662 * for example says that something can't both be a printable and a control.
1663 * But Windows treats the \t control as a printable, and does such things
1664 * as making superscripts into both digits and punctuation. This tames
1665 * these flaws by assuming that the definitions of both controls and space
1666 * are correct, and then making sure that other definitions don't have
1667 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1668 * Not all possible weirdnesses are checked for, just the ones that were
1669 * detected on actual Microsoft code pages */
1670
b7d90381
KW
1671# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1672# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1673
1674# define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1675 && isALPHANUMERIC_LC(c))
1676# define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1677 ! isPUNCT_LC(c))
1678# define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1679 isALPHANUMERIC_LC(c))
1680# define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1681# define isIDFIRST_LC(c) (((c) == '_') \
1682 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1683# define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1684# define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1685# define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1686# define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
f05550c0 1687# define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
b7d90381
KW
1688# define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1689 && isALPHANUMERIC_LC(c))
f05550c0
BF
1690
1691# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1692# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1693# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1694
1695#elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
4650c663 1696 /* For most other platforms */
beab9ebe 1697
f05550c0
BF
1698# define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1699# define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1700# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1701# define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1702# define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1703# define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1704# define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1705# define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1706# define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1707# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1708# define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1709# define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1710# define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1711
1712
1713# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1714# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1715# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1716
1717#else /* The final fallback position */
1718
b7d90381
KW
1719# define isALPHA_LC(c) (isascii(c) && isalpha(c))
1720# define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1721# define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1722# define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1723# define isGRAPH_LC(c) (isascii(c) && isgraph(c))
f05550c0 1724# define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
b7d90381
KW
1725# define isLOWER_LC(c) (isascii(c) && islower(c))
1726# define isPRINT_LC(c) (isascii(c) && isprint(c))
1727# define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1728# define isSPACE_LC(c) (isascii(c) && isspace(c))
1729# define isUPPER_LC(c) (isascii(c) && isupper(c))
f05550c0 1730# define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
b7d90381 1731# define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
f05550c0
BF
1732
1733# define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1734# define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1735# define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
bbce6d69 1736
f05550c0 1737#endif
55204971 1738
eba68aa0
KW
1739#define isIDCONT(c) isWORDCHAR(c)
1740#define isIDCONT_A(c) isWORDCHAR_A(c)
1741#define isIDCONT_L1(c) isWORDCHAR_L1(c)
1742#define isIDCONT_LC(c) isWORDCHAR_LC(c)
13380643 1743#define isPSXSPC_LC(c) isSPACE_LC(c)
aaa51d5e 1744
4650c663 1745/* For internal core Perl use only: the base macros for defining macros like
d0da05db
KW
1746 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1747 * number defined earlier in this file. _generic_uvchr() is used for POSIX
4650c663
KW
1748 * classes where there is a macro or function 'above_latin1' that takes the
1749 * single argument 'c' and returns the desired value. These exist for those
1750 * classes which have simple definitions, avoiding the overhead of a hash
d0da05db 1751 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
4650c663 1752 * for classes where that overhead is faster than a direct lookup.
