3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * 'So that was the job I felt I had to do when I started,' thought Sam.
14 * [p.934 of _The Lord of the Rings_, VI/iii: "Mount Doom"]
17 /* This file contains some common functions needed to carry out certain
18 * ops. For example, both pp_sprintf() and pp_prtf() call the function
19 * do_sprintf() found in this file.
23 #define PERL_IN_DOOP_C
25 #include "invlist_inline.h"
30 /* Helper function for do_trans().
31 * Handles cases where the search and replacement charlists aren't UTF-8,
32 * aren't identical, and neither the /d nor /s flag is present.
34 * sv may or may not be utf8. Note that no code point above 255 can possibly
35 * be in the to-translate set
39 S_do_trans_simple(pTHX_ SV * const sv, const OPtrans_map * const tbl)
43 U8 *s = (U8*)SvPV_nomg(sv,len);
44 U8 * const send = s+len;
46 PERL_ARGS_ASSERT_DO_TRANS_SIMPLE;
47 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d: entering do_trans_simple:"
52 /* First, take care of non-UTF-8 input strings, because they're easy */
55 const short ch = tbl->map[*s];
65 const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS);
69 /* Allow for worst-case expansion: Each input byte can become 2. For a
70 * given input character, this happens when it occupies a single byte
71 * under UTF-8, but is to be translated to something that occupies two:
72 * $_="a".chr(400); tr/a/\xFE/, FE needs encoding. */
82 /* Need to check this, otherwise 128..255 won't match */
83 const UV c = utf8n_to_uvchr(s, send - s, &ulen, UTF8_ALLOW_DEFAULT);
84 if (c < 0x100 && (ch = tbl->map[c]) >= 0) {
86 d = uvchr_to_utf8(d, (UV)ch);
89 else { /* No match -> copy */
96 sv_setpvn(sv, (char*)dstart, d - dstart);
101 SvCUR_set(sv, d - dstart);
106 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d: returning %zu\n",
107 __FILE__, __LINE__, matches));
108 DEBUG_y(sv_dump(sv));
113 /* Helper function for do_trans().
114 * Handles cases where the search and replacement charlists are identical and
115 * non-utf8: so the string isn't modified, and only a count of modifiable
118 * Note that it doesn't handle /d or /s, since these modify the string even if
119 * the replacement list is empty.
121 * sv may or may not be utf8. Note that no code point above 255 can possibly
122 * be in the to-translate set
126 S_do_trans_count(pTHX_ SV * const sv, const OPtrans_map * const tbl)
129 const U8 *s = (const U8*)SvPV_nomg_const(sv, len);
130 const U8 * const send = s + len;
133 PERL_ARGS_ASSERT_DO_TRANS_COUNT;
135 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d: entering do_trans_count:"
137 __FILE__, __LINE__));
138 DEBUG_y(sv_dump(sv));
142 if (tbl->map[*s++] >= 0)
147 const bool complement = cBOOL(PL_op->op_private & OPpTRANS_COMPLEMENT);
150 const UV c = utf8n_to_uvchr(s, send - s, &ulen, UTF8_ALLOW_DEFAULT);
152 if (tbl->map[c] >= 0)
154 } else if (complement)
160 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d: count returning %zu\n",
161 __FILE__, __LINE__, matches));
166 /* Helper function for do_trans().
167 * Handles cases where the search and replacement charlists aren't identical
168 * and both are non-utf8, and one or both of /d, /s is specified.