d0da05db 1753 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
4650c663
KW
1754 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1755 * have duplicate checks here, so could create versions of the macros that
1756 * don't, but experiments show that gcc optimizes them out anyway. */
66c17564
KW
1757
1758/* Note that all ignore 'use bytes' */
1e222e4f
KW
1759#define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1760 ? _generic_isCC(c, classnum) \
cd500f2f 1761 : above_latin1(c))
1e222e4f
KW
1762#define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1763 ? _generic_isCC(c, classnum) \
922e8cb4 1764 : _is_uni_FOO(classnum, c))
d0da05db
KW
1765#define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1766#define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1767#define isASCII_uvchr(c) isASCII(c)
1768#define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1769#define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1770#define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1771#define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1e222e4f
KW
1772#define isIDCONT_uvchr(c) \
1773 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1774#define isIDFIRST_uvchr(c) \
1775 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
d0da05db
KW
1776#define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1777#define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1778
1779#define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1780#define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1781#define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1782
1783#define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1784#define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1785#define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1786#define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1787
1788#define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1789#define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1790#define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1791#define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1792
1793/* For backwards compatibility, even though '_uni' should mean official Unicode
1794 * code points, in Perl it means native for those below 256 */
1795#define isALPHA_uni(c) isALPHA_uvchr(c)
1796#define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1797#define isASCII_uni(c) isASCII_uvchr(c)
1798#define isBLANK_uni(c) isBLANK_uvchr(c)
1799#define isCNTRL_uni(c) isCNTRL_uvchr(c)
1800#define isDIGIT_uni(c) isDIGIT_uvchr(c)
1801#define isGRAPH_uni(c) isGRAPH_uvchr(c)
1802#define isIDCONT_uni(c) isIDCONT_uvchr(c)
1803#define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1804#define isLOWER_uni(c) isLOWER_uvchr(c)
1805#define isPRINT_uni(c) isPRINT_uvchr(c)
1806#define isPUNCT_uni(c) isPUNCT_uvchr(c)
1807#define isSPACE_uni(c) isSPACE_uvchr(c)
1808#define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1809#define isUPPER_uni(c) isUPPER_uvchr(c)
1810#define isVERTWS_uni(c) isVERTWS_uvchr(c)
1811#define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1812#define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1813#define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1814#define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1815#define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1816#define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
a0ed51b3 1817
4650c663
KW
1818/* For internal core Perl use only: the base macros for defining macros like
1819 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
d0da05db 1820 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
4650c663 1821 * simple class definitions; the other for more complex. These are like
d0da05db 1822 * _generic_uvchr, so see it for more info. */
cd500f2f
KW
1823#define _generic_LC_uvchr(latin1, above_latin1, c) \
1824 (c < 256 ? latin1(c) : above_latin1(c))
1825#define _generic_LC_swash_uvchr(latin1, classnum, c) \
1826 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1827
1828#define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1829#define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1830 _CC_ALPHANUMERIC, c)
b7d90381
KW
1831#define isASCII_LC_uvchr(c) isASCII_LC(c)
1832#define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1833 is_HORIZWS_cp_high, c)
feeab5a9 1834#define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
cd500f2f
KW
1835#define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1836#define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
b7d90381 1837#define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
eba68aa0 1838 _is_uni_perl_idcont, c)
b7d90381 1839#define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
cd500f2f
KW
1840 _is_uni_perl_idstart, c)
1841#define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1842#define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
b7d90381 1843#define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
cd500f2f 1844#define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
b7d90381 1845#define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
509fb054 1846 is_XPERLSPACE_cp_high, c)
cd500f2f 1847#define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
b7d90381 1848#define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
cd500f2f 1849 _CC_WORDCHAR, c)
b7d90381
KW
1850#define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1851 is_XDIGIT_cp_high, c)
e712593e 1852
b7d90381 1853#define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
aaa51d5e 1854
4650c663
KW
1855/* For internal core Perl use only: the base macros for defining macros like
1856 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1857 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1858 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1859 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1860 * same representation whether utf8 or not. Note that it assumes that the utf8
1861 * has been validated, and ignores 'use bytes' */
34aeb2e9
KW
1862#define _base_generic_utf8(enum_name, name, p, use_locale ) \
1863 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1864 (const U8 *) p, \
1865 "is" STRINGIFY(name) "_utf8", \
1866 "is" STRINGIFY(name) "_utf8_safe", \
1867 1, use_locale, __FILE__,__LINE__)
1868
1869#define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
da8c1a98
KW
1870
1871/* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1872 * they don't otherwise go out of their way to look for malformed UTF-8. If
1873 * they can return accurate results without knowing if the input is otherwise
1874 * malformed, they do so. For example isASCII is accurate in spite of any
1875 * non-length malformations because it looks only at a single byte. Likewise
1876 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1877 * variant ones return FALSE. But, if the input has to be well-formed in order
1878 * for the results to be accurate, the macros will test and if malformed will
1879 * call a routine to die
1880 *
1881 * Except for toke.c, the macros do assume that e > p, asserting that on
1882 * DEBUGGING builds. Much code that calls these depends on this being true,
1883 * for other reasons. toke.c is treated specially as using the regular
1884 * assertion breaks it in many ways. All strings that these operate on there
1885 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1886 * bunch of code in toke.c assumes that this is true, so the assertion allows
1887 * for that */
1888#ifdef PERL_IN_TOKE_C
1889# define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1890#else
1891# define _utf8_safe_assert(p,e) ((e) > (p))
1892#endif
1893
1894#define _generic_utf8_safe(classnum, p, e, above_latin1) \
1895 (__ASSERT_(_utf8_safe_assert(p, e)) \
1896 (UTF8_IS_INVARIANT(*(p))) \
1897 ? _generic_isCC(*(p), classnum) \
1898 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1899 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1900 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1901 classnum) \
1902 : (_force_out_malformed_utf8_message( \
1903 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1904 : above_latin1))
b7d90381
KW
1905/* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1906 * 'above_latin1' can be a macro */
da8c1a98
KW
1907#define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1908 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1909#define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1910 _generic_utf8_safe(classnum, p, e, \
1911 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1912 ? (_force_out_malformed_utf8_message( \
1913 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1914 : above_latin1(p)))
4650c663 1915/* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
cd500f2f 1916 * 'above_latin1' parameter */
da8c1a98
KW
1917#define _generic_swash_utf8_safe(classnum, p, e) \
1918_generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
922e8cb4 1919
cc8ab7c0 1920/* Like the above, but should be used only when it is known that there are no
ff7ecfc3
KW
1921 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1922 * class is TRUE for. Hence it can skip the tests for this range.
1923 * 'above_latin1' should include its arguments */
da8c1a98
KW
1924#define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1925 (__ASSERT_(_utf8_safe_assert(p, e)) \
1926 (UTF8_IS_INVARIANT(*(p))) \
1927 ? _generic_isCC(*(p), classnum) \
1928 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1929 ? 0 /* Note that doesn't check validity for latin1 */ \
1930 : above_latin1)
1931
84238efa 1932
34aeb2e9
KW
1933#define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1934#define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1935#define isASCII_utf8(p) _generic_utf8(ASCII, p)
1936#define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1937#define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1938#define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1939#define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1940#define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1941#define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1942#define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1943#define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1944#define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1945#define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1946#define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1947#define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1948#define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1949#define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1950#define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
e8fa43e2 1951
da8c1a98
KW
1952#define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1953#define isALPHANUMERIC_utf8_safe(p, e) \
1954 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1955#define isASCII_utf8_safe(p, e) \
1956 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1957 * works */ \
1958 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1959#define isBLANK_utf8_safe(p, e) \
1960 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1961
e8fa43e2
KW
1962#ifdef EBCDIC
1963 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1964 * more efficient macro instead of the more general one */
da8c1a98 1965# define isCNTRL_utf8_safe(p, e) \
56d02b8c 1966 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
e8fa43e2 1967#else
da8c1a98 1968# define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
e8fa43e2
KW
1969#endif
1970
da8c1a98
KW
1971#define isDIGIT_utf8_safe(p, e) \
1972 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1973 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1974#define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1975#define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1976 _is_utf8_perl_idcont_with_len, p, e)
e5dcd934 1977
c11ff943
KW
1978/* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1979 * IDFIRST is an alnum. See
c8362b00 1980 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
f91dcd13
KW
1981 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1982 * This used to be not the XID version, but we decided to go with the more
1983 * modern Unicode definition */
da8c1a98
KW
1984#define isIDFIRST_utf8_safe(p, e) \
1985 _generic_func_utf8_safe(_CC_IDFIRST, \
1986 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1987
1988#define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1989#define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1990#define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1991#define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1992#define isSPACE_utf8_safe(p, e) \
1993 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1994#define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1995#define isVERTWS_utf8_safe(p, e) \
1996 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1997#define isWORDCHAR_utf8_safe(p, e) \
1998 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1999#define isXDIGIT_utf8_safe(p, e) \
2000 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
2001 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2002 ? (_force_out_malformed_utf8_message( \
2003 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2004 : is_XDIGIT_high(p)))
a0ed51b3 2005
7d2b61dd 2006#define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
509fb054
KW
2007#define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
2008#define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
2009#define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
2e8adce6 2010
567b353c 2011/* For internal core use only, subject to change */
607313a1
KW
2012#define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
2013#define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
2014#define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
2015#define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
a1a5ec35
KW
2016
2017#define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
2018#define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
2019#define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
2020#define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
567b353c 2021
4650c663
KW
2022/* For internal core Perl use only: the base macros for defining macros like
2023 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
2024 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
2025 * 'macro' parameter */
34aeb2e9
KW
2026#define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
2027
2028#define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
2029#define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
2030#define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
2031#define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