170 * sv may or may not be utf8. Note that no code point above 255 can possibly
171 * be in the to-translate set
175 S_do_trans_complex(pTHX_ SV * const sv, const OPtrans_map * const tbl)
178 U8 *s = (U8*)SvPV_nomg(sv, len);
179 U8 * const send = s+len;
181 const bool complement = cBOOL(PL_op->op_private & OPpTRANS_COMPLEMENT);
183 PERL_ARGS_ASSERT_DO_TRANS_COMPLEX;
185 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d: entering do_trans_complex:"
187 __FILE__, __LINE__));
188 DEBUG_y(sv_dump(sv));
192 U8 * const dstart = d;
194 if (PL_op->op_private & OPpTRANS_SQUASH) {
196 /* What the mapping of the previous character was to. If the new
197 * character has the same mapping, it is squashed from the output
198 * (but still is included in the count) */
199 short previous_map = (short) TR_OOB;
202 const short this_map = tbl->map[*s];
205 if (this_map != previous_map) {
207 previous_map = this_map;
211 if (this_map == (short) TR_UNMAPPED) {
213 previous_map = (short) TR_OOB;
216 assert(this_map == (short) TR_DELETE);
224 else { /* Not to squash */
226 const short this_map = tbl->map[*s];
231 else if (this_map == (short) TR_UNMAPPED)
233 else if (this_map == (short) TR_DELETE)
239 SvCUR_set(sv, d - dstart);
242 const bool squash = cBOOL(PL_op->op_private & OPpTRANS_SQUASH);
243 const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS);
246 Size_t size = tbl->size;
248 /* What the mapping of the previous character was to. If the new
249 * character has the same mapping, it is squashed from the output (but
250 * still is included in the count) */
254 /* Allow for worst-case expansion: Each input byte can become 2.
255 * For a given input character, this happens when it occupies a
256 * single byte under UTF-8, but is to be translated to something
257 * that occupies two: */
258 Newx(d, len*2+1, U8);
265 const UV comp = utf8n_to_uvchr(s, send - s, &len,
273 ? (short) TR_UNMAPPED
280 if (LIKELY(!squash || ch != pch)) {
281 d = uvchr_to_utf8(d, ch);
287 else if (sch == (short) TR_UNMAPPED) {
292 else if (sch == (short) TR_DELETE)
295 assert(sch == (short) TR_R_EMPTY); /* empty replacement */
304 sv_setpvn(sv, (char*)dstart, d - dstart);
309 SvCUR_set(sv, d - dstart);
314 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d: returning %zu\n",
315 __FILE__, __LINE__, matches));
316 DEBUG_y(sv_dump(sv));
321 /* Helper function for do_trans().
322 * Handles cases where an inversion map implementation is to be used and the
323 * search and replacement charlists are identical: so the string isn't
324 * modified, and only a count of modifiable chars is needed.
326 * Note that it doesn't handle /d nor /s, since these modify the string
327 * even if the replacement charlist is empty.
329 * sv may or may not be utf8.
333 S_do_trans_count_invmap(pTHX_ SV * const sv, AV * const invmap)
339 SV** const from_invlist_ptr = av_fetch(invmap, 0, TRUE);
340 SV** const to_invmap_ptr = av_fetch(invmap, 1, TRUE);
341 SV* from_invlist = *from_invlist_ptr;
342 SV* to_invmap_sv = *to_invmap_ptr;
343 UV* map = (UV *) SvPVX(to_invmap_sv);
345 PERL_ARGS_ASSERT_DO_TRANS_COUNT_INVMAP;
347 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d:"
348 "entering do_trans_count_invmap:"
350 __FILE__, __LINE__));
351 DEBUG_y(sv_dump(sv));
352 DEBUG_y(PerlIO_printf(Perl_debug_log, "mapping:\n"));
353 DEBUG_y(invmap_dump(from_invlist, (UV *) SvPVX(to_invmap_sv)));
355 s = (U8*)SvPV_nomg(sv, len);
364 /* Get the code point of the next character in the string */
365 if (! SvUTF8(sv) || UTF8_IS_INVARIANT(*s)) {
370 from = utf8_to_uvchr_buf(s, send, &s_len);
371 if (from == 0 && *s != '\0') {
372 _force_out_malformed_utf8_message(s, send, 0, /*die*/TRUE);
376 /* Look the code point up in the data structure for this tr/// to get
378 i = _invlist_search(from_invlist, from);
381 if (map[i] != (UV) TR_UNLISTED) {
388 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d: returning %zu\n",
389 __FILE__, __LINE__, matches));
393 /* Helper function for do_trans().
394 * Handles cases where an inversion map implementation is to be used and the
395 * search and replacement charlists are either not identical or flags are
398 * sv may or may not be utf8.