2032#define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
2033#define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
2034#define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
2035#define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
2036#define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
2037#define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
2038#define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
2039#define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
2040#define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
2041#define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
2042#define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
2043#define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
2044#define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
2045
da8c1a98
KW
2046/* For internal core Perl use only: the base macros for defining macros like
2047 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
2048 * point in 'p' is within the 0-255 range, it uses locale rules from the
2049 * passed-in 'macro' parameter */
2050#define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2051 (__ASSERT_(_utf8_safe_assert(p, e)) \
2052 (UTF8_IS_INVARIANT(*(p))) \
2053 ? macro(*(p)) \
2054 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2055 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2056 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2057 : (_force_out_malformed_utf8_message( \
2058 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2059 : above_latin1))
2060
2061#define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2062 _generic_LC_utf8_safe(macro, p, e, \
2063 _is_utf8_FOO_with_len(classnum, p, e))
2064
2065#define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2066 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2067
2068#define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2069 _generic_LC_utf8_safe(classnum, p, e, \
2070 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2071 ? (_force_out_malformed_utf8_message( \
2072 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2073 : above_latin1(p)))
2074
2075#define isALPHANUMERIC_LC_utf8_safe(p, e) \
2076 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2077 _CC_ALPHANUMERIC, p, e)
2078#define isALPHA_LC_utf8_safe(p, e) \
2079 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2080#define isASCII_LC_utf8_safe(p, e) \
2081 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2082#define isBLANK_LC_utf8_safe(p, e) \
2083 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2084#define isCNTRL_LC_utf8_safe(p, e) \
2085 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2086#define isDIGIT_LC_utf8_safe(p, e) \
2087 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2088#define isGRAPH_LC_utf8_safe(p, e) \
2089 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2090#define isIDCONT_LC_utf8_safe(p, e) \
2091 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2092 _is_utf8_perl_idcont_with_len, p, e)
2093#define isIDFIRST_LC_utf8_safe(p, e) \
2094 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2095 _is_utf8_perl_idstart_with_len, p, e)
2096#define isLOWER_LC_utf8_safe(p, e) \
2097 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2098#define isPRINT_LC_utf8_safe(p, e) \
2099 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2100#define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2101#define isPUNCT_LC_utf8_safe(p, e) \
2102 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2103#define isSPACE_LC_utf8_safe(p, e) \
2104 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2105#define isUPPER_LC_utf8_safe(p, e) \
2106 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2107#define isWORDCHAR_LC_utf8_safe(p, e) \
2108 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2109#define isXDIGIT_LC_utf8_safe(p, e) \
2110 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
aaa51d5e 2111
fbc19f27
KW
2112/* Macros for backwards compatibility and for completeness when the ASCII and
2113 * Latin1 values are identical */
b7d90381
KW
2114#define isALPHAU(c) isALPHA_L1(c)
2115#define isDIGIT_L1(c) isDIGIT_A(c)
2116#define isOCTAL(c) isOCTAL_A(c)
2117#define isOCTAL_L1(c) isOCTAL_A(c)
2118#define isXDIGIT_L1(c) isXDIGIT_A(c)
2119#define isALNUM(c) isWORDCHAR(c)
a377c856 2120#define isALNUM_A(c) isALNUM(c)
b7d90381
KW
2121#define isALNUMU(c) isWORDCHAR_L1(c)
2122#define isALNUM_LC(c) isWORDCHAR_LC(c)
2123#define isALNUM_uni(c) isWORDCHAR_uni(c)
2e28f0b9 2124#define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
b7d90381
KW
2125#define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2126#define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2127#define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2128#define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2129#define isALNUMC(c) isALPHANUMERIC(c)
2130#define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2131#define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
15861f94 2132#define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
b7d90381 2133#define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
15861f94 2134#define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
fbc19f27 2135
2bd1cbf6
KW
2136/* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2137 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2138 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2139 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2140 * the outlier from the block that contains the other controls, just like
2141 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2142 * block. If it weren't special cased, it would yield a non-control.