402 S_do_trans_invmap(pTHX_ SV * const sv, AV * const invmap)
411 SV** const from_invlist_ptr = av_fetch(invmap, 0, TRUE);
412 SV** const to_invmap_ptr = av_fetch(invmap, 1, TRUE);
413 SV** const to_expansion_ptr = av_fetch(invmap, 2, TRUE);
414 NV max_expansion = SvNV(*to_expansion_ptr);
415 SV* from_invlist = *from_invlist_ptr;
416 SV* to_invmap_sv = *to_invmap_ptr;
417 UV* map = (UV *) SvPVX(to_invmap_sv);
418 UV previous_map = TR_OOB;
419 const bool squash = cBOOL(PL_op->op_private & OPpTRANS_SQUASH);
420 const bool delete_unfound = cBOOL(PL_op->op_private & OPpTRANS_DELETE);
421 bool inplace = ! cBOOL(PL_op->op_private & OPpTRANS_GROWS);
422 const UV* from_array = invlist_array(from_invlist);
423 UV final_map = TR_OOB;
424 bool out_is_utf8 = cBOOL(SvUTF8(sv));
427 PERL_ARGS_ASSERT_DO_TRANS_INVMAP;
429 /* A third element in the array indicates that the replacement list was
430 * shorter than the search list, and this element contains the value to use
431 * for the items that don't correspond */
432 if (av_top_index(invmap) >= 3) {
433 SV** const final_map_ptr = av_fetch(invmap, 3, TRUE);
434 SV* const final_map_sv = *final_map_ptr;
435 final_map = SvUV(final_map_sv);
438 /* If there is something in the transliteration that could force the input
439 * to be changed to UTF-8, we don't know if we can do it in place, so
441 if (! out_is_utf8 && (PL_op->op_private & OPpTRANS_CAN_FORCE_UTF8)) {
445 s = (U8*)SvPV_nomg(sv, len);
446 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d: entering do_trans_invmap:"
448 __FILE__, __LINE__));
449 DEBUG_y(sv_dump(sv));
450 DEBUG_y(PerlIO_printf(Perl_debug_log, "mapping:\n"));
451 DEBUG_y(invmap_dump(from_invlist, map));
456 /* We know by now if there are some possible input strings whose
457 * transliterations are longer than the input. If none can, we just edit
463 /* Here, we can't edit in place. We have no idea how much, if any,
464 * this particular input string will grow. However, the compilation
465 * calculated the maximum expansion possible. Use that to allocate
466 * based on the worst case scenario. (First +1 is to round up; 2nd is
468 Newx(d, (STRLEN) (len * max_expansion + 1 + 1), U8);
474 /* Do the actual transliteration */
481 /* Get the code point of the next character in the string */
482 if (! SvUTF8(sv) || UTF8_IS_INVARIANT(*s)) {
487 from = utf8_to_uvchr_buf(s, send, &s_len);
488 if (from == 0 && *s != '\0') {
489 _force_out_malformed_utf8_message(s, send, 0, /*die*/TRUE);
493 /* Look the code point up in the data structure for this tr/// to get
495 i = _invlist_search(from_invlist, from);
500 if (to == (UV) TR_UNLISTED) { /* Just copy the unreplaced character */
501 if (UVCHR_IS_INVARIANT(from) || ! out_is_utf8) {
504 else if (SvUTF8(sv)) {
505 Move(s, d, s_len, U8);
508 else { /* Convert to UTF-8 */
509 append_utf8_from_native_byte(*s, &d);
517 /* Everything else is counted as a match */
520 if (to == (UV) TR_SPECIAL_HANDLING) {
521 if (delete_unfound) {
526 /* Use the final character in the replacement list */
529 else { /* Here the input code point is to be remapped. The actual
530 value is offset from the base of this entry */
531 to += from - from_array[i];
534 /* If copying all occurrences, or this is the first occurrence, copy it
536 if (! squash || to != previous_map) {
538 d = uvchr_to_utf8(d, to);
541 if (to >= 256) { /* If need to convert to UTF-8, restart */
559 sv_setpvn(sv, (char*)d0, d - d0);
564 SvCUR_set(sv, d - d0);
567 if (! SvUTF8(sv) && out_is_utf8) {
572 DEBUG_y(PerlIO_printf(Perl_debug_log, "%s: %d: returning %zu\n",
573 __FILE__, __LINE__, matches));
574 DEBUG_y(sv_dump(sv));
578 /* Execute a tr//. sv is the value to be translated, while PL_op
579 * should be an OP_TRANS or OP_TRANSR op, whose op_pv field contains a
580 * translation table or whose op_sv field contains an inversion map.
582 * Returns a count of number of characters translated
586 Perl_do_trans(pTHX_ SV *sv)
589 const U8 flags = PL_op->op_private;
590 bool use_utf8_fcns = cBOOL(flags & OPpTRANS_USE_SVOP);
591 bool identical = cBOOL(flags & OPpTRANS_IDENTICAL);
593 PERL_ARGS_ASSERT_DO_TRANS;
595 if (SvREADONLY(sv) && ! identical) {
596 Perl_croak_no_modify();
598 (void)SvPV_const(sv, len);
602 if (!SvPOKp(sv) || SvTHINKFIRST(sv))
603 (void)SvPV_force_nomg(sv, len);
604 (void)SvPOK_only_UTF8(sv);
610 PAD_SVl(cPADOP->op_padix);
612 MUTABLE_SV(cSVOP->op_sv);
616 return do_trans_count_invmap(sv, (AV *) map);
619 return do_trans_invmap(sv, (AV *) map);
623 const OPtrans_map * const map = (OPtrans_map*)cPVOP->op_pv;
626 return do_trans_count(sv, map);
628 else if (flags & (OPpTRANS_SQUASH|OPpTRANS_DELETE|OPpTRANS_COMPLEMENT)) {
629 return do_trans_complex(sv, map);
632 return do_trans_simple(sv, map);
637 /* make it small to exercise the logic */
638 # define JOIN_DELIM_BUFSIZE 2
640 # define JOIN_DELIM_BUFSIZE 40
644 =for apidoc_section $string
647 This performs a Perl L<C<join>|perlfunc/join>, placing the joined output
650 The elements to join are in SVs, stored in a C array of pointers to SVs, from
651 C<**mark> to S<C<**sp - 1>>. Hence C<*mark> is a reference to the first SV.
652 Each SV will be coerced into a PV if not one already.
654 C<delim> contains the string (or coerced into a string) that is to separate
655 each of the joined elements.
657 If any component is in UTF-8, the result will be as well, and all non-UTF-8
658 components will be converted to UTF-8 as necessary.
660 Magic and tainting are handled.
666 Perl_do_join(pTHX_ SV *sv, SV *delim, SV **mark, SV **sp)
668 PERL_ARGS_ASSERT_DO_JOIN;
670 SV ** const oldmark = mark;
671 SSize_t items = sp - mark;
674 const char * delimpv = SvPV_const(delim, delimlen);
675 char delim_buf[JOIN_DELIM_BUFSIZE];
676 bool delim_do_utf8 = DO_UTF8(delim);
679 /* Make a copy of the delim, since G or A magic may modify the delim SV.
680 Use a local buffer if possible to avoid the cost of allocation and
683 if (delimlen <= JOIN_DELIM_BUFSIZE) {
684 Copy(delimpv, delim_buf, delimlen, char);
688 delimpv = savepvn(delimpv, delimlen);
694 len = (items > 0 ? (delimlen * (items - 1) ) : 0);
695 SvUPGRADE(sv, SVt_PV);
696 if (SvLEN(sv) < len + items) { /* current length is way too short */
697 while (items-- > 0) {
698 if (*mark && !SvGAMAGIC(*mark) && SvOK(*mark)) {
700 SvPV_const(*mark, tmplen);
705 SvGROW(sv, len + 1); /* so try to pre-extend */
713 /* sv_setpv retains old UTF8ness [perl #24846] */
716 if (TAINTING_get && SvMAGICAL(sv))
726 const U32 delimflag = delim_do_utf8 ? SV_CATUTF8 : SV_CATBYTES;
727 for (; items > 0; items--,mark++) {
730 sv_catpvn_flags(sv, delimpv, delimlen, delimflag);
731 s = SvPV_const(*mark,len);
732 sv_catpvn_flags(sv,s,len,
733 DO_UTF8(*mark) ? SV_CATUTF8 : SV_CATBYTES);
737 for (; items > 0; items--,mark++)
740 const char *s = SvPV_const(*mark,len);
741 sv_catpvn_flags(sv,s,len,
742 DO_UTF8(*mark) ? SV_CATUTF8 : SV_CATBYTES);
749 =for apidoc_section $string
750 =for apidoc do_sprintf
752 This performs a Perl L<C<sprintf>|perlfunc/sprintf> placing the string output
755 The elements to format are in SVs, stored in a C array of pointers to SVs of
756 length C<len>> and beginning at C<**sarg>. The element referenced by C<*sarg>