88794300
KW
2143 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2144 * etc. */
2bd1cbf6 2145#ifndef EBCDIC
75763b3a 2146# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2bd1cbf6 2147#else
75763b3a
KW
2148# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2149 ((isPRINT_A(c)) \
2150 ? (UNLIKELY((c) == '?') \
2151 ? QUESTION_MARK_CTRL \
2152 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2153 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2154 ? '?' \
2155 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2bd1cbf6 2156#endif
bbce6d69 2157
dea28490
JJ
2158/* Line numbers are unsigned, 32 bits. */
2159typedef U32 line_t;
e5dcd934 2160#define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
378cc40b 2161
91152fc1
DG
2162/* Helpful alias for version prescan */
2163#define is_LAX_VERSION(a,b) \
2164 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2165
2166#define is_STRICT_VERSION(a,b) \
2167 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2168
2169#define BADVERSION(a,b,c) \
2170 if (b) { \
2171 *b = c; \
2172 } \
2173 return a;
8c52afec 2174
cb27eebd
KW
2175/* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2176 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2177 * which is advanced. The input is validated only by an assert() in DEBUGGING
2178 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2179 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2180#define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2181 ? (c) \
2182 : ((c) + 9))))
902008b8
KW
2183#define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2184 ? (*(s)++) \
2185 : (*(s)++ + 9))))
95a59cab 2186
cb27eebd
KW
2187/* Converts a character known to represent an octal digit (0-7) to its numeric
2188 * value. The input is validated only by an assert() in DEBUGGING builds. In
2189 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2190#define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2191
305b8651
KW
2192/* Efficiently returns a boolean as to if two native characters are equivalent
2193 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2194 * the ALPHA in the name is to remind you of that. This is asserted() in
2195 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2196 * works (on valid input) for both non- and UTF-8-encoded bytes.
2197 *
2198 * When one of the inputs is a compile-time constant and gets folded by the
2199 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2200 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2201 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
96ca48da
KW
2202 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2203 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2204 * just a single 0, in the bit position where the upper- and lowercase differ.
2205 * */
305b8651
KW
2206#define isALPHA_FOLD_EQ(c1, c2) \
2207 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2208 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2209#define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2210
8e84507e 2211/*
ccfc67b7
JH
2212=head1 Memory Management
2213
a02a5408 2214=for apidoc Am|void|Newx|void* ptr|int nitems|type
954c1994
GS
2215The XSUB-writer's interface to the C C<malloc> function.
2216
596f7718 2217Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2218
c5008215
JC
2219In 5.9.3, Newx() and friends replace the older New() API, and drops
2220the first parameter, I<x>, a debug aid which allowed callers to identify
37b8b4c9 2221themselves. This aid has been superseded by a new build option,
d10b4965 2222PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
c5008215
JC
2223there for use in XS modules supporting older perls.
2224
a02a5408 2225=for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
954c1994 2226The XSUB-writer's interface to the C C<malloc> function, with
fbe13c60 2227cast. See also C<L</Newx>>.
954c1994 2228
596f7718 2229Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2230
a02a5408 2231=for apidoc Am|void|Newxz|void* ptr|int nitems|type
954c1994 2232The XSUB-writer's interface to the C C<malloc> function. The allocated
fbe13c60 2233memory is zeroed with C<memzero>. See also C<L</Newx>>.
a02a5408 2234
596f7718 2235Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2236
954c1994
GS
2237=for apidoc Am|void|Renew|void* ptr|int nitems|type
2238The XSUB-writer's interface to the C C<realloc> function.
2239
596f7718 2240Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2241
954c1994
GS
2242=for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2243The XSUB-writer's interface to the C C<realloc> function, with
2244cast.
2245
596f7718 2246Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2247
49b8b560 2248=for apidoc Am|void|Safefree|void* ptr
954c1994
GS
2249The XSUB-writer's interface to the C C<free> function.
2250
596f7718 2251This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
0d7b2759 2252
954c1994
GS
2253=for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2254The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
926bb54c 2255source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2256C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
954c1994 2257
e90e2364 2258=for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
796b6530 2259Like C<Move> but returns C<dest>. Useful
72d33970 2260for encouraging compilers to tail-call
e90e2364
NC
2261optimise.
2262
954c1994
GS
2263=for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2264The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
926bb54c 2265source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2266C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
954c1994 2267
e90e2364
NC
2268=for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2269
796b6530 2270Like C<Copy> but returns C<dest>. Useful
72d33970 2271for encouraging compilers to tail-call
e90e2364
NC
2272optimise.