759 Magic and tainting are handled.
765 Perl_do_sprintf(pTHX_ SV *sv, SSize_t len, SV **sarg)
768 const char * const pat = SvPV_const(*sarg, patlen);
769 bool do_taint = FALSE;
771 PERL_ARGS_ASSERT_DO_SPRINTF;
774 if (SvTAINTED(*sarg))
776 (PL_op && PL_op->op_type < OP_max)
777 ? (PL_op->op_type == OP_PRTF)
779 : PL_op_name[PL_op->op_type]
785 sv_vsetpvfn(sv, pat, patlen, NULL, sarg + 1, (Size_t)(len - 1), &do_taint);
792 Perl_do_vecget(pTHX_ SV *sv, STRLEN offset, int size)
795 const I32 svpv_flags = ((PL_op->op_flags & OPf_MOD || LVRET)
796 ? SV_UNDEF_RETURNS_NULL : 0);
797 unsigned char *s = (unsigned char *)
798 SvPV_flags(sv, srclen, (svpv_flags|SV_GMAGIC));
802 s = (unsigned char *)"";
805 PERL_ARGS_ASSERT_DO_VECGET;
807 if (size < 1 || ! isPOWER_OF_2(size))
808 Perl_croak(aTHX_ "Illegal number of bits in vec");
811 if (Perl_sv_utf8_downgrade_flags(aTHX_ sv, TRUE, 0)) {
812 /* PVX may have changed */
813 s = (unsigned char *) SvPV_flags(sv, srclen, svpv_flags);
816 Perl_croak(aTHX_ "Use of strings with code points over 0xFF"
817 " as arguments to vec is forbidden");
822 STRLEN bitoffs = ((offset % 8) * size) % 8;
823 STRLEN uoffset = offset / (8 / size);
825 if (uoffset >= srclen)
828 retnum = (s[uoffset] >> bitoffs) & nBIT_MASK(size);
831 int n = size / 8; /* required number of bytes */
837 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
838 "Bit vector size > 32 non-portable");
841 if (offset > Size_t_MAX / n - 1) /* would overflow */
844 uoffset = offset * n;
846 /* Switch on the number of bytes available, but no more than the number
848 switch (MIN(n, (SSize_t) srclen - uoffset)) {
853 retnum += ((UV) s[uoffset + 7]);
856 retnum += ((UV) s[uoffset + 6] << 8); /* = size - 56 */
859 retnum += ((UV) s[uoffset + 5] << 16); /* = size - 48 */
862 retnum += ((UV) s[uoffset + 4] << 24); /* = size - 40 */
866 retnum += ((UV) s[uoffset + 3] << (size - 32));
869 retnum += ((UV) s[uoffset + 2] << (size - 24));
872 retnum += ((UV) s[uoffset + 1] << (size - 16));
875 retnum += ((UV) s[uoffset ] << (size - 8));
886 /* currently converts input to bytes if possible but doesn't sweat failures,
887 * although it does ensure that the string it clobbers is not marked as
888 * utf8-valid any more
891 Perl_do_vecset(pTHX_ SV *sv)
893 STRLEN offset, bitoffs = 0;
900 SV * const targ = LvTARG(sv);
901 char errflags = LvFLAGS(sv);
903 PERL_ARGS_ASSERT_DO_VECSET;
905 /* some out-of-range errors have been deferred if/until the LV is
906 * actually written to: f(vec($s,-1,8)) is not always fatal */
908 assert(!(errflags & ~(LVf_NEG_OFF|LVf_OUT_OF_RANGE)));
909 if (errflags & LVf_NEG_OFF)
910 Perl_croak_nocontext("Negative offset to vec in lvalue context");
911 Perl_croak_nocontext("Out of memory during vec in lvalue context");
916 s = (unsigned char*)SvPV_force_flags(targ, targlen,
917 SV_GMAGIC | SV_UNDEF_RETURNS_NULL);
919 /* This is handled by the SvPOK_only below...