2273
954c1994
GS
2274=for apidoc Am|void|Zero|void* dest|int nitems|type
2275
2276The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2277destination, C<nitems> is the number of items, and C<type> is the type.
2278
e90e2364
NC
2279=for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2280
72d33970
FC
2281Like C<Zero> but returns dest. Useful
2282for encouraging compilers to tail-call
e90e2364
NC
2283optimise.
2284
da5d8dbb 2285=for apidoc Am|void|StructCopy|type *src|type *dest|type
4375e838 2286This is an architecture-independent macro to copy one structure to another.
954c1994 2287
7e337ee0
JH
2288=for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2289
2290Fill up memory with a byte pattern (a byte repeated over and over
2291again) that hopefully catches attempts to access uninitialized memory.
2292
2293=for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2294
2295PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2296
1c12ffb4 2297=for apidoc Am|void|PoisonFree|void* dest|int nitems|type
7e337ee0
JH
2298
2299PoisonWith(0xEF) for catching access to freed memory.
2300
9965345d
JH
2301=for apidoc Am|void|Poison|void* dest|int nitems|type
2302
7e337ee0 2303PoisonWith(0xEF) for catching access to freed memory.
9965345d
JH
2304
2305=cut */
954c1994 2306
561b68a9
SH
2307/* Maintained for backwards-compatibility only. Use newSV() instead. */
2308#ifndef PERL_CORE
ff06c60c 2309#define NEWSV(x,len) newSV(len)
561b68a9 2310#endif
ff06c60c 2311
b7112dce 2312#define MEM_SIZE_MAX ((MEM_SIZE)-1)
19a94d75 2313
a500027b 2314#define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
e6bdf523 2315
27d5b266 2316#ifdef PERL_MALLOC_WRAP
e6bdf523
DM
2317
2318/* This expression will be constant-folded at compile time. It checks
2319 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2320 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2321 * As well as avoiding the need for a run-time check in some cases, it's
2322 * designed to avoid compiler warnings like:
2323 * comparison is always false due to limited range of data type
73e8ff00
DM
2324 * It's mathematically equivalent to
2325 * max(n) * sizeof(t) > MEM_SIZE_MAX
e6bdf523
DM
2326 */
2327
2328# define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
445198b9
LM
2329 ( sizeof(MEM_SIZE) < sizeof(n) \
2330 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
e6bdf523 2331
88f9f128 2332/* This is written in a slightly odd way to avoid various spurious
d98e5cde
DM
2333 * compiler warnings. We *want* to write the expression as
2334 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2335 * (for some compile-time constant C), but even when the LHS
2336 * constant-folds to false at compile-time, g++ insists on emitting
2337 * warnings about the RHS (e.g. "comparison is always false"), so instead
2338 * we write it as
e6bdf523 2339 *
d98e5cde 2340 * (cond ? n : X) > C
88f9f128 2341 *
d98e5cde
DM
2342 * where X is a constant with X > C always false. Choosing a value for X
2343 * is tricky. If 0, some compilers will complain about 0 > C always being
2344 * false; if 1, Coverity complains when n happens to be the constant value
2345 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2346 * for X and hope that nothing else whines.
e6bdf523
DM
2347 */
2348
2349# define _MEM_WRAP_WILL_WRAP(n,t) \
88f9f128
DM
2350 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2351 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
e6bdf523
DM
2352
2353# define MEM_WRAP_CHECK(n,t) \
2354 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2355 && (croak_memory_wrap(),0))
2356
2357# define MEM_WRAP_CHECK_1(n,t,a) \
2358 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2359 && (Perl_croak_nocontext("%s",(a)),0))
2360
814eedc8
DD
2361/* "a" arg must be a string literal */
2362# define MEM_WRAP_CHECK_s(n,t,a) \
2363 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2364 && (Perl_croak_nocontext("" a ""),0))
2365
8b44ba4c 2366#define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
27d5b266 2367
a500027b 2368#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))
27d5b266
JH
2369#else
2370
410319be
NC
2371#define MEM_WRAP_CHECK(n,t)
2372#define MEM_WRAP_CHECK_1(n,t,a)
814eedc8 2373#define MEM_WRAP_CHECK_s(n,t,a)
8b44ba4c
NC
2374#define MEM_WRAP_CHECK_(n,t)
2375
a500027b 2376#define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
27d5b266 2377
1936d2a7 2378#endif
8b44ba4c 2379
fe4f188c 2380#ifdef PERL_MEM_LOG
46c6c7e2 2381/*
9f653bb5 2382 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
46c6c7e2
JH
2383 * go through functions, which are handy for debugging breakpoints, but
2384 * which more importantly get the immediate calling environment (file and
e352bcff
JH
2385 * line number, and C function name if available) passed in. This info can
2386 * then be used for logging the calls, for which one gets a sample
73d1d973 2387 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
3609ea0d 2388 *
46c6c7e2 2389 * Known problems:
94e892a6 2390 * - not all memory allocs get logged, only those
46c6c7e2 2391 * that go through Newx() and derivatives (while all
94e892a6 2392 * Safefrees do get logged)
46c6c7e2
JH
2393 * - __FILE__ and __LINE__ do not work everywhere
2394 * - __func__ or __FUNCTION__ even less so
2395 * - I think more goes on after the perlio frees but
2396 * the thing is that STDERR gets closed (as do all
2397 * the file descriptors)
2398 * - no deeper calling stack than the caller of the Newx()
2399 * or the kind, but do I look like a C reflection/introspection
2400 * utility to you?
2401 * - the function prototypes for the logging functions
2402 * probably should maybe be somewhere else than handy.h
2403 * - one could consider inlining (macrofying) the logging
2404 * for speed, but I am too lazy
2405 * - one could imagine recording the allocations in a hash,
2406 * (keyed by the allocation address?), and maintain that
2407 * through reallocs and frees, but how to do that without
2408 * any News() happening...?
73d1d973 2409 * - lots of -Ddefines to get useful/controllable output
b953482e 2410 * - lots of ENV reads
46c6c7e2
JH
2411 */
2412
0b0ab801 2413# ifdef PERL_CORE
73d1d973 2414# ifndef PERL_MEM_LOG_NOIMPL
0b0ab801
MHM
2415enum mem_log_type {
2416 MLT_ALLOC,
2417 MLT_REALLOC,
d7a2c63c
MHM
2418 MLT_FREE,
2419 MLT_NEW_SV,
2420 MLT_DEL_SV
0b0ab801
MHM
2421};
2422# endif
12754f92 2423# if defined(PERL_IN_SV_C) /* those are only used in sv.c */
d7a2c63c
MHM
2424void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2425void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
12754f92 2426# endif
0b0ab801
MHM
2427# endif
2428
fe4f188c
JH
2429#endif
2430
2431#ifdef PERL_MEM_LOG
d1401ee9
MHM
2432#define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2433#define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
46c6c7e2 2434#define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
fe4f188c
JH
2435#endif
2436
2437#ifndef MEM_LOG_ALLOC
2438#define MEM_LOG_ALLOC(n,t,a) (a)
2439#endif
2440#ifndef MEM_LOG_REALLOC
2441#define MEM_LOG_REALLOC(n,t,v,a) (a)
2442#endif
2443#ifndef MEM_LOG_FREE
2444#define MEM_LOG_FREE(a) (a)
2445#endif
2446
d1401ee9
MHM
2447#define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2448#define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2449#define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
a6f6820f
NC
2450
2451#ifndef PERL_CORE
a02a5408
JC
2452/* pre 5.9.x compatibility */
2453#define New(x,v,n,t) Newx(v,n,t)
2454#define Newc(x,v,n,t,c) Newxc(v,n,t,c)
4541904d 2455#define Newz(x,v,n,t) Newxz(v,n,t)
a6f6820f 2456#endif
a02a5408 2457
ff68c719 2458#define Renew(v,n,t) \
d1401ee9 2459 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
ff68c719 2460#define Renewc(v,n,t,c) \
d1401ee9 2461 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
94010e71
NC
2462
2463#ifdef PERL_POISON
2464#define Safefree(d) \
06c0cc96 2465 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
94010e71 2466#else
fe4f188c 2467#define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
94010e71 2468#endif
55497cff 2469
dbb57106
YO
2470/* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2471 * as it handles cases like constant string arguments without throwing warnings *
2472 * the cast is required, as is the inequality check, to avoid warnings */
45908e4d 2473#define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
55497cff 2474
45908e4d
YO
2475
2476#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)))
2477#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)))
2478#define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2479
bdd1531d 2480/* Like above, but returns a pointer to 'd' */
45908e4d
YO
2481#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)))
2482#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)))
45908e4d 2483#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
e90e2364 2484
7e337ee0
JH
2485#define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2486#define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2487#define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2488#define Poison(d,n,t) PoisonFree(d,n,t)
27d5b266 2489
caa674f3
DD
2490#ifdef PERL_POISON
2491# define PERL_POISON_EXPR(x) x
2492#else
2493# define PERL_POISON_EXPR(x)
2494#endif
2495
ff68c719 2496#define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2cc61e15 2497
1b7e2294
KW
2498/*
2499=head1 Handy Values
2500
2501=for apidoc Am|STRLEN|C_ARRAY_LENGTH|void *a
2502
2503Returns the number of elements in the input C array (so you want your
2504zero-based indices to be less than but not equal to).