920 if (!Perl_sv_utf8_downgrade_flags(aTHX_ targ, TRUE, 0))
923 (void) Perl_sv_utf8_downgrade_flags(aTHX_ targ, TRUE, 0);
926 (void)SvPOK_only(targ);
928 offset = LvTARGOFF(sv);
929 size = LvTARGLEN(sv);
931 if (size < 1 || (size & (size-1))) /* size < 1 or not a power of two */
932 Perl_croak(aTHX_ "Illegal number of bits in vec");
935 bitoffs = ((offset%8)*size)%8;
940 if (offset > Size_t_MAX / n - 1) /* would overflow */
941 Perl_croak_nocontext("Out of memory during vec in lvalue context");
945 len = (bitoffs + size + 7)/8; /* required number of bytes */
946 if (targlen < offset || targlen - offset < len) {
947 STRLEN newlen = offset > Size_t_MAX - len - 1 ? /* avoid overflow */
948 Size_t_MAX : offset + len + 1;
949 s = (unsigned char*)SvGROW(targ, newlen);
950 (void)memzero((char *)(s + targlen), newlen - targlen);
951 SvCUR_set(targ, newlen - 1);
955 mask = nBIT_MASK(size);
957 s[offset] &= ~(mask << bitoffs);
958 s[offset] |= lval << bitoffs;
965 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
966 "Bit vector size > 32 non-portable");
967 s[offset+7] = (U8)( lval ); /* = size - 64 */
968 s[offset+6] = (U8)( lval >> 8); /* = size - 56 */
969 s[offset+5] = (U8)( lval >> 16); /* = size - 48 */
970 s[offset+4] = (U8)( lval >> 24); /* = size - 40 */
974 s[offset+3] = (U8)( lval >> (size - 32));
975 s[offset+2] = (U8)( lval >> (size - 24));
978 s[offset+1] = (U8)( lval >> (size - 16));
981 s[offset ] = (U8)( lval >> (size - 8));
987 Perl_do_vop(pTHX_ I32 optype, SV *sv, SV *left, SV *right)
1002 bool result_needs_to_be_utf8 = FALSE;
1003 bool left_utf8 = FALSE;
1004 bool right_utf8 = FALSE;
1005 U8 * left_non_downgraded = NULL;
1006 U8 * right_non_downgraded = NULL;
1007 Size_t left_non_downgraded_len = 0;
1008 Size_t right_non_downgraded_len = 0;
1009 char * non_downgraded = NULL;
1010 Size_t non_downgraded_len = 0;
1012 PERL_ARGS_ASSERT_DO_VOP;
1014 if (sv != left || (optype != OP_BIT_AND && !SvOK(sv)))
1015 SvPVCLEAR(sv); /* avoid undef warning on |= and ^= */
1017 lc = SvPV_force_nomg(left, leftlen);
1020 lc = SvPV_nomg_const(left, leftlen);
1021 SvPV_force_nomg_nolen(sv);
1023 rc = SvPV_nomg_const(right, rightlen);
1025 /* This needs to come after SvPV to ensure that string overloading has
1028 /* Create downgraded temporaries of any UTF-8 encoded operands */
1029 if (DO_UTF8(left)) {
1030 const U8 * save_lc = (U8 *) lc;
1033 result_needs_to_be_utf8 = TRUE;
1035 left_non_downgraded_len = leftlen;
1036 lc = (char *) bytes_from_utf8_loc((const U8 *) lc, &leftlen,
1038 (const U8 **) &left_non_downgraded);
1039 /* Calculate the number of trailing unconvertible bytes. This quantity
1040 * is the original length minus the length of the converted portion. */
1041 left_non_downgraded_len -= left_non_downgraded - save_lc;
1044 if (DO_UTF8(right)) {
1045 const U8 * save_rc = (U8 *) rc;
1048 result_needs_to_be_utf8 = TRUE;
1050 right_non_downgraded_len = rightlen;
1051 rc = (char *) bytes_from_utf8_loc((const U8 *) rc, &rightlen,
1053 (const U8 **) &right_non_downgraded);
1054 right_non_downgraded_len -= right_non_downgraded - save_rc;
1058 /* We set 'len' to the length that the operation actually operates on. The
1059 * dangling part of the longer operand doesn't actually participate in the
1060 * operation. What happens is that we pretend that the shorter operand has
1061 * been extended to the right by enough imaginary zeros to match the length
1062 * of the longer one. But we know in advance the result of the operation
1063 * on zeros without having to do it. In the case of '&', the result is
1064 * zero, and the dangling portion is simply discarded. For '|' and '^', the
1065 * result is the same as the other operand, so the dangling part is just
1066 * appended to the final result, unchanged. As of perl-5.32, we no longer
1067 * accept above-FF code points in the dangling portion.