2505
2506=for apidoc Am|void *|C_ARRAY_END|void *a
2507
2508Returns a pointer to one element past the final element of the input C array.
2509
2510=cut
2511
2512C_ARRAY_END is one past the last: half-open/half-closed range, not
2513last-inclusive range.
2514*/
622913ab 2515#define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
c3caa5c3 2516#define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
622913ab 2517
2cc61e15
DD
2518#ifdef NEED_VA_COPY
2519# ifdef va_copy
2520# define Perl_va_copy(s, d) va_copy(d, s)
07798b17
AC
2521# elif defined(__va_copy)
2522# define Perl_va_copy(s, d) __va_copy(d, s)
2cc61e15 2523# else
07798b17 2524# define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2cc61e15
DD
2525# endif
2526#endif
2527
472d47bc
SB
2528/* convenience debug macros */
2529#ifdef USE_ITHREADS
2530#define pTHX_FORMAT "Perl interpreter: 0x%p"
2531#define pTHX__FORMAT ", Perl interpreter: 0x%p"
f54cb97a
AL
2532#define pTHX_VALUE_ (void *)my_perl,
2533#define pTHX_VALUE (void *)my_perl
2534#define pTHX__VALUE_ ,(void *)my_perl,
2535#define pTHX__VALUE ,(void *)my_perl
472d47bc 2536#else
3609ea0d 2537#define pTHX_FORMAT
472d47bc 2538#define pTHX__FORMAT
3609ea0d 2539#define pTHX_VALUE_
472d47bc 2540#define pTHX_VALUE
3609ea0d 2541#define pTHX__VALUE_
472d47bc
SB
2542#define pTHX__VALUE
2543#endif /* USE_ITHREADS */
3609ea0d 2544
2acdbac1
NC
2545/* Perl_deprecate was not part of the public API, and did not have a deprecate()
2546 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2547 CPAN::Unpack show any users outside the core. */
2548#ifdef PERL_CORE
dc6e8de0
A
2549# define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2550 "Use of " s " is deprecated")
c9680906
A
2551# define deprecate_disappears_in(when,message) \
2552 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2553 message ", and will disappear in Perl " when)
ac641426
A
2554# define deprecate_fatal_in(when,message) \
2555 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2556 message ". Its use will be fatal in Perl " when)
2acdbac1
NC
2557#endif
2558
dfff4baf
BF
2559/* Internal macros to deal with gids and uids */
2560#ifdef PERL_CORE
2561
2562# if Uid_t_size > IVSIZE
2563# define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2564# define SvUID(sv) SvNV(sv)
07798b17
AC
2565# elif Uid_t_sign <= 0
2566# define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2567# define SvUID(sv) SvIV(sv)
dfff4baf 2568# else
07798b17
AC
2569# define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2570# define SvUID(sv) SvUV(sv)
dfff4baf
BF
2571# endif /* Uid_t_size */
2572
2573# if Gid_t_size > IVSIZE
2574# define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2575# define SvGID(sv) SvNV(sv)
07798b17
AC
2576# elif Gid_t_sign <= 0
2577# define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2578# define SvGID(sv) SvIV(sv)
dfff4baf 2579# else
07798b17
AC
2580# define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2581# define SvGID(sv) SvUV(sv)
dfff4baf
BF
2582# endif /* Gid_t_size */
2583
2584#endif
2585
6a5bc5ac 2586#endif /* PERL_HANDY_H_ */
9d745869 2587
e9a8c099 2588/*
14d04a33 2589 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 2590 */