1069 if (left_utf8 || right_utf8) {
1070 Perl_croak(aTHX_ FATAL_ABOVE_FF_MSG, PL_op_desc[optype]);
1072 else { /* Neither is UTF-8 */
1073 len = MIN(leftlen, rightlen);
1080 (void)SvPOK_only(sv);
1081 if (SvOK(sv) || SvTYPE(sv) > SVt_PVMG) {
1082 dc = SvPV_force_nomg_nolen(sv);
1083 if (SvLEN(sv) < len + 1) {
1084 dc = SvGROW(sv, len + 1);
1085 (void)memzero(dc + SvCUR(sv), len - SvCUR(sv) + 1);
1089 needlen = optype == OP_BIT_AND
1090 ? len : (leftlen > rightlen ? leftlen : rightlen);
1091 Newxz(dc, needlen + 1, char);
1092 sv_usepvn_flags(sv, dc, needlen, SV_HAS_TRAILING_NUL);
1093 dc = SvPVX(sv); /* sv_usepvn() calls Renew() */
1097 if (len >= sizeof(long)*4 &&
1098 !(PTR2nat(dc) % sizeof(long)) &&
1099 !(PTR2nat(lc) % sizeof(long)) &&
1100 !(PTR2nat(rc) % sizeof(long))) /* It's almost always aligned... */
1102 const STRLEN remainder = len % (sizeof(long)*4);
1103 len /= (sizeof(long)*4);
1112 *dl++ = *ll++ & *rl++;
1113 *dl++ = *ll++ & *rl++;
1114 *dl++ = *ll++ & *rl++;
1115 *dl++ = *ll++ & *rl++;
1120 *dl++ = *ll++ ^ *rl++;
1121 *dl++ = *ll++ ^ *rl++;
1122 *dl++ = *ll++ ^ *rl++;
1123 *dl++ = *ll++ ^ *rl++;
1128 *dl++ = *ll++ | *rl++;
1129 *dl++ = *ll++ | *rl++;
1130 *dl++ = *ll++ | *rl++;
1131 *dl++ = *ll++ | *rl++;
1145 *dc++ = *lc++ & *rc++;
1150 *dc++ = *lc++ ^ *rc++;
1154 *dc++ = *lc++ | *rc++;
1157 if (rightlen > len) {
1159 SvCUR_set(sv, rightlen);
1161 sv_catpvn_nomg(sv, rsave + len, rightlen - len);
1163 else if (leftlen > len) {
1165 SvCUR_set(sv, leftlen);
1167 sv_catpvn_nomg(sv, lsave + len, leftlen - len);
1171 /* If there is trailing stuff that couldn't be converted from UTF-8, it
1172 * is appended as-is for the ^ and | operators. This preserves
1173 * backwards compatibility */
1174 if (right_non_downgraded) {
1175 non_downgraded = (char *) right_non_downgraded;
1176 non_downgraded_len = right_non_downgraded_len;
1178 else if (left_non_downgraded) {
1179 non_downgraded = (char *) left_non_downgraded;
1180 non_downgraded_len = left_non_downgraded_len;
1186 if (result_needs_to_be_utf8) {
1187 sv_utf8_upgrade_nomg(sv);
1189 /* Append any trailing UTF-8 as-is. */
1190 if (non_downgraded) {
1191 sv_catpvn_nomg(sv, non_downgraded, non_downgraded_len);
1199 /* Perl_do_kv() may be:
1200 * * called directly as the pp function for pp_keys() and pp_values();
1201 * * It may also be called directly when the op is OP_AVHVSWITCH, to
1202 * implement CORE::keys(), CORE::values().
1204 * In all cases it expects an HV on the stack and returns a list of keys,
1205 * values, or key-value pairs, depending on PL_op.
1210 HV * const keys = MUTABLE_HV(*PL_stack_sp);
1211 const U8 gimme = GIMME_V;
1213 const I32 dokeys = (PL_op->op_type == OP_KEYS)
1214 || ( PL_op->op_type == OP_AVHVSWITCH
1215 && (PL_op->op_private & OPpAVHVSWITCH_MASK)
1216 + OP_EACH == OP_KEYS);
1218 const I32 dovalues = (PL_op->op_type == OP_VALUES)
1219 || ( PL_op->op_type == OP_AVHVSWITCH
1220 && (PL_op->op_private & OPpAVHVSWITCH_MASK)
1221 + OP_EACH == OP_VALUES);
1223 assert( PL_op->op_type == OP_KEYS
1224 || PL_op->op_type == OP_VALUES
1225 || PL_op->op_type == OP_AVHVSWITCH);
1227 assert(!( PL_op->op_type == OP_VALUES
1228 && (PL_op->op_private & OPpMAYBE_LVSUB)));
1230 (void)hv_iterinit(keys); /* always reset iterator regardless */
1232 if (gimme == G_VOID) {
1237 if (gimme == G_SCALAR) {
1238 if (PL_op->op_flags & OPf_MOD || LVRET) { /* lvalue */
1239 SV * const ret = newSV_type_mortal(SVt_PVLV); /* Not TARG RT#67838 */
1240 sv_magic(ret, NULL, PERL_MAGIC_nkeys, NULL, 0);
1242 LvTARG(ret) = SvREFCNT_inc_simple(keys);
1243 rpp_replace_1_1(ret);
1249 /* note that in 'scalar(keys %h)' the OP_KEYS is usually
1250 * optimised away and the action is performed directly by the
1251 * padhv or rv2hv op. We now only get here via OP_AVHVSWITCH
1254 if (! SvTIED_mg((const SV *)keys, PERL_MAGIC_tied) ) {
1255 i = HvUSEDKEYS(keys);
1259 while (hv_iternext(keys)) i++;
1262 rpp_replace_1_1(targ);
1267 /* list context only here */
1269 if (UNLIKELY(PL_op->op_private & OPpMAYBE_LVSUB)) {
1270 const I32 flags = is_lvalue_sub();
1271 if (flags && !(flags & OPpENTERSUB_INARGS))
1272 /* diag_listed_as: Can't modify %s in %s */
1273 Perl_croak(aTHX_ "Can't modify keys in list assignment");
1276 /* push all keys and/or values onto stack */
1277 #ifdef PERL_RC_STACK
1278 SSize_t sp_base = PL_stack_sp - PL_stack_base;
1279 hv_pushkv(keys, (dokeys | (dovalues << 1)));
1280 /* Now safe to free the original arg on the stack and shuffle
1281 * down one place anything pushed on top of it */
1282 SSize_t nitems = PL_stack_sp - (PL_stack_base + sp_base);
1283 SV *old_sv = PL_stack_sp[-nitems];
1285 Move(PL_stack_sp - nitems + 1,
1286 PL_stack_sp - nitems, nitems, SV*);
1288 SvREFCNT_dec_NN(old_sv);
1291 hv_pushkv(keys, (dokeys | (dovalues << 1)));
1297 * ex: set ts=8 sts=4 sw=4 et: