3 perlguts - Introduction to the Perl API
7 This document attempts to describe how to use the Perl API, as well as
8 to provide some info on the basic workings of the Perl core. It is far
9 from complete and probably contains many errors. Please refer any
10 questions or comments to the author below.
16 Perl has three typedefs that handle Perl's three main data types:
22 Each typedef has specific routines that manipulate the various data types.
24 =head2 What is an "IV"?
26 Perl uses a special typedef IV which is a simple signed integer type that is
27 guaranteed to be large enough to hold a pointer (as well as an integer).
28 Additionally, there is the UV, which is simply an unsigned IV.
30 Perl also uses two special typedefs, I32 and I16, which will always be at
31 least 32-bits and 16-bits long, respectively. (Again, there are U32 and U16,
32 as well.) They will usually be exactly 32 and 16 bits long, but on Crays
33 they will both be 64 bits.
35 =head2 Working with SVs
37 An SV can be created and loaded with one command. There are five types of
38 values that can be loaded: an integer value (IV), an unsigned integer
39 value (UV), a double (NV), a string (PV), and another scalar (SV).
40 ("PV" stands for "Pointer Value". You might think that it is misnamed
41 because it is described as pointing only to strings. However, it is
42 possible to have it point to other things. For example, it could point
43 to an array of UVs. But,
44 using it for non-strings requires care, as the underlying assumption of
45 much of the internals is that PVs are just for strings. Often, for
46 example, a trailing C<NUL> is tacked on automatically. The non-string use
47 is documented only in this paragraph.)
49 The seven routines are:
54 SV* newSVpv(const char*, STRLEN);
55 SV* newSVpvn(const char*, STRLEN);
56 SV* newSVpvf(const char*, ...);
59 C<STRLEN> is an integer type (C<Size_t>, usually defined as C<size_t> in
60 F<config.h>) guaranteed to be large enough to represent the size of
61 any string that perl can handle.
63 In the unlikely case of a SV requiring more complex initialization, you
64 can create an empty SV with newSV(len). If C<len> is 0 an empty SV of
65 type NULL is returned, else an SV of type PV is returned with len + 1 (for
66 the C<NUL>) bytes of storage allocated, accessible via SvPVX. In both cases
67 the SV has the undef value.
69 SV *sv = newSV(0); /* no storage allocated */
70 SV *sv = newSV(10); /* 10 (+1) bytes of uninitialised storage
73 To change the value of an I<already-existing> SV, there are eight routines:
75 void sv_setiv(SV*, IV);
76 void sv_setuv(SV*, UV);
77 void sv_setnv(SV*, double);
78 void sv_setpv(SV*, const char*);
79 void sv_setpvn(SV*, const char*, STRLEN)
80 void sv_setpvf(SV*, const char*, ...);
81 void sv_vsetpvfn(SV*, const char*, STRLEN, va_list *,
82 SV **, Size_t, bool *);
83 void sv_setsv(SV*, SV*);
85 Notice that you can choose to specify the length of the string to be
86 assigned by using C<sv_setpvn>, C<newSVpvn>, or C<newSVpv>, or you may
87 allow Perl to calculate the length by using C<sv_setpv> or by specifying
88 0 as the second argument to C<newSVpv>. Be warned, though, that Perl will
89 determine the string's length by using C<strlen>, which depends on the
90 string terminating with a C<NUL> character, and not otherwise containing
93 The arguments of C<sv_setpvf> are processed like C<sprintf>, and the
94 formatted output becomes the value.
96 C<sv_vsetpvfn> is an analogue of C<vsprintf>, but it allows you to specify
97 either a pointer to a variable argument list or the address and length of
98 an array of SVs. The last argument points to a boolean; on return, if that
99 boolean is true, then locale-specific information has been used to format
100 the string, and the string's contents are therefore untrustworthy (see
101 L<perlsec>). This pointer may be NULL if that information is not
102 important. Note that this function requires you to specify the length of
105 The C<sv_set*()> functions are not generic enough to operate on values
106 that have "magic". See L</Magic Virtual Tables> later in this document.
108 All SVs that contain strings should be terminated with a C<NUL> character.
109 If it is not C<NUL>-terminated there is a risk of
110 core dumps and corruptions from code which passes the string to C
111 functions or system calls which expect a C<NUL>-terminated string.
112 Perl's own functions typically add a trailing C<NUL> for this reason.
113 Nevertheless, you should be very careful when you pass a string stored
114 in an SV to a C function or system call.
116 To access the actual value that an SV points to, you can use the macros:
121 SvPV(SV*, STRLEN len)
124 which will automatically coerce the actual scalar type into an IV, UV, double,
127 In the C<SvPV> macro, the length of the string returned is placed into the
128 variable C<len> (this is a macro, so you do I<not> use C<&len>). If you do
129 not care what the length of the data is, use the C<SvPV_nolen> macro.
130 Historically the C<SvPV> macro with the global variable C<PL_na> has been
131 used in this case. But that can be quite inefficient because C<PL_na> must
132 be accessed in thread-local storage in threaded Perl. In any case, remember
133 that Perl allows arbitrary strings of data that may both contain NULs and
134 might not be terminated by a C<NUL>.
136 Also remember that C doesn't allow you to safely say C<foo(SvPV(s, len),
137 len);>. It might work with your
138 compiler, but it won't work for everyone.
139 Break this sort of statement up into separate assignments:
147 If you want to know if the scalar value is TRUE, you can use:
151 Although Perl will automatically grow strings for you, if you need to force
152 Perl to allocate more memory for your SV, you can use the macro
154 SvGROW(SV*, STRLEN newlen)
156 which will determine if more memory needs to be allocated. If so, it will
157 call the function C<sv_grow>. Note that C<SvGROW> can only increase, not
158 decrease, the allocated memory of an SV and that it does not automatically
159 add space for the trailing C<NUL> byte (perl's own string functions typically do
160 C<SvGROW(sv, len + 1)>).
162 If you want to write to an existing SV's buffer and set its value to a
163 string, use SvPV_force() or one of its variants to force the SV to be
164 a PV. This will remove any of various types of non-stringness from
165 the SV while preserving the content of the SV in the PV. This can be
166 used, for example, to append data from an API function to a buffer
167 without extra copying:
169 (void)SvPVbyte_force(sv, len);
170 s = SvGROW(sv, len + needlen + 1);
171 /* something that modifies up to needlen bytes at s+len, but
172 modifies newlen bytes
173 eg. newlen = read(fd, s + len, needlen);
174 ignoring errors for these examples
176 s[len + newlen] = '\0';
177 SvCUR_set(sv, len + newlen);
181 If you already have the data in memory or if you want to keep your
182 code simple, you can use one of the sv_cat*() variants, such as
183 sv_catpvn(). If you want to insert anywhere in the string you can use
184 sv_insert() or sv_insert_flags().
186 If you don't need the existing content of the SV, you can avoid some
190 s = SvGROW(sv, needlen + 1);
191 /* something that modifies up to needlen bytes at s, but modifies
193 eg. newlen = read(fd, s. needlen);
196 SvCUR_set(sv, newlen);
197 SvPOK_only(sv); /* also clears SVf_UTF8 */
200 Again, if you already have the data in memory or want to avoid the
201 complexity of the above, you can use sv_setpvn().
203 If you have a buffer allocated with Newx() and want to set that as the
204 SV's value, you can use sv_usepvn_flags(). That has some requirements
205 if you want to avoid perl re-allocating the buffer to fit the trailing
208 Newx(buf, somesize+1, char);
209 /* ... fill in buf ... */
210 buf[somesize] = '\0';
211 sv_usepvn_flags(sv, buf, somesize, SV_SMAGIC | SV_HAS_TRAILING_NUL);
212 /* buf now belongs to perl, don't release it */
214 If you have an SV and want to know what kind of data Perl thinks is stored
215 in it, you can use the following macros to check the type of SV you have.
221 You can get and set the current length of the string stored in an SV with
222 the following macros:
225 SvCUR_set(SV*, I32 val)
227 You can also get a pointer to the end of the string stored in the SV
232 But note that these last three macros are valid only if C<SvPOK()> is true.
234 If you want to append something to the end of string stored in an C<SV*>,
235 you can use the following functions:
237 void sv_catpv(SV*, const char*);
238 void sv_catpvn(SV*, const char*, STRLEN);
239 void sv_catpvf(SV*, const char*, ...);
240 void sv_vcatpvfn(SV*, const char*, STRLEN, va_list *, SV **,
242 void sv_catsv(SV*, SV*);
244 The first function calculates the length of the string to be appended by
245 using C<strlen>. In the second, you specify the length of the string
246 yourself. The third function processes its arguments like C<sprintf> and
247 appends the formatted output. The fourth function works like C<vsprintf>.
248 You can specify the address and length of an array of SVs instead of the
249 va_list argument. The fifth function
250 extends the string stored in the first
251 SV with the string stored in the second SV. It also forces the second SV
252 to be interpreted as a string.
254 The C<sv_cat*()> functions are not generic enough to operate on values that
255 have "magic". See L</Magic Virtual Tables> later in this document.
257 If you know the name of a scalar variable, you can get a pointer to its SV
258 by using the following:
260 SV* get_sv("package::varname", 0);
262 This returns NULL if the variable does not exist.
264 If you want to know if this variable (or any other SV) is actually C<defined>,
269 The scalar C<undef> value is stored in an SV instance called C<PL_sv_undef>.
271 Its address can be used whenever an C<SV*> is needed. Make sure that
272 you don't try to compare a random sv with C<&PL_sv_undef>. For example
273 when interfacing Perl code, it'll work correctly for:
277 But won't work when called as:
282 So to repeat always use SvOK() to check whether an sv is defined.
284 Also you have to be careful when using C<&PL_sv_undef> as a value in
285 AVs or HVs (see L</AVs, HVs and undefined values>).
287 There are also the two values C<PL_sv_yes> and C<PL_sv_no>, which contain
288 boolean TRUE and FALSE values, respectively. Like C<PL_sv_undef>, their
289 addresses can be used whenever an C<SV*> is needed.
291 Do not be fooled into thinking that C<(SV *) 0> is the same as C<&PL_sv_undef>.
295 if (I-am-to-return-a-real-value) {
296 sv = sv_2mortal(newSViv(42));
300 This code tries to return a new SV (which contains the value 42) if it should
301 return a real value, or undef otherwise. Instead it has returned a NULL
302 pointer which, somewhere down the line, will cause a segmentation violation,
303 bus error, or just weird results. Change the zero to C<&PL_sv_undef> in the
304 first line and all will be well.
306 To free an SV that you've created, call C<SvREFCNT_dec(SV*)>. Normally this
307 call is not necessary (see L</Reference Counts and Mortality>).
311 Perl provides the function C<sv_chop> to efficiently remove characters
312 from the beginning of a string; you give it an SV and a pointer to
313 somewhere inside the PV, and it discards everything before the
314 pointer. The efficiency comes by means of a little hack: instead of
315 actually removing the characters, C<sv_chop> sets the flag C<OOK>
316 (offset OK) to signal to other functions that the offset hack is in
317 effect, and it moves the PV pointer (called C<SvPVX>) forward
318 by the number of bytes chopped off, and adjusts C<SvCUR> and C<SvLEN>
319 accordingly. (A portion of the space between the old and new PV
320 pointers is used to store the count of chopped bytes.)
322 Hence, at this point, the start of the buffer that we allocated lives
323 at C<SvPVX(sv) - SvIV(sv)> in memory and the PV pointer is pointing
324 into the middle of this allocated storage.
326 This is best demonstrated by example. Normally copy-on-write will prevent
327 the substitution from operator from using this hack, but if you can craft a
328 string for which copy-on-write is not possible, you can see it in play. In
329 the current implementation, the final byte of a string buffer is used as a
330 copy-on-write reference count. If the buffer is not big enough, then
331 copy-on-write is skipped. First have a look at an empty string:
333 % ./perl -Ilib -MDevel::Peek -le '$a=""; $a .= ""; Dump $a'
334 SV = PV(0x7ffb7c008a70) at 0x7ffb7c030390
337 PV = 0x7ffb7bc05b50 ""\0
341 Notice here the LEN is 10. (It may differ on your platform.) Extend the
342 length of the string to one less than 10, and do a substitution:
344 % ./perl -Ilib -MDevel::Peek -le '$a=""; $a.="123456789"; $a=~s/.//; \
346 SV = PV(0x7ffa04008a70) at 0x7ffa04030390
348 FLAGS = (POK,OOK,pPOK)
350 PV = 0x7ffa03c05b61 ( "\1" . ) "23456789"\0
354 Here the number of bytes chopped off (1) is shown next as the OFFSET. The
355 portion of the string between the "real" and the "fake" beginnings is
356 shown in parentheses, and the values of C<SvCUR> and C<SvLEN> reflect
357 the fake beginning, not the real one. (The first character of the string
358 buffer happens to have changed to "\1" here, not "1", because the current
359 implementation stores the offset count in the string buffer. This is
362 Something similar to the offset hack is performed on AVs to enable
363 efficient shifting and splicing off the beginning of the array; while
364 C<AvARRAY> points to the first element in the array that is visible from
365 Perl, C<AvALLOC> points to the real start of the C array. These are
366 usually the same, but a C<shift> operation can be carried out by
367 increasing C<AvARRAY> by one and decreasing C<AvFILL> and C<AvMAX>.
368 Again, the location of the real start of the C array only comes into
369 play when freeing the array. See C<av_shift> in F<av.c>.
371 =head2 What's Really Stored in an SV?
373 Recall that the usual method of determining the type of scalar you have is
374 to use C<Sv*OK> macros. Because a scalar can be both a number and a string,
375 usually these macros will always return TRUE and calling the C<Sv*V>
376 macros will do the appropriate conversion of string to integer/double or
377 integer/double to string.
379 If you I<really> need to know if you have an integer, double, or string
380 pointer in an SV, you can use the following three macros instead:
386 These will tell you if you truly have an integer, double, or string pointer
387 stored in your SV. The "p" stands for private.
389 There are various ways in which the private and public flags may differ.
390 For example, in perl 5.16 and earlier a tied SV may have a valid
391 underlying value in the IV slot (so SvIOKp is true), but the data
392 should be accessed via the FETCH routine rather than directly,
393 so SvIOK is false. (In perl 5.18 onwards, tied scalars use
394 the flags the same way as untied scalars.) Another is when
395 numeric conversion has occurred and precision has been lost: only the
396 private flag is set on 'lossy' values. So when an NV is converted to an
397 IV with loss, SvIOKp, SvNOKp and SvNOK will be set, while SvIOK wont be.
399 In general, though, it's best to use the C<Sv*V> macros.
401 =head2 Working with AVs
403 There are two ways to create and load an AV. The first method creates an
408 The second method both creates the AV and initially populates it with SVs:
410 AV* av_make(SSize_t num, SV **ptr);
412 The second argument points to an array containing C<num> C<SV*>'s. Once the
413 AV has been created, the SVs can be destroyed, if so desired.
415 Once the AV has been created, the following operations are possible on it:
417 void av_push(AV*, SV*);
420 void av_unshift(AV*, SSize_t num);
422 These should be familiar operations, with the exception of C<av_unshift>.
423 This routine adds C<num> elements at the front of the array with the C<undef>
424 value. You must then use C<av_store> (described below) to assign values
425 to these new elements.
427 Here are some other functions:
429 SSize_t av_top_index(AV*);
430 SV** av_fetch(AV*, SSize_t key, I32 lval);
431 SV** av_store(AV*, SSize_t key, SV* val);
433 The C<av_top_index> function returns the highest index value in an array (just
434 like $#array in Perl). If the array is empty, -1 is returned. The
435 C<av_fetch> function returns the value at index C<key>, but if C<lval>
436 is non-zero, then C<av_fetch> will store an undef value at that index.
437 The C<av_store> function stores the value C<val> at index C<key>, and does
438 not increment the reference count of C<val>. Thus the caller is responsible
439 for taking care of that, and if C<av_store> returns NULL, the caller will
440 have to decrement the reference count to avoid a memory leak. Note that
441 C<av_fetch> and C<av_store> both return C<SV**>'s, not C<SV*>'s as their
448 void av_extend(AV*, SSize_t key);
450 The C<av_clear> function deletes all the elements in the AV* array, but
451 does not actually delete the array itself. The C<av_undef> function will
452 delete all the elements in the array plus the array itself. The
453 C<av_extend> function extends the array so that it contains at least C<key+1>
454 elements. If C<key+1> is less than the currently allocated length of the array,
455 then nothing is done.
457 If you know the name of an array variable, you can get a pointer to its AV
458 by using the following:
460 AV* get_av("package::varname", 0);
462 This returns NULL if the variable does not exist.
464 See L</Understanding the Magic of Tied Hashes and Arrays> for more
465 information on how to use the array access functions on tied arrays.
467 =head2 Working with HVs
469 To create an HV, you use the following routine:
473 Once the HV has been created, the following operations are possible on it:
475 SV** hv_store(HV*, const char* key, U32 klen, SV* val, U32 hash);
476 SV** hv_fetch(HV*, const char* key, U32 klen, I32 lval);
478 The C<klen> parameter is the length of the key being passed in (Note that
479 you cannot pass 0 in as a value of C<klen> to tell Perl to measure the
480 length of the key). The C<val> argument contains the SV pointer to the
481 scalar being stored, and C<hash> is the precomputed hash value (zero if
482 you want C<hv_store> to calculate it for you). The C<lval> parameter
483 indicates whether this fetch is actually a part of a store operation, in
484 which case a new undefined value will be added to the HV with the supplied
485 key and C<hv_fetch> will return as if the value had already existed.
487 Remember that C<hv_store> and C<hv_fetch> return C<SV**>'s and not just
488 C<SV*>. To access the scalar value, you must first dereference the return
489 value. However, you should check to make sure that the return value is
490 not NULL before dereferencing it.
492 The first of these two functions checks if a hash table entry exists, and the
495 bool hv_exists(HV*, const char* key, U32 klen);
496 SV* hv_delete(HV*, const char* key, U32 klen, I32 flags);
498 If C<flags> does not include the C<G_DISCARD> flag then C<hv_delete> will
499 create and return a mortal copy of the deleted value.
501 And more miscellaneous functions:
506 Like their AV counterparts, C<hv_clear> deletes all the entries in the hash
507 table but does not actually delete the hash table. The C<hv_undef> deletes
508 both the entries and the hash table itself.
510 Perl keeps the actual data in a linked list of structures with a typedef of HE.
511 These contain the actual key and value pointers (plus extra administrative
512 overhead). The key is a string pointer; the value is an C<SV*>. However,
513 once you have an C<HE*>, to get the actual key and value, use the routines
516 I32 hv_iterinit(HV*);
517 /* Prepares starting point to traverse hash table */
518 HE* hv_iternext(HV*);
519 /* Get the next entry, and return a pointer to a
520 structure that has both the key and value */
521 char* hv_iterkey(HE* entry, I32* retlen);
522 /* Get the key from an HE structure and also return
523 the length of the key string */
524 SV* hv_iterval(HV*, HE* entry);
525 /* Return an SV pointer to the value of the HE
527 SV* hv_iternextsv(HV*, char** key, I32* retlen);
528 /* This convenience routine combines hv_iternext,
529 hv_iterkey, and hv_iterval. The key and retlen
530 arguments are return values for the key and its
531 length. The value is returned in the SV* argument */
533 If you know the name of a hash variable, you can get a pointer to its HV
534 by using the following:
536 HV* get_hv("package::varname", 0);
538 This returns NULL if the variable does not exist.
540 The hash algorithm is defined in the C<PERL_HASH> macro:
542 PERL_HASH(hash, key, klen)
544 The exact implementation of this macro varies by architecture and version
545 of perl, and the return value may change per invocation, so the value
546 is only valid for the duration of a single perl process.
548 See L</Understanding the Magic of Tied Hashes and Arrays> for more
549 information on how to use the hash access functions on tied hashes.
551 =for apidoc Amh|void|PERL_HASH|U32 hash|char *key|STRLEN klen
553 =head2 Hash API Extensions
555 Beginning with version 5.004, the following functions are also supported:
557 HE* hv_fetch_ent (HV* tb, SV* key, I32 lval, U32 hash);
558 HE* hv_store_ent (HV* tb, SV* key, SV* val, U32 hash);
560 bool hv_exists_ent (HV* tb, SV* key, U32 hash);
561 SV* hv_delete_ent (HV* tb, SV* key, I32 flags, U32 hash);
563 SV* hv_iterkeysv (HE* entry);
565 Note that these functions take C<SV*> keys, which simplifies writing
566 of extension code that deals with hash structures. These functions
567 also allow passing of C<SV*> keys to C<tie> functions without forcing
568 you to stringify the keys (unlike the previous set of functions).
570 They also return and accept whole hash entries (C<HE*>), making their
571 use more efficient (since the hash number for a particular string
572 doesn't have to be recomputed every time). See L<perlapi> for detailed
575 The following macros must always be used to access the contents of hash
576 entries. Note that the arguments to these macros must be simple
577 variables, since they may get evaluated more than once. See
578 L<perlapi> for detailed descriptions of these macros.
580 HePV(HE* he, STRLEN len)
584 HeSVKEY_force(HE* he)
585 HeSVKEY_set(HE* he, SV* sv)
587 These two lower level macros are defined, but must only be used when
588 dealing with keys that are not C<SV*>s:
593 Note that both C<hv_store> and C<hv_store_ent> do not increment the
594 reference count of the stored C<val>, which is the caller's responsibility.
595 If these functions return a NULL value, the caller will usually have to
596 decrement the reference count of C<val> to avoid a memory leak.
598 =head2 AVs, HVs and undefined values
600 Sometimes you have to store undefined values in AVs or HVs. Although
601 this may be a rare case, it can be tricky. That's because you're
602 used to using C<&PL_sv_undef> if you need an undefined SV.
604 For example, intuition tells you that this XS code:
607 av_store( av, 0, &PL_sv_undef );
609 is equivalent to this Perl code:
614 Unfortunately, this isn't true. In perl 5.18 and earlier, AVs use C<&PL_sv_undef> as a marker
615 for indicating that an array element has not yet been initialized.
616 Thus, C<exists $av[0]> would be true for the above Perl code, but
617 false for the array generated by the XS code. In perl 5.20, storing
618 &PL_sv_undef will create a read-only element, because the scalar
619 &PL_sv_undef itself is stored, not a copy.
621 Similar problems can occur when storing C<&PL_sv_undef> in HVs:
623 hv_store( hv, "key", 3, &PL_sv_undef, 0 );
625 This will indeed make the value C<undef>, but if you try to modify
626 the value of C<key>, you'll get the following error:
628 Modification of non-creatable hash value attempted
630 In perl 5.8.0, C<&PL_sv_undef> was also used to mark placeholders
631 in restricted hashes. This caused such hash entries not to appear
632 when iterating over the hash or when checking for the keys
633 with the C<hv_exists> function.
635 You can run into similar problems when you store C<&PL_sv_yes> or
636 C<&PL_sv_no> into AVs or HVs. Trying to modify such elements
637 will give you the following error:
639 Modification of a read-only value attempted
641 To make a long story short, you can use the special variables
642 C<&PL_sv_undef>, C<&PL_sv_yes> and C<&PL_sv_no> with AVs and
643 HVs, but you have to make sure you know what you're doing.
645 Generally, if you want to store an undefined value in an AV
646 or HV, you should not use C<&PL_sv_undef>, but rather create a
647 new undefined value using the C<newSV> function, for example:
649 av_store( av, 42, newSV(0) );
650 hv_store( hv, "foo", 3, newSV(0), 0 );
654 References are a special type of scalar that point to other data types
655 (including other references).
657 To create a reference, use either of the following functions:
659 SV* newRV_inc((SV*) thing);
660 SV* newRV_noinc((SV*) thing);
662 The C<thing> argument can be any of an C<SV*>, C<AV*>, or C<HV*>. The
663 functions are identical except that C<newRV_inc> increments the reference
664 count of the C<thing>, while C<newRV_noinc> does not. For historical
665 reasons, C<newRV> is a synonym for C<newRV_inc>.
667 Once you have a reference, you can use the following macro to dereference
672 then call the appropriate routines, casting the returned C<SV*> to either an
673 C<AV*> or C<HV*>, if required.
675 To determine if an SV is a reference, you can use the following macro:
679 To discover what type of value the reference refers to, use the following
680 macro and then check the return value.
684 The most useful types that will be returned are:
689 SVt_PVGV Glob (possibly a file handle)
691 Any numerical value returned which is less than SVt_PVAV will be a scalar
694 See L<perlapi/svtype> for more details.
696 =head2 Blessed References and Class Objects
698 References are also used to support object-oriented programming. In perl's
699 OO lexicon, an object is simply a reference that has been blessed into a
700 package (or class). Once blessed, the programmer may now use the reference
701 to access the various methods in the class.
703 A reference can be blessed into a package with the following function:
705 SV* sv_bless(SV* sv, HV* stash);
707 The C<sv> argument must be a reference value. The C<stash> argument
708 specifies which class the reference will belong to. See
709 L</Stashes and Globs> for information on converting class names into stashes.
711 /* Still under construction */
713 The following function upgrades rv to reference if not already one.
714 Creates a new SV for rv to point to. If C<classname> is non-null, the SV
715 is blessed into the specified class. SV is returned.
717 SV* newSVrv(SV* rv, const char* classname);
719 The following three functions copy integer, unsigned integer or double
720 into an SV whose reference is C<rv>. SV is blessed if C<classname> is
723 SV* sv_setref_iv(SV* rv, const char* classname, IV iv);
724 SV* sv_setref_uv(SV* rv, const char* classname, UV uv);
725 SV* sv_setref_nv(SV* rv, const char* classname, NV iv);
727 The following function copies the pointer value (I<the address, not the
728 string!>) into an SV whose reference is rv. SV is blessed if C<classname>
731 SV* sv_setref_pv(SV* rv, const char* classname, void* pv);
733 The following function copies a string into an SV whose reference is C<rv>.
734 Set length to 0 to let Perl calculate the string length. SV is blessed if
735 C<classname> is non-null.
737 SV* sv_setref_pvn(SV* rv, const char* classname, char* pv,
740 The following function tests whether the SV is blessed into the specified
741 class. It does not check inheritance relationships.
743 int sv_isa(SV* sv, const char* name);
745 The following function tests whether the SV is a reference to a blessed object.
747 int sv_isobject(SV* sv);
749 The following function tests whether the SV is derived from the specified
750 class. SV can be either a reference to a blessed object or a string
751 containing a class name. This is the function implementing the
752 C<UNIVERSAL::isa> functionality.
754 bool sv_derived_from(SV* sv, const char* name);
756 To check if you've got an object derived from a specific class you have
759 if (sv_isobject(sv) && sv_derived_from(sv, class)) { ... }
761 =head2 Creating New Variables
763 To create a new Perl variable with an undef value which can be accessed from
764 your Perl script, use the following routines, depending on the variable type.
766 SV* get_sv("package::varname", GV_ADD);
767 AV* get_av("package::varname", GV_ADD);
768 HV* get_hv("package::varname", GV_ADD);
770 Notice the use of GV_ADD as the second parameter. The new variable can now
771 be set, using the routines appropriate to the data type.
773 There are additional macros whose values may be bitwise OR'ed with the
774 C<GV_ADD> argument to enable certain extra features. Those bits are:
780 Marks the variable as multiply defined, thus preventing the:
782 Name <varname> used only once: possible typo
790 Had to create <varname> unexpectedly
792 if the variable did not exist before the function was called.
796 If you do not specify a package name, the variable is created in the current
799 =head2 Reference Counts and Mortality
801 Perl uses a reference count-driven garbage collection mechanism. SVs,
802 AVs, or HVs (xV for short in the following) start their life with a
803 reference count of 1. If the reference count of an xV ever drops to 0,
804 then it will be destroyed and its memory made available for reuse.
805 At the most basic internal level, reference counts can be manipulated
806 with the following macros:
808 int SvREFCNT(SV* sv);
809 SV* SvREFCNT_inc(SV* sv);
810 void SvREFCNT_dec(SV* sv);
812 (There are also suffixed versions of the increment and decrement macros,
813 for situations where the full generality of these basic macros can be
814 exchanged for some performance.)
816 However, the way a programmer should think about references is not so
817 much in terms of the bare reference count, but in terms of I<ownership>
818 of references. A reference to an xV can be owned by any of a variety
819 of entities: another xV, the Perl interpreter, an XS data structure,
820 a piece of running code, or a dynamic scope. An xV generally does not
821 know what entities own the references to it; it only knows how many
822 references there are, which is the reference count.
824 To correctly maintain reference counts, it is essential to keep track
825 of what references the XS code is manipulating. The programmer should
826 always know where a reference has come from and who owns it, and be
827 aware of any creation or destruction of references, and any transfers
828 of ownership. Because ownership isn't represented explicitly in the xV
829 data structures, only the reference count need be actually maintained
830 by the code, and that means that this understanding of ownership is not
831 actually evident in the code. For example, transferring ownership of a
832 reference from one owner to another doesn't change the reference count
833 at all, so may be achieved with no actual code. (The transferring code
834 doesn't touch the referenced object, but does need to ensure that the
835 former owner knows that it no longer owns the reference, and that the
836 new owner knows that it now does.)
838 An xV that is visible at the Perl level should not become unreferenced
839 and thus be destroyed. Normally, an object will only become unreferenced
840 when it is no longer visible, often by the same means that makes it
841 invisible. For example, a Perl reference value (RV) owns a reference to
842 its referent, so if the RV is overwritten that reference gets destroyed,
843 and the no-longer-reachable referent may be destroyed as a result.
845 Many functions have some kind of reference manipulation as
846 part of their purpose. Sometimes this is documented in terms
847 of ownership of references, and sometimes it is (less helpfully)
848 documented in terms of changes to reference counts. For example, the
849 L<newRV_inc()|perlapi/newRV_inc> function is documented to create a new RV
850 (with reference count 1) and increment the reference count of the referent
851 that was supplied by the caller. This is best understood as creating
852 a new reference to the referent, which is owned by the created RV,
853 and returning to the caller ownership of the sole reference to the RV.
854 The L<newRV_noinc()|perlapi/newRV_noinc> function instead does not
855 increment the reference count of the referent, but the RV nevertheless
856 ends up owning a reference to the referent. It is therefore implied
857 that the caller of C<newRV_noinc()> is relinquishing a reference to the
858 referent, making this conceptually a more complicated operation even
859 though it does less to the data structures.
861 For example, imagine you want to return a reference from an XSUB
862 function. Inside the XSUB routine, you create an SV which initially
863 has just a single reference, owned by the XSUB routine. This reference
864 needs to be disposed of before the routine is complete, otherwise it
865 will leak, preventing the SV from ever being destroyed. So to create
866 an RV referencing the SV, it is most convenient to pass the SV to
867 C<newRV_noinc()>, which consumes that reference. Now the XSUB routine
868 no longer owns a reference to the SV, but does own a reference to the RV,
869 which in turn owns a reference to the SV. The ownership of the reference
870 to the RV is then transferred by the process of returning the RV from
873 There are some convenience functions available that can help with the
874 destruction of xVs. These functions introduce the concept of "mortality".
875 Much documentation speaks of an xV itself being mortal, but this is
876 misleading. It is really I<a reference to> an xV that is mortal, and it
877 is possible for there to be more than one mortal reference to a single xV.
878 For a reference to be mortal means that it is owned by the temps stack,
879 one of perl's many internal stacks, which will destroy that reference
880 "a short time later". Usually the "short time later" is the end of
881 the current Perl statement. However, it gets more complicated around
882 dynamic scopes: there can be multiple sets of mortal references hanging
883 around at the same time, with different death dates. Internally, the
884 actual determinant for when mortal xV references are destroyed depends
885 on two macros, SAVETMPS and FREETMPS. See L<perlcall> and L<perlxs>
886 and L</Temporaries Stack> below for more details on these macros.
888 Mortal references are mainly used for xVs that are placed on perl's
889 main stack. The stack is problematic for reference tracking, because it
890 contains a lot of xV references, but doesn't own those references: they
891 are not counted. Currently, there are many bugs resulting from xVs being
892 destroyed while referenced by the stack, because the stack's uncounted
893 references aren't enough to keep the xVs alive. So when putting an
894 (uncounted) reference on the stack, it is vitally important to ensure that
895 there will be a counted reference to the same xV that will last at least
896 as long as the uncounted reference. But it's also important that that
897 counted reference be cleaned up at an appropriate time, and not unduly
898 prolong the xV's life. For there to be a mortal reference is often the
899 best way to satisfy this requirement, especially if the xV was created
900 especially to be put on the stack and would otherwise be unreferenced.
902 To create a mortal reference, use the functions:
905 SV* sv_mortalcopy(SV*)
908 C<sv_newmortal()> creates an SV (with the undefined value) whose sole
909 reference is mortal. C<sv_mortalcopy()> creates an xV whose value is a
910 copy of a supplied xV and whose sole reference is mortal. C<sv_2mortal()>
911 mortalises an existing xV reference: it transfers ownership of a reference
912 from the caller to the temps stack. Because C<sv_newmortal> gives the new
913 SV no value, it must normally be given one via C<sv_setpv>, C<sv_setiv>,
916 SV *tmp = sv_newmortal();
917 sv_setiv(tmp, an_integer);
919 As that is multiple C statements it is quite common so see this idiom instead:
921 SV *tmp = sv_2mortal(newSViv(an_integer));
923 The mortal routines are not just for SVs; AVs and HVs can be
924 made mortal by passing their address (type-casted to C<SV*>) to the
925 C<sv_2mortal> or C<sv_mortalcopy> routines.
927 =head2 Stashes and Globs
929 A B<stash> is a hash that contains all variables that are defined
930 within a package. Each key of the stash is a symbol
931 name (shared by all the different types of objects that have the same
932 name), and each value in the hash table is a GV (Glob Value). This GV
933 in turn contains references to the various objects of that name,
934 including (but not limited to) the following:
943 There is a single stash called C<PL_defstash> that holds the items that exist
944 in the C<main> package. To get at the items in other packages, append the
945 string "::" to the package name. The items in the C<Foo> package are in
946 the stash C<Foo::> in PL_defstash. The items in the C<Bar::Baz> package are
947 in the stash C<Baz::> in C<Bar::>'s stash.
949 To get the stash pointer for a particular package, use the function:
951 HV* gv_stashpv(const char* name, I32 flags)
952 HV* gv_stashsv(SV*, I32 flags)
954 The first function takes a literal string, the second uses the string stored
955 in the SV. Remember that a stash is just a hash table, so you get back an
956 C<HV*>. The C<flags> flag will create a new package if it is set to GV_ADD.
958 The name that C<gv_stash*v> wants is the name of the package whose symbol table
959 you want. The default package is called C<main>. If you have multiply nested
960 packages, pass their names to C<gv_stash*v>, separated by C<::> as in the Perl
963 Alternately, if you have an SV that is a blessed reference, you can find
964 out the stash pointer by using:
966 HV* SvSTASH(SvRV(SV*));
968 then use the following to get the package name itself:
970 char* HvNAME(HV* stash);
972 If you need to bless or re-bless an object you can use the following
975 SV* sv_bless(SV*, HV* stash)
977 where the first argument, an C<SV*>, must be a reference, and the second
978 argument is a stash. The returned C<SV*> can now be used in the same way
981 For more information on references and blessings, consult L<perlref>.
983 =head2 Double-Typed SVs
985 Scalar variables normally contain only one type of value, an integer,
986 double, pointer, or reference. Perl will automatically convert the
987 actual scalar data from the stored type into the requested type.
989 Some scalar variables contain more than one type of scalar data. For
990 example, the variable C<$!> contains either the numeric value of C<errno>
991 or its string equivalent from either C<strerror> or C<sys_errlist[]>.
993 To force multiple data values into an SV, you must do two things: use the
994 C<sv_set*v> routines to add the additional scalar type, then set a flag
995 so that Perl will believe it contains more than one type of data. The
996 four macros to set the flags are:
1003 The particular macro you must use depends on which C<sv_set*v> routine
1004 you called first. This is because every C<sv_set*v> routine turns on
1005 only the bit for the particular type of data being set, and turns off
1008 For example, to create a new Perl variable called "dberror" that contains
1009 both the numeric and descriptive string error values, you could use the
1013 extern char *dberror_list;
1015 SV* sv = get_sv("dberror", GV_ADD);
1016 sv_setiv(sv, (IV) dberror);
1017 sv_setpv(sv, dberror_list[dberror]);
1020 If the order of C<sv_setiv> and C<sv_setpv> had been reversed, then the
1021 macro C<SvPOK_on> would need to be called instead of C<SvIOK_on>.
1023 =head2 Read-Only Values
1025 In Perl 5.16 and earlier, copy-on-write (see the next section) shared a
1026 flag bit with read-only scalars. So the only way to test whether
1027 C<sv_setsv>, etc., will raise a "Modification of a read-only value" error
1028 in those versions is:
1030 SvREADONLY(sv) && !SvIsCOW(sv)
1032 Under Perl 5.18 and later, SvREADONLY only applies to read-only variables,
1033 and, under 5.20, copy-on-write scalars can also be read-only, so the above
1034 check is incorrect. You just want:
1038 If you need to do this check often, define your own macro like this:
1040 #if PERL_VERSION >= 18
1041 # define SvTRULYREADONLY(sv) SvREADONLY(sv)
1043 # define SvTRULYREADONLY(sv) (SvREADONLY(sv) && !SvIsCOW(sv))
1046 =head2 Copy on Write
1048 Perl implements a copy-on-write (COW) mechanism for scalars, in which
1049 string copies are not immediately made when requested, but are deferred
1050 until made necessary by one or the other scalar changing. This is mostly
1051 transparent, but one must take care not to modify string buffers that are
1052 shared by multiple SVs.
1054 You can test whether an SV is using copy-on-write with C<SvIsCOW(sv)>.
1056 You can force an SV to make its own copy of its string buffer by calling C<sv_force_normal(sv)> or SvPV_force_nolen(sv).
1058 If you want to make the SV drop its string buffer, use
1059 C<sv_force_normal_flags(sv, SV_COW_DROP_PV)> or simply
1060 C<sv_setsv(sv, NULL)>.
1062 All of these functions will croak on read-only scalars (see the previous
1063 section for more on those).
1065 To test that your code is behaving correctly and not modifying COW buffers,
1066 on systems that support L<mmap(2)> (i.e., Unix) you can configure perl with
1067 C<-Accflags=-DPERL_DEBUG_READONLY_COW> and it will turn buffer violations
1068 into crashes. You will find it to be marvellously slow, so you may want to
1069 skip perl's own tests.
1071 =head2 Magic Variables
1073 [This section still under construction. Ignore everything here. Post no
1074 bills. Everything not permitted is forbidden.]
1076 Any SV may be magical, that is, it has special features that a normal
1077 SV does not have. These features are stored in the SV structure in a
1078 linked list of C<struct magic>'s, typedef'ed to C<MAGIC>.
1081 MAGIC* mg_moremagic;
1091 Note this is current as of patchlevel 0, and could change at any time.
1093 =head2 Assigning Magic
1095 Perl adds magic to an SV using the sv_magic function:
1097 void sv_magic(SV* sv, SV* obj, int how, const char* name, I32 namlen);
1099 The C<sv> argument is a pointer to the SV that is to acquire a new magical
1102 If C<sv> is not already magical, Perl uses the C<SvUPGRADE> macro to
1103 convert C<sv> to type C<SVt_PVMG>.
1104 Perl then continues by adding new magic
1105 to the beginning of the linked list of magical features. Any prior entry
1106 of the same type of magic is deleted. Note that this can be overridden,
1107 and multiple instances of the same type of magic can be associated with an
1110 The C<name> and C<namlen> arguments are used to associate a string with
1111 the magic, typically the name of a variable. C<namlen> is stored in the
1112 C<mg_len> field and if C<name> is non-null then either a C<savepvn> copy of
1113 C<name> or C<name> itself is stored in the C<mg_ptr> field, depending on
1114 whether C<namlen> is greater than zero or equal to zero respectively. As a
1115 special case, if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
1116 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
1118 The sv_magic function uses C<how> to determine which, if any, predefined
1119 "Magic Virtual Table" should be assigned to the C<mg_virtual> field.
1120 See the L</Magic Virtual Tables> section below. The C<how> argument is also
1121 stored in the C<mg_type> field. The value of
1122 C<how> should be chosen from the set of macros
1123 C<PERL_MAGIC_foo> found in F<perl.h>. Note that before
1124 these macros were added, Perl internals used to directly use character
1125 literals, so you may occasionally come across old code or documentation
1126 referring to 'U' magic rather than C<PERL_MAGIC_uvar> for example.
1128 The C<obj> argument is stored in the C<mg_obj> field of the C<MAGIC>
1129 structure. If it is not the same as the C<sv> argument, the reference
1130 count of the C<obj> object is incremented. If it is the same, or if
1131 the C<how> argument is C<PERL_MAGIC_arylen>, C<PERL_MAGIC_regdatum>,
1132 C<PERL_MAGIC_regdata>, or if it is a NULL pointer, then C<obj> is merely
1133 stored, without the reference count being incremented.
1135 See also C<sv_magicext> in L<perlapi> for a more flexible way to add magic
1138 There is also a function to add magic to an C<HV>:
1140 void hv_magic(HV *hv, GV *gv, int how);
1142 This simply calls C<sv_magic> and coerces the C<gv> argument into an C<SV>.
1144 To remove the magic from an SV, call the function sv_unmagic:
1146 int sv_unmagic(SV *sv, int type);
1148 The C<type> argument should be equal to the C<how> value when the C<SV>
1149 was initially made magical.
1151 However, note that C<sv_unmagic> removes all magic of a certain C<type> from the
1152 C<SV>. If you want to remove only certain
1153 magic of a C<type> based on the magic
1154 virtual table, use C<sv_unmagicext> instead:
1156 int sv_unmagicext(SV *sv, int type, MGVTBL *vtbl);
1158 =head2 Magic Virtual Tables
1160 The C<mg_virtual> field in the C<MAGIC> structure is a pointer to an
1161 C<MGVTBL>, which is a structure of function pointers and stands for
1162 "Magic Virtual Table" to handle the various operations that might be
1163 applied to that variable.
1165 The C<MGVTBL> has five (or sometimes eight) pointers to the following
1168 int (*svt_get) (pTHX_ SV* sv, MAGIC* mg);
1169 int (*svt_set) (pTHX_ SV* sv, MAGIC* mg);
1170 U32 (*svt_len) (pTHX_ SV* sv, MAGIC* mg);
1171 int (*svt_clear)(pTHX_ SV* sv, MAGIC* mg);
1172 int (*svt_free) (pTHX_ SV* sv, MAGIC* mg);
1174 int (*svt_copy) (pTHX_ SV *sv, MAGIC* mg, SV *nsv,
1175 const char *name, I32 namlen);
1176 int (*svt_dup) (pTHX_ MAGIC *mg, CLONE_PARAMS *param);
1177 int (*svt_local)(pTHX_ SV *nsv, MAGIC *mg);
1180 This MGVTBL structure is set at compile-time in F<perl.h> and there are
1181 currently 32 types. These different structures contain pointers to various
1182 routines that perform additional actions depending on which function is
1185 Function pointer Action taken
1186 ---------------- ------------
1187 svt_get Do something before the value of the SV is
1189 svt_set Do something after the SV is assigned a value.
1190 svt_len Report on the SV's length.
1191 svt_clear Clear something the SV represents.
1192 svt_free Free any extra storage associated with the SV.
1194 svt_copy copy tied variable magic to a tied element
1195 svt_dup duplicate a magic structure during thread cloning
1196 svt_local copy magic to local value during 'local'
1198 For instance, the MGVTBL structure called C<vtbl_sv> (which corresponds
1199 to an C<mg_type> of C<PERL_MAGIC_sv>) contains:
1201 { magic_get, magic_set, magic_len, 0, 0 }
1203 Thus, when an SV is determined to be magical and of type C<PERL_MAGIC_sv>,
1204 if a get operation is being performed, the routine C<magic_get> is
1205 called. All the various routines for the various magical types begin
1206 with C<magic_>. NOTE: the magic routines are not considered part of
1207 the Perl API, and may not be exported by the Perl library.
1209 The last three slots are a recent addition, and for source code
1210 compatibility they are only checked for if one of the three flags
1211 MGf_COPY, MGf_DUP or MGf_LOCAL is set in mg_flags.
1212 This means that most code can continue declaring
1213 a vtable as a 5-element value. These three are
1214 currently used exclusively by the threading code, and are highly subject
1217 The current kinds of Magic Virtual Tables are:
1220 This table is generated by regen/mg_vtable.pl. Any changes made here
1223 =for mg_vtable.pl begin
1226 (old-style char and macro) MGVTBL Type of magic
1227 -------------------------- ------ -------------
1228 \0 PERL_MAGIC_sv vtbl_sv Special scalar variable
1229 # PERL_MAGIC_arylen vtbl_arylen Array length ($#ary)
1230 % PERL_MAGIC_rhash (none) Extra data for restricted
1232 * PERL_MAGIC_debugvar vtbl_debugvar $DB::single, signal, trace
1234 . PERL_MAGIC_pos vtbl_pos pos() lvalue
1235 : PERL_MAGIC_symtab (none) Extra data for symbol
1237 < PERL_MAGIC_backref vtbl_backref For weak ref data
1238 @ PERL_MAGIC_arylen_p (none) To move arylen out of XPVAV
1239 B PERL_MAGIC_bm vtbl_regexp Boyer-Moore
1240 (fast string search)
1241 c PERL_MAGIC_overload_table vtbl_ovrld Holds overload table
1243 D PERL_MAGIC_regdata vtbl_regdata Regex match position data
1245 d PERL_MAGIC_regdatum vtbl_regdatum Regex match position data
1247 E PERL_MAGIC_env vtbl_env %ENV hash
1248 e PERL_MAGIC_envelem vtbl_envelem %ENV hash element
1249 f PERL_MAGIC_fm vtbl_regexp Formline
1251 g PERL_MAGIC_regex_global vtbl_mglob m//g target
1252 H PERL_MAGIC_hints vtbl_hints %^H hash
1253 h PERL_MAGIC_hintselem vtbl_hintselem %^H hash element
1254 I PERL_MAGIC_isa vtbl_isa @ISA array
1255 i PERL_MAGIC_isaelem vtbl_isaelem @ISA array element
1256 k PERL_MAGIC_nkeys vtbl_nkeys scalar(keys()) lvalue
1257 L PERL_MAGIC_dbfile (none) Debugger %_<filename
1258 l PERL_MAGIC_dbline vtbl_dbline Debugger %_<filename
1260 N PERL_MAGIC_shared (none) Shared between threads
1261 n PERL_MAGIC_shared_scalar (none) Shared between threads
1262 o PERL_MAGIC_collxfrm vtbl_collxfrm Locale transformation
1263 P PERL_MAGIC_tied vtbl_pack Tied array or hash
1264 p PERL_MAGIC_tiedelem vtbl_packelem Tied array or hash element
1265 q PERL_MAGIC_tiedscalar vtbl_packelem Tied scalar or handle
1266 r PERL_MAGIC_qr vtbl_regexp Precompiled qr// regex
1267 S PERL_MAGIC_sig (none) %SIG hash
1268 s PERL_MAGIC_sigelem vtbl_sigelem %SIG hash element
1269 t PERL_MAGIC_taint vtbl_taint Taintedness
1270 U PERL_MAGIC_uvar vtbl_uvar Available for use by
1272 u PERL_MAGIC_uvar_elem (none) Reserved for use by
1274 V PERL_MAGIC_vstring (none) SV was vstring literal
1275 v PERL_MAGIC_vec vtbl_vec vec() lvalue
1276 w PERL_MAGIC_utf8 vtbl_utf8 Cached UTF-8 information
1277 x PERL_MAGIC_substr vtbl_substr substr() lvalue
1278 Y PERL_MAGIC_nonelem vtbl_nonelem Array element that does not
1280 y PERL_MAGIC_defelem vtbl_defelem Shadow "foreach" iterator
1281 variable / smart parameter
1283 \ PERL_MAGIC_lvref vtbl_lvref Lvalue reference
1285 ] PERL_MAGIC_checkcall vtbl_checkcall Inlining/mutation of call
1287 ~ PERL_MAGIC_ext (none) Available for use by
1291 =for apidoc Amnh||PERL_MAGIC_sv
1292 =for apidoc Amnh||PERL_MAGIC_arylen
1293 =for apidoc Amnh||PERL_MAGIC_rhash
1294 =for apidoc Amnh||PERL_MAGIC_debugvar
1295 =for apidoc Amnh||PERL_MAGIC_pos
1296 =for apidoc Amnh||PERL_MAGIC_symtab
1297 =for apidoc Amnh||PERL_MAGIC_backref
1298 =for apidoc Amnh||PERL_MAGIC_arylen_p
1299 =for apidoc Amnh||PERL_MAGIC_bm
1300 =for apidoc Amnh||PERL_MAGIC_overload_table
1301 =for apidoc Amnh||PERL_MAGIC_regdata
1302 =for apidoc Amnh||PERL_MAGIC_regdatum
1303 =for apidoc Amnh||PERL_MAGIC_env
1304 =for apidoc Amnh||PERL_MAGIC_envelem
1305 =for apidoc Amnh||PERL_MAGIC_fm
1306 =for apidoc Amnh||PERL_MAGIC_regex_global
1307 =for apidoc Amnh||PERL_MAGIC_hints
1308 =for apidoc Amnh||PERL_MAGIC_hintselem
1309 =for apidoc Amnh||PERL_MAGIC_isa
1310 =for apidoc Amnh||PERL_MAGIC_isaelem
1311 =for apidoc Amnh||PERL_MAGIC_nkeys
1312 =for apidoc Amnh||PERL_MAGIC_dbfile
1313 =for apidoc Amnh||PERL_MAGIC_dbline
1314 =for apidoc Amnh||PERL_MAGIC_shared
1315 =for apidoc Amnh||PERL_MAGIC_shared_scalar
1316 =for apidoc Amnh||PERL_MAGIC_collxfrm
1317 =for apidoc Amnh||PERL_MAGIC_tied
1318 =for apidoc Amnh||PERL_MAGIC_tiedelem
1319 =for apidoc Amnh||PERL_MAGIC_tiedscalar
1320 =for apidoc Amnh||PERL_MAGIC_qr
1321 =for apidoc Amnh||PERL_MAGIC_sig
1322 =for apidoc Amnh||PERL_MAGIC_sigelem
1323 =for apidoc Amnh||PERL_MAGIC_taint
1324 =for apidoc Amnh||PERL_MAGIC_uvar
1325 =for apidoc Amnh||PERL_MAGIC_uvar_elem
1326 =for apidoc Amnh||PERL_MAGIC_vstring
1327 =for apidoc Amnh||PERL_MAGIC_vec
1328 =for apidoc Amnh||PERL_MAGIC_utf8
1329 =for apidoc Amnh||PERL_MAGIC_substr
1330 =for apidoc Amnh||PERL_MAGIC_nonelem
1331 =for apidoc Amnh||PERL_MAGIC_defelem
1332 =for apidoc Amnh||PERL_MAGIC_lvref
1333 =for apidoc Amnh||PERL_MAGIC_checkcall
1334 =for apidoc Amnh||PERL_MAGIC_ext
1336 =for mg_vtable.pl end
1338 When an uppercase and lowercase letter both exist in the table, then the
1339 uppercase letter is typically used to represent some kind of composite type
1340 (a list or a hash), and the lowercase letter is used to represent an element
1341 of that composite type. Some internals code makes use of this case
1342 relationship. However, 'v' and 'V' (vec and v-string) are in no way related.
1344 The C<PERL_MAGIC_ext> and C<PERL_MAGIC_uvar> magic types are defined
1345 specifically for use by extensions and will not be used by perl itself.
1346 Extensions can use C<PERL_MAGIC_ext> magic to 'attach' private information
1347 to variables (typically objects). This is especially useful because
1348 there is no way for normal perl code to corrupt this private information
1349 (unlike using extra elements of a hash object).
1351 Similarly, C<PERL_MAGIC_uvar> magic can be used much like tie() to call a
1352 C function any time a scalar's value is used or changed. The C<MAGIC>'s
1353 C<mg_ptr> field points to a C<ufuncs> structure:
1356 I32 (*uf_val)(pTHX_ IV, SV*);
1357 I32 (*uf_set)(pTHX_ IV, SV*);
1361 When the SV is read from or written to, the C<uf_val> or C<uf_set>
1362 function will be called with C<uf_index> as the first arg and a pointer to
1363 the SV as the second. A simple example of how to add C<PERL_MAGIC_uvar>
1364 magic is shown below. Note that the ufuncs structure is copied by
1365 sv_magic, so you can safely allocate it on the stack.
1373 uf.uf_val = &my_get_fn;
1374 uf.uf_set = &my_set_fn;
1376 sv_magic(sv, 0, PERL_MAGIC_uvar, (char*)&uf, sizeof(uf));
1378 Attaching C<PERL_MAGIC_uvar> to arrays is permissible but has no effect.
1380 For hashes there is a specialized hook that gives control over hash
1381 keys (but not values). This hook calls C<PERL_MAGIC_uvar> 'get' magic
1382 if the "set" function in the C<ufuncs> structure is NULL. The hook
1383 is activated whenever the hash is accessed with a key specified as
1384 an C<SV> through the functions C<hv_store_ent>, C<hv_fetch_ent>,
1385 C<hv_delete_ent>, and C<hv_exists_ent>. Accessing the key as a string
1386 through the functions without the C<..._ent> suffix circumvents the
1387 hook. See L<Hash::Util::FieldHash/GUTS> for a detailed description.
1389 Note that because multiple extensions may be using C<PERL_MAGIC_ext>
1390 or C<PERL_MAGIC_uvar> magic, it is important for extensions to take
1391 extra care to avoid conflict. Typically only using the magic on
1392 objects blessed into the same class as the extension is sufficient.
1393 For C<PERL_MAGIC_ext> magic, it is usually a good idea to define an
1394 C<MGVTBL>, even if all its fields will be C<0>, so that individual
1395 C<MAGIC> pointers can be identified as a particular kind of magic
1396 using their magic virtual table. C<mg_findext> provides an easy way
1399 STATIC MGVTBL my_vtbl = { 0, 0, 0, 0, 0, 0, 0, 0 };
1402 if ((mg = mg_findext(sv, PERL_MAGIC_ext, &my_vtbl))) {
1403 /* this is really ours, not another module's PERL_MAGIC_ext */
1404 my_priv_data_t *priv = (my_priv_data_t *)mg->mg_ptr;
1408 Also note that the C<sv_set*()> and C<sv_cat*()> functions described
1409 earlier do B<not> invoke 'set' magic on their targets. This must
1410 be done by the user either by calling the C<SvSETMAGIC()> macro after
1411 calling these functions, or by using one of the C<sv_set*_mg()> or
1412 C<sv_cat*_mg()> functions. Similarly, generic C code must call the
1413 C<SvGETMAGIC()> macro to invoke any 'get' magic if they use an SV
1414 obtained from external sources in functions that don't handle magic.
1415 See L<perlapi> for a description of these functions.
1416 For example, calls to the C<sv_cat*()> functions typically need to be
1417 followed by C<SvSETMAGIC()>, but they don't need a prior C<SvGETMAGIC()>
1418 since their implementation handles 'get' magic.
1420 =head2 Finding Magic
1422 MAGIC *mg_find(SV *sv, int type); /* Finds the magic pointer of that
1425 This routine returns a pointer to a C<MAGIC> structure stored in the SV.
1426 If the SV does not have that magical
1427 feature, C<NULL> is returned. If the
1428 SV has multiple instances of that magical feature, the first one will be
1429 returned. C<mg_findext> can be used
1430 to find a C<MAGIC> structure of an SV
1431 based on both its magic type and its magic virtual table:
1433 MAGIC *mg_findext(SV *sv, int type, MGVTBL *vtbl);
1435 Also, if the SV passed to C<mg_find> or C<mg_findext> is not of type
1436 SVt_PVMG, Perl may core dump.
1438 int mg_copy(SV* sv, SV* nsv, const char* key, STRLEN klen);
1440 This routine checks to see what types of magic C<sv> has. If the mg_type
1441 field is an uppercase letter, then the mg_obj is copied to C<nsv>, but
1442 the mg_type field is changed to be the lowercase letter.
1444 =head2 Understanding the Magic of Tied Hashes and Arrays
1446 Tied hashes and arrays are magical beasts of the C<PERL_MAGIC_tied>
1449 WARNING: As of the 5.004 release, proper usage of the array and hash
1450 access functions requires understanding a few caveats. Some
1451 of these caveats are actually considered bugs in the API, to be fixed
1452 in later releases, and are bracketed with [MAYCHANGE] below. If
1453 you find yourself actually applying such information in this section, be
1454 aware that the behavior may change in the future, umm, without warning.
1456 The perl tie function associates a variable with an object that implements
1457 the various GET, SET, etc methods. To perform the equivalent of the perl
1458 tie function from an XSUB, you must mimic this behaviour. The code below
1459 carries out the necessary steps -- firstly it creates a new hash, and then
1460 creates a second hash which it blesses into the class which will implement
1461 the tie methods. Lastly it ties the two hashes together, and returns a
1462 reference to the new tied hash. Note that the code below does NOT call the
1463 TIEHASH method in the MyTie class -
1464 see L</Calling Perl Routines from within C Programs> for details on how
1475 tie = newRV_noinc((SV*)newHV());
1476 stash = gv_stashpv("MyTie", GV_ADD);
1477 sv_bless(tie, stash);
1478 hv_magic(hash, (GV*)tie, PERL_MAGIC_tied);
1479 RETVAL = newRV_noinc(hash);
1483 The C<av_store> function, when given a tied array argument, merely
1484 copies the magic of the array onto the value to be "stored", using
1485 C<mg_copy>. It may also return NULL, indicating that the value did not
1486 actually need to be stored in the array. [MAYCHANGE] After a call to
1487 C<av_store> on a tied array, the caller will usually need to call
1488 C<mg_set(val)> to actually invoke the perl level "STORE" method on the
1489 TIEARRAY object. If C<av_store> did return NULL, a call to
1490 C<SvREFCNT_dec(val)> will also be usually necessary to avoid a memory
1493 The previous paragraph is applicable verbatim to tied hash access using the
1494 C<hv_store> and C<hv_store_ent> functions as well.
1496 C<av_fetch> and the corresponding hash functions C<hv_fetch> and
1497 C<hv_fetch_ent> actually return an undefined mortal value whose magic
1498 has been initialized using C<mg_copy>. Note the value so returned does not
1499 need to be deallocated, as it is already mortal. [MAYCHANGE] But you will
1500 need to call C<mg_get()> on the returned value in order to actually invoke
1501 the perl level "FETCH" method on the underlying TIE object. Similarly,
1502 you may also call C<mg_set()> on the return value after possibly assigning
1503 a suitable value to it using C<sv_setsv>, which will invoke the "STORE"
1504 method on the TIE object. [/MAYCHANGE]
1507 In other words, the array or hash fetch/store functions don't really
1508 fetch and store actual values in the case of tied arrays and hashes. They
1509 merely call C<mg_copy> to attach magic to the values that were meant to be
1510 "stored" or "fetched". Later calls to C<mg_get> and C<mg_set> actually
1511 do the job of invoking the TIE methods on the underlying objects. Thus
1512 the magic mechanism currently implements a kind of lazy access to arrays
1515 Currently (as of perl version 5.004), use of the hash and array access
1516 functions requires the user to be aware of whether they are operating on
1517 "normal" hashes and arrays, or on their tied variants. The API may be
1518 changed to provide more transparent access to both tied and normal data
1519 types in future versions.
1522 You would do well to understand that the TIEARRAY and TIEHASH interfaces
1523 are mere sugar to invoke some perl method calls while using the uniform hash
1524 and array syntax. The use of this sugar imposes some overhead (typically
1525 about two to four extra opcodes per FETCH/STORE operation, in addition to
1526 the creation of all the mortal variables required to invoke the methods).
1527 This overhead will be comparatively small if the TIE methods are themselves
1528 substantial, but if they are only a few statements long, the overhead
1529 will not be insignificant.
1531 =head2 Localizing changes
1533 Perl has a very handy construction
1540 This construction is I<approximately> equivalent to
1549 The biggest difference is that the first construction would
1550 reinstate the initial value of $var, irrespective of how control exits
1551 the block: C<goto>, C<return>, C<die>/C<eval>, etc. It is a little bit
1552 more efficient as well.
1554 There is a way to achieve a similar task from C via Perl API: create a
1555 I<pseudo-block>, and arrange for some changes to be automatically
1556 undone at the end of it, either explicit, or via a non-local exit (via
1557 die()). A I<block>-like construct is created by a pair of
1558 C<ENTER>/C<LEAVE> macros (see L<perlcall/"Returning a Scalar">).
1559 Such a construct may be created specially for some important localized
1560 task, or an existing one (like boundaries of enclosing Perl
1561 subroutine/block, or an existing pair for freeing TMPs) may be
1562 used. (In the second case the overhead of additional localization must
1563 be almost negligible.) Note that any XSUB is automatically enclosed in
1564 an C<ENTER>/C<LEAVE> pair.
1566 Inside such a I<pseudo-block> the following service is available:
1570 =item C<SAVEINT(int i)>
1572 =item C<SAVEIV(IV i)>
1574 =item C<SAVEI32(I32 i)>
1576 =item C<SAVELONG(long i)>
1578 These macros arrange things to restore the value of integer variable
1579 C<i> at the end of enclosing I<pseudo-block>.
1581 =item C<SAVESPTR(s)>
1583 =item C<SAVEPPTR(p)>
1585 These macros arrange things to restore the value of pointers C<s> and
1586 C<p>. C<s> must be a pointer of a type which survives conversion to
1587 C<SV*> and back, C<p> should be able to survive conversion to C<char*>
1590 =item C<SAVEFREESV(SV *sv)>
1592 The refcount of C<sv> will be decremented at the end of
1593 I<pseudo-block>. This is similar to C<sv_2mortal> in that it is also a
1594 mechanism for doing a delayed C<SvREFCNT_dec>. However, while C<sv_2mortal>
1595 extends the lifetime of C<sv> until the beginning of the next statement,
1596 C<SAVEFREESV> extends it until the end of the enclosing scope. These
1597 lifetimes can be wildly different.
1599 Also compare C<SAVEMORTALIZESV>.
1601 =item C<SAVEMORTALIZESV(SV *sv)>
1603 Just like C<SAVEFREESV>, but mortalizes C<sv> at the end of the current
1604 scope instead of decrementing its reference count. This usually has the
1605 effect of keeping C<sv> alive until the statement that called the currently
1606 live scope has finished executing.
1608 =item C<SAVEFREEOP(OP *op)>
1610 The C<OP *> is op_free()ed at the end of I<pseudo-block>.
1612 =item C<SAVEFREEPV(p)>
1614 The chunk of memory which is pointed to by C<p> is Safefree()ed at the
1615 end of I<pseudo-block>.
1617 =item C<SAVECLEARSV(SV *sv)>
1619 Clears a slot in the current scratchpad which corresponds to C<sv> at
1620 the end of I<pseudo-block>.
1622 =item C<SAVEDELETE(HV *hv, char *key, I32 length)>
1624 The key C<key> of C<hv> is deleted at the end of I<pseudo-block>. The
1625 string pointed to by C<key> is Safefree()ed. If one has a I<key> in
1626 short-lived storage, the corresponding string may be reallocated like
1629 SAVEDELETE(PL_defstash, savepv(tmpbuf), strlen(tmpbuf));
1631 =item C<SAVEDESTRUCTOR(DESTRUCTORFUNC_NOCONTEXT_t f, void *p)>
1633 At the end of I<pseudo-block> the function C<f> is called with the
1636 =item C<SAVEDESTRUCTOR_X(DESTRUCTORFUNC_t f, void *p)>
1638 At the end of I<pseudo-block> the function C<f> is called with the
1639 implicit context argument (if any), and C<p>.
1641 =item C<SAVESTACK_POS()>
1643 The current offset on the Perl internal stack (cf. C<SP>) is restored
1644 at the end of I<pseudo-block>.
1648 The following API list contains functions, thus one needs to
1649 provide pointers to the modifiable data explicitly (either C pointers,
1650 or Perlish C<GV *>s). Where the above macros take C<int>, a similar
1651 function takes C<int *>.
1655 =item C<SV* save_scalar(GV *gv)>
1657 =for apidoc save_scalar
1659 Equivalent to Perl code C<local $gv>.
1661 =item C<AV* save_ary(GV *gv)>
1663 =for apidoc save_ary
1665 =item C<HV* save_hash(GV *gv)>
1667 =for apidoc save_hash
1669 Similar to C<save_scalar>, but localize C<@gv> and C<%gv>.
1671 =item C<void save_item(SV *item)>
1673 =for apidoc save_item
1675 Duplicates the current value of C<SV>. On the exit from the current
1676 C<ENTER>/C<LEAVE> I<pseudo-block> the value of C<SV> will be restored
1677 using the stored value. It doesn't handle magic. Use C<save_scalar> if
1680 =item C<void save_list(SV **sarg, I32 maxsarg)>
1682 =for apidoc save_list
1684 A variant of C<save_item> which takes multiple arguments via an array
1685 C<sarg> of C<SV*> of length C<maxsarg>.
1687 =item C<SV* save_svref(SV **sptr)>
1689 =for apidoc save_svref
1691 Similar to C<save_scalar>, but will reinstate an C<SV *>.
1693 =item C<void save_aptr(AV **aptr)>
1695 =item C<void save_hptr(HV **hptr)>
1697 =for apidoc save_aptr
1698 =for apidoc save_hptr
1700 Similar to C<save_svref>, but localize C<AV *> and C<HV *>.
1704 The C<Alias> module implements localization of the basic types within the
1705 I<caller's scope>. People who are interested in how to localize things in
1706 the containing scope should take a look there too.
1710 =head2 XSUBs and the Argument Stack
1712 The XSUB mechanism is a simple way for Perl programs to access C subroutines.
1713 An XSUB routine will have a stack that contains the arguments from the Perl
1714 program, and a way to map from the Perl data structures to a C equivalent.
1716 The stack arguments are accessible through the C<ST(n)> macro, which returns
1717 the C<n>'th stack argument. Argument 0 is the first argument passed in the
1718 Perl subroutine call. These arguments are C<SV*>, and can be used anywhere
1721 Most of the time, output from the C routine can be handled through use of
1722 the RETVAL and OUTPUT directives. However, there are some cases where the
1723 argument stack is not already long enough to handle all the return values.
1724 An example is the POSIX tzname() call, which takes no arguments, but returns
1725 two, the local time zone's standard and summer time abbreviations.
1727 To handle this situation, the PPCODE directive is used and the stack is
1728 extended using the macro:
1732 where C<SP> is the macro that represents the local copy of the stack pointer,
1733 and C<num> is the number of elements the stack should be extended by.
1735 Now that there is room on the stack, values can be pushed on it using C<PUSHs>
1736 macro. The pushed values will often need to be "mortal" (See
1737 L</Reference Counts and Mortality>):
1739 PUSHs(sv_2mortal(newSViv(an_integer)))
1740 PUSHs(sv_2mortal(newSVuv(an_unsigned_integer)))
1741 PUSHs(sv_2mortal(newSVnv(a_double)))
1742 PUSHs(sv_2mortal(newSVpv("Some String",0)))
1743 /* Although the last example is better written as the more
1745 PUSHs(newSVpvs_flags("Some String", SVs_TEMP))
1747 And now the Perl program calling C<tzname>, the two values will be assigned
1750 ($standard_abbrev, $summer_abbrev) = POSIX::tzname;
1752 An alternate (and possibly simpler) method to pushing values on the stack is
1757 This macro automatically adjusts the stack for you, if needed. Thus, you
1758 do not need to call C<EXTEND> to extend the stack.
1760 Despite their suggestions in earlier versions of this document the macros
1761 C<(X)PUSH[iunp]> are I<not> suited to XSUBs which return multiple results.
1762 For that, either stick to the C<(X)PUSHs> macros shown above, or use the new
1763 C<m(X)PUSH[iunp]> macros instead; see L</Putting a C value on Perl stack>.
1765 For more information, consult L<perlxs> and L<perlxstut>.
1767 =head2 Autoloading with XSUBs
1769 If an AUTOLOAD routine is an XSUB, as with Perl subroutines, Perl puts the
1770 fully-qualified name of the autoloaded subroutine in the $AUTOLOAD variable
1771 of the XSUB's package.
1773 But it also puts the same information in certain fields of the XSUB itself:
1775 HV *stash = CvSTASH(cv);
1776 const char *subname = SvPVX(cv);
1777 STRLEN name_length = SvCUR(cv); /* in bytes */
1778 U32 is_utf8 = SvUTF8(cv);
1780 C<SvPVX(cv)> contains just the sub name itself, not including the package.
1781 For an AUTOLOAD routine in UNIVERSAL or one of its superclasses,
1782 C<CvSTASH(cv)> returns NULL during a method call on a nonexistent package.
1784 B<Note>: Setting $AUTOLOAD stopped working in 5.6.1, which did not support
1785 XS AUTOLOAD subs at all. Perl 5.8.0 introduced the use of fields in the
1786 XSUB itself. Perl 5.16.0 restored the setting of $AUTOLOAD. If you need
1787 to support 5.8-5.14, use the XSUB's fields.
1789 =head2 Calling Perl Routines from within C Programs
1791 There are four routines that can be used to call a Perl subroutine from
1792 within a C program. These four are:
1794 I32 call_sv(SV*, I32);
1795 I32 call_pv(const char*, I32);
1796 I32 call_method(const char*, I32);
1797 I32 call_argv(const char*, I32, char**);
1799 The routine most often used is C<call_sv>. The C<SV*> argument
1800 contains either the name of the Perl subroutine to be called, or a
1801 reference to the subroutine. The second argument consists of flags
1802 that control the context in which the subroutine is called, whether
1803 or not the subroutine is being passed arguments, how errors should be
1804 trapped, and how to treat return values.
1806 All four routines return the number of arguments that the subroutine returned
1809 These routines used to be called C<perl_call_sv>, etc., before Perl v5.6.0,
1810 but those names are now deprecated; macros of the same name are provided for
1813 When using any of these routines (except C<call_argv>), the programmer
1814 must manipulate the Perl stack. These include the following macros and
1829 For a detailed description of calling conventions from C to Perl,
1830 consult L<perlcall>.
1832 =head2 Putting a C value on Perl stack
1834 A lot of opcodes (this is an elementary operation in the internal perl
1835 stack machine) put an SV* on the stack. However, as an optimization
1836 the corresponding SV is (usually) not recreated each time. The opcodes
1837 reuse specially assigned SVs (I<target>s) which are (as a corollary)
1838 not constantly freed/created.
1840 Each of the targets is created only once (but see
1841 L</Scratchpads and recursion> below), and when an opcode needs to put
1842 an integer, a double, or a string on stack, it just sets the
1843 corresponding parts of its I<target> and puts the I<target> on stack.
1845 The macro to put this target on stack is C<PUSHTARG>, and it is
1846 directly used in some opcodes, as well as indirectly in zillions of
1847 others, which use it via C<(X)PUSH[iunp]>.
1849 Because the target is reused, you must be careful when pushing multiple
1850 values on the stack. The following code will not do what you think:
1855 This translates as "set C<TARG> to 10, push a pointer to C<TARG> onto
1856 the stack; set C<TARG> to 20, push a pointer to C<TARG> onto the stack".
1857 At the end of the operation, the stack does not contain the values 10
1858 and 20, but actually contains two pointers to C<TARG>, which we have set
1861 If you need to push multiple different values then you should either use
1862 the C<(X)PUSHs> macros, or else use the new C<m(X)PUSH[iunp]> macros,
1863 none of which make use of C<TARG>. The C<(X)PUSHs> macros simply push an
1864 SV* on the stack, which, as noted under L</XSUBs and the Argument Stack>,
1865 will often need to be "mortal". The new C<m(X)PUSH[iunp]> macros make
1866 this a little easier to achieve by creating a new mortal for you (via
1867 C<(X)PUSHmortal>), pushing that onto the stack (extending it if necessary
1868 in the case of the C<mXPUSH[iunp]> macros), and then setting its value.
1869 Thus, instead of writing this to "fix" the example above:
1871 XPUSHs(sv_2mortal(newSViv(10)))
1872 XPUSHs(sv_2mortal(newSViv(20)))
1874 you can simply write:
1879 On a related note, if you do use C<(X)PUSH[iunp]>, then you're going to
1880 need a C<dTARG> in your variable declarations so that the C<*PUSH*>
1881 macros can make use of the local variable C<TARG>. See also C<dTARGET>
1886 The question remains on when the SVs which are I<target>s for opcodes
1887 are created. The answer is that they are created when the current
1888 unit--a subroutine or a file (for opcodes for statements outside of
1889 subroutines)--is compiled. During this time a special anonymous Perl
1890 array is created, which is called a scratchpad for the current unit.
1892 A scratchpad keeps SVs which are lexicals for the current unit and are
1893 targets for opcodes. A previous version of this document
1894 stated that one can deduce that an SV lives on a scratchpad
1895 by looking on its flags: lexicals have C<SVs_PADMY> set, and
1896 I<target>s have C<SVs_PADTMP> set. But this has never been fully true.
1897 C<SVs_PADMY> could be set on a variable that no longer resides in any pad.
1898 While I<target>s do have C<SVs_PADTMP> set, it can also be set on variables
1899 that have never resided in a pad, but nonetheless act like I<target>s. As
1900 of perl 5.21.5, the C<SVs_PADMY> flag is no longer used and is defined as
1901 0. C<SvPADMY()> now returns true for anything without C<SVs_PADTMP>.
1903 The correspondence between OPs and I<target>s is not 1-to-1. Different
1904 OPs in the compile tree of the unit can use the same target, if this
1905 would not conflict with the expected life of the temporary.
1907 =head2 Scratchpads and recursion
1909 In fact it is not 100% true that a compiled unit contains a pointer to
1910 the scratchpad AV. In fact it contains a pointer to an AV of
1911 (initially) one element, and this element is the scratchpad AV. Why do
1912 we need an extra level of indirection?
1914 The answer is B<recursion>, and maybe B<threads>. Both
1915 these can create several execution pointers going into the same
1916 subroutine. For the subroutine-child not write over the temporaries
1917 for the subroutine-parent (lifespan of which covers the call to the
1918 child), the parent and the child should have different
1919 scratchpads. (I<And> the lexicals should be separate anyway!)
1921 So each subroutine is born with an array of scratchpads (of length 1).
1922 On each entry to the subroutine it is checked that the current
1923 depth of the recursion is not more than the length of this array, and
1924 if it is, new scratchpad is created and pushed into the array.
1926 The I<target>s on this scratchpad are C<undef>s, but they are already
1927 marked with correct flags.
1929 =head1 Memory Allocation
1933 All memory meant to be used with the Perl API functions should be manipulated
1934 using the macros described in this section. The macros provide the necessary
1935 transparency between differences in the actual malloc implementation that is
1938 The following three macros are used to initially allocate memory :
1940 Newx(pointer, number, type);
1941 Newxc(pointer, number, type, cast);
1942 Newxz(pointer, number, type);
1944 The first argument C<pointer> should be the name of a variable that will
1945 point to the newly allocated memory.
1947 The second and third arguments C<number> and C<type> specify how many of
1948 the specified type of data structure should be allocated. The argument
1949 C<type> is passed to C<sizeof>. The final argument to C<Newxc>, C<cast>,
1950 should be used if the C<pointer> argument is different from the C<type>
1953 Unlike the C<Newx> and C<Newxc> macros, the C<Newxz> macro calls C<memzero>
1954 to zero out all the newly allocated memory.
1958 Renew(pointer, number, type);
1959 Renewc(pointer, number, type, cast);
1962 These three macros are used to change a memory buffer size or to free a
1963 piece of memory no longer needed. The arguments to C<Renew> and C<Renewc>
1964 match those of C<New> and C<Newc> with the exception of not needing the
1965 "magic cookie" argument.
1969 Move(source, dest, number, type);
1970 Copy(source, dest, number, type);
1971 Zero(dest, number, type);
1973 These three macros are used to move, copy, or zero out previously allocated
1974 memory. The C<source> and C<dest> arguments point to the source and
1975 destination starting points. Perl will move, copy, or zero out C<number>
1976 instances of the size of the C<type> data structure (using the C<sizeof>
1981 The most recent development releases of Perl have been experimenting with
1982 removing Perl's dependency on the "normal" standard I/O suite and allowing
1983 other stdio implementations to be used. This involves creating a new
1984 abstraction layer that then calls whichever implementation of stdio Perl
1985 was compiled with. All XSUBs should now use the functions in the PerlIO
1986 abstraction layer and not make any assumptions about what kind of stdio
1989 For a complete description of the PerlIO abstraction, consult L<perlapio>.
1991 =head1 Compiled code
1995 Here we describe the internal form your code is converted to by
1996 Perl. Start with a simple example:
2000 This is converted to a tree similar to this one:
2008 (but slightly more complicated). This tree reflects the way Perl
2009 parsed your code, but has nothing to do with the execution order.
2010 There is an additional "thread" going through the nodes of the tree
2011 which shows the order of execution of the nodes. In our simplified
2012 example above it looks like:
2014 $b ---> $c ---> + ---> $a ---> assign-to
2016 But with the actual compile tree for C<$a = $b + $c> it is different:
2017 some nodes I<optimized away>. As a corollary, though the actual tree
2018 contains more nodes than our simplified example, the execution order
2019 is the same as in our example.
2021 =head2 Examining the tree
2023 If you have your perl compiled for debugging (usually done with
2024 C<-DDEBUGGING> on the C<Configure> command line), you may examine the
2025 compiled tree by specifying C<-Dx> on the Perl command line. The
2026 output takes several lines per node, and for C<$b+$c> it looks like
2031 FLAGS = (SCALAR,KIDS)
2033 TYPE = null ===> (4)
2035 FLAGS = (SCALAR,KIDS)
2037 3 TYPE = gvsv ===> 4
2043 TYPE = null ===> (5)
2045 FLAGS = (SCALAR,KIDS)
2047 4 TYPE = gvsv ===> 5
2053 This tree has 5 nodes (one per C<TYPE> specifier), only 3 of them are
2054 not optimized away (one per number in the left column). The immediate
2055 children of the given node correspond to C<{}> pairs on the same level
2056 of indentation, thus this listing corresponds to the tree:
2064 The execution order is indicated by C<===E<gt>> marks, thus it is C<3
2065 4 5 6> (node C<6> is not included into above listing), i.e.,
2066 C<gvsv gvsv add whatever>.
2068 Each of these nodes represents an op, a fundamental operation inside the
2069 Perl core. The code which implements each operation can be found in the
2070 F<pp*.c> files; the function which implements the op with type C<gvsv>
2071 is C<pp_gvsv>, and so on. As the tree above shows, different ops have
2072 different numbers of children: C<add> is a binary operator, as one would
2073 expect, and so has two children. To accommodate the various different
2074 numbers of children, there are various types of op data structure, and
2075 they link together in different ways.
2077 The simplest type of op structure is C<OP>: this has no children. Unary
2078 operators, C<UNOP>s, have one child, and this is pointed to by the
2079 C<op_first> field. Binary operators (C<BINOP>s) have not only an
2080 C<op_first> field but also an C<op_last> field. The most complex type of
2081 op is a C<LISTOP>, which has any number of children. In this case, the
2082 first child is pointed to by C<op_first> and the last child by
2083 C<op_last>. The children in between can be found by iteratively
2084 following the C<OpSIBLING> pointer from the first child to the last (but
2087 There are also some other op types: a C<PMOP> holds a regular expression,
2088 and has no children, and a C<LOOP> may or may not have children. If the
2089 C<op_children> field is non-zero, it behaves like a C<LISTOP>. To
2090 complicate matters, if a C<UNOP> is actually a C<null> op after
2091 optimization (see L</Compile pass 2: context propagation>) it will still
2092 have children in accordance with its former type.
2094 Finally, there is a C<LOGOP>, or logic op. Like a C<LISTOP>, this has one
2095 or more children, but it doesn't have an C<op_last> field: so you have to
2096 follow C<op_first> and then the C<OpSIBLING> chain itself to find the
2097 last child. Instead it has an C<op_other> field, which is comparable to
2098 the C<op_next> field described below, and represents an alternate
2099 execution path. Operators like C<and>, C<or> and C<?> are C<LOGOP>s. Note
2100 that in general, C<op_other> may not point to any of the direct children
2103 Starting in version 5.21.2, perls built with the experimental
2104 define C<-DPERL_OP_PARENT> add an extra boolean flag for each op,
2105 C<op_moresib>. When not set, this indicates that this is the last op in an
2106 C<OpSIBLING> chain. This frees up the C<op_sibling> field on the last
2107 sibling to point back to the parent op. Under this build, that field is
2108 also renamed C<op_sibparent> to reflect its joint role. The macro
2109 C<OpSIBLING(o)> wraps this special behaviour, and always returns NULL on
2110 the last sibling. With this build the C<op_parent(o)> function can be
2111 used to find the parent of any op. Thus for forward compatibility, you
2112 should always use the C<OpSIBLING(o)> macro rather than accessing
2113 C<op_sibling> directly.
2115 Another way to examine the tree is to use a compiler back-end module, such
2118 =head2 Compile pass 1: check routines
2120 The tree is created by the compiler while I<yacc> code feeds it
2121 the constructions it recognizes. Since I<yacc> works bottom-up, so does
2122 the first pass of perl compilation.
2124 What makes this pass interesting for perl developers is that some
2125 optimization may be performed on this pass. This is optimization by
2126 so-called "check routines". The correspondence between node names
2127 and corresponding check routines is described in F<opcode.pl> (do not
2128 forget to run C<make regen_headers> if you modify this file).
2130 A check routine is called when the node is fully constructed except
2131 for the execution-order thread. Since at this time there are no
2132 back-links to the currently constructed node, one can do most any
2133 operation to the top-level node, including freeing it and/or creating
2134 new nodes above/below it.
2136 The check routine returns the node which should be inserted into the
2137 tree (if the top-level node was not modified, check routine returns
2140 By convention, check routines have names C<ck_*>. They are usually
2141 called from C<new*OP> subroutines (or C<convert>) (which in turn are
2142 called from F<perly.y>).
2144 =head2 Compile pass 1a: constant folding
2146 Immediately after the check routine is called the returned node is
2147 checked for being compile-time executable. If it is (the value is
2148 judged to be constant) it is immediately executed, and a I<constant>
2149 node with the "return value" of the corresponding subtree is
2150 substituted instead. The subtree is deleted.
2152 If constant folding was not performed, the execution-order thread is
2155 =head2 Compile pass 2: context propagation
2157 When a context for a part of compile tree is known, it is propagated
2158 down through the tree. At this time the context can have 5 values
2159 (instead of 2 for runtime context): void, boolean, scalar, list, and
2160 lvalue. In contrast with the pass 1 this pass is processed from top
2161 to bottom: a node's context determines the context for its children.
2163 Additional context-dependent optimizations are performed at this time.
2164 Since at this moment the compile tree contains back-references (via
2165 "thread" pointers), nodes cannot be free()d now. To allow
2166 optimized-away nodes at this stage, such nodes are null()ified instead
2167 of free()ing (i.e. their type is changed to OP_NULL).
2169 =head2 Compile pass 3: peephole optimization
2171 After the compile tree for a subroutine (or for an C<eval> or a file)
2172 is created, an additional pass over the code is performed. This pass
2173 is neither top-down or bottom-up, but in the execution order (with
2174 additional complications for conditionals). Optimizations performed
2175 at this stage are subject to the same restrictions as in the pass 2.
2177 Peephole optimizations are done by calling the function pointed to
2178 by the global variable C<PL_peepp>. By default, C<PL_peepp> just
2179 calls the function pointed to by the global variable C<PL_rpeepp>.
2180 By default, that performs some basic op fixups and optimisations along
2181 the execution-order op chain, and recursively calls C<PL_rpeepp> for
2182 each side chain of ops (resulting from conditionals). Extensions may
2183 provide additional optimisations or fixups, hooking into either the
2184 per-subroutine or recursive stage, like this:
2186 static peep_t prev_peepp;
2187 static void my_peep(pTHX_ OP *o)
2189 /* custom per-subroutine optimisation goes here */
2190 prev_peepp(aTHX_ o);
2191 /* custom per-subroutine optimisation may also go here */
2194 prev_peepp = PL_peepp;
2197 static peep_t prev_rpeepp;
2198 static void my_rpeep(pTHX_ OP *o)
2201 for(; o; o = o->op_next) {
2202 /* custom per-op optimisation goes here */
2204 prev_rpeepp(aTHX_ orig_o);
2207 prev_rpeepp = PL_rpeepp;
2208 PL_rpeepp = my_rpeep;
2210 =head2 Pluggable runops
2212 The compile tree is executed in a runops function. There are two runops
2213 functions, in F<run.c> and in F<dump.c>. C<Perl_runops_debug> is used
2214 with DEBUGGING and C<Perl_runops_standard> is used otherwise. For fine
2215 control over the execution of the compile tree it is possible to provide
2216 your own runops function.
2218 It's probably best to copy one of the existing runops functions and
2219 change it to suit your needs. Then, in the BOOT section of your XS
2222 PL_runops = my_runops;
2224 This function should be as efficient as possible to keep your programs
2225 running as fast as possible.
2227 =head2 Compile-time scope hooks
2229 As of perl 5.14 it is possible to hook into the compile-time lexical
2230 scope mechanism using C<Perl_blockhook_register>. This is used like
2233 STATIC void my_start_hook(pTHX_ int full);
2234 STATIC BHK my_hooks;
2237 BhkENTRY_set(&my_hooks, bhk_start, my_start_hook);
2238 Perl_blockhook_register(aTHX_ &my_hooks);
2240 This will arrange to have C<my_start_hook> called at the start of
2241 compiling every lexical scope. The available hooks are:
2245 =item C<void bhk_start(pTHX_ int full)>
2247 This is called just after starting a new lexical scope. Note that Perl
2252 creates two scopes: the first starts at the C<(> and has C<full == 1>,
2253 the second starts at the C<{> and has C<full == 0>. Both end at the
2254 C<}>, so calls to C<start> and C<pre>/C<post_end> will match. Anything
2255 pushed onto the save stack by this hook will be popped just before the
2256 scope ends (between the C<pre_> and C<post_end> hooks, in fact).
2258 =item C<void bhk_pre_end(pTHX_ OP **o)>
2260 This is called at the end of a lexical scope, just before unwinding the
2261 stack. I<o> is the root of the optree representing the scope; it is a
2262 double pointer so you can replace the OP if you need to.
2264 =item C<void bhk_post_end(pTHX_ OP **o)>
2266 This is called at the end of a lexical scope, just after unwinding the
2267 stack. I<o> is as above. Note that it is possible for calls to C<pre_>
2268 and C<post_end> to nest, if there is something on the save stack that
2271 =item C<void bhk_eval(pTHX_ OP *const o)>
2273 This is called just before starting to compile an C<eval STRING>, C<do
2274 FILE>, C<require> or C<use>, after the eval has been set up. I<o> is the
2275 OP that requested the eval, and will normally be an C<OP_ENTEREVAL>,
2276 C<OP_DOFILE> or C<OP_REQUIRE>.
2280 Once you have your hook functions, you need a C<BHK> structure to put
2281 them in. It's best to allocate it statically, since there is no way to
2282 free it once it's registered. The function pointers should be inserted
2283 into this structure using the C<BhkENTRY_set> macro, which will also set
2284 flags indicating which entries are valid. If you do need to allocate
2285 your C<BHK> dynamically for some reason, be sure to zero it before you
2288 Once registered, there is no mechanism to switch these hooks off, so if
2289 that is necessary you will need to do this yourself. An entry in C<%^H>
2290 is probably the best way, so the effect is lexically scoped; however it
2291 is also possible to use the C<BhkDISABLE> and C<BhkENABLE> macros to
2292 temporarily switch entries on and off. You should also be aware that
2293 generally speaking at least one scope will have opened before your
2294 extension is loaded, so you will see some C<pre>/C<post_end> pairs that
2295 didn't have a matching C<start>.
2297 =head1 Examining internal data structures with the C<dump> functions
2299 To aid debugging, the source file F<dump.c> contains a number of
2300 functions which produce formatted output of internal data structures.
2302 The most commonly used of these functions is C<Perl_sv_dump>; it's used
2303 for dumping SVs, AVs, HVs, and CVs. The C<Devel::Peek> module calls
2304 C<sv_dump> to produce debugging output from Perl-space, so users of that
2305 module should already be familiar with its format.
2307 C<Perl_op_dump> can be used to dump an C<OP> structure or any of its
2308 derivatives, and produces output similar to C<perl -Dx>; in fact,
2309 C<Perl_dump_eval> will dump the main root of the code being evaluated,
2310 exactly like C<-Dx>.
2312 Other useful functions are C<Perl_dump_sub>, which turns a C<GV> into an
2313 op tree, C<Perl_dump_packsubs> which calls C<Perl_dump_sub> on all the
2314 subroutines in a package like so: (Thankfully, these are all xsubs, so
2315 there is no op tree)
2317 (gdb) print Perl_dump_packsubs(PL_defstash)
2319 SUB attributes::bootstrap = (xsub 0x811fedc 0)
2321 SUB UNIVERSAL::can = (xsub 0x811f50c 0)
2323 SUB UNIVERSAL::isa = (xsub 0x811f304 0)
2325 SUB UNIVERSAL::VERSION = (xsub 0x811f7ac 0)
2327 SUB DynaLoader::boot_DynaLoader = (xsub 0x805b188 0)
2329 and C<Perl_dump_all>, which dumps all the subroutines in the stash and
2330 the op tree of the main root.
2332 =head1 How multiple interpreters and concurrency are supported
2334 =head2 Background and PERL_IMPLICIT_CONTEXT
2336 The Perl interpreter can be regarded as a closed box: it has an API
2337 for feeding it code or otherwise making it do things, but it also has
2338 functions for its own use. This smells a lot like an object, and
2339 there are ways for you to build Perl so that you can have multiple
2340 interpreters, with one interpreter represented either as a C structure,
2341 or inside a thread-specific structure. These structures contain all
2342 the context, the state of that interpreter.
2344 One macro controls the major Perl build flavor: MULTIPLICITY. The
2345 MULTIPLICITY build has a C structure that packages all the interpreter
2346 state. With multiplicity-enabled perls, PERL_IMPLICIT_CONTEXT is also
2347 normally defined, and enables the support for passing in a "hidden" first
2348 argument that represents all three data structures. MULTIPLICITY makes
2349 multi-threaded perls possible (with the ithreads threading model, related
2350 to the macro USE_ITHREADS.)
2352 Two other "encapsulation" macros are the PERL_GLOBAL_STRUCT and
2353 PERL_GLOBAL_STRUCT_PRIVATE (the latter turns on the former, and the
2354 former turns on MULTIPLICITY.) The PERL_GLOBAL_STRUCT causes all the
2355 internal variables of Perl to be wrapped inside a single global struct,
2356 struct perl_vars, accessible as (globals) &PL_Vars or PL_VarsPtr or
2357 the function Perl_GetVars(). The PERL_GLOBAL_STRUCT_PRIVATE goes
2358 one step further, there is still a single struct (allocated in main()
2359 either from heap or from stack) but there are no global data symbols
2360 pointing to it. In either case the global struct should be initialized
2361 as the very first thing in main() using Perl_init_global_struct() and
2362 correspondingly tear it down after perl_free() using Perl_free_global_struct(),
2363 please see F<miniperlmain.c> for usage details. You may also need
2364 to use C<dVAR> in your coding to "declare the global variables"
2365 when you are using them. dTHX does this for you automatically.
2367 =for apidoc Amnh||dVAR
2369 To see whether you have non-const data you can use a BSD (or GNU)
2372 nm libperl.a | grep -v ' [TURtr] '
2374 If this displays any C<D> or C<d> symbols (or possibly C<C> or C<c>),
2375 you have non-const data. The symbols the C<grep> removed are as follows:
2376 C<Tt> are I<text>, or code, the C<Rr> are I<read-only> (const) data,
2377 and the C<U> is <undefined>, external symbols referred to.
2379 The test F<t/porting/libperl.t> does this kind of symbol sanity
2380 checking on C<libperl.a>.
2382 For backward compatibility reasons defining just PERL_GLOBAL_STRUCT
2383 doesn't actually hide all symbols inside a big global struct: some
2384 PerlIO_xxx vtables are left visible. The PERL_GLOBAL_STRUCT_PRIVATE
2385 then hides everything (see how the PERLIO_FUNCS_DECL is used).
2387 All this obviously requires a way for the Perl internal functions to be
2388 either subroutines taking some kind of structure as the first
2389 argument, or subroutines taking nothing as the first argument. To
2390 enable these two very different ways of building the interpreter,
2391 the Perl source (as it does in so many other situations) makes heavy
2392 use of macros and subroutine naming conventions.
2394 First problem: deciding which functions will be public API functions and
2395 which will be private. All functions whose names begin C<S_> are private
2396 (think "S" for "secret" or "static"). All other functions begin with
2397 "Perl_", but just because a function begins with "Perl_" does not mean it is
2398 part of the API. (See L</Internal
2399 Functions>.) The easiest way to be B<sure> a
2400 function is part of the API is to find its entry in L<perlapi>.
2401 If it exists in L<perlapi>, it's part of the API. If it doesn't, and you
2402 think it should be (i.e., you need it for your extension), submit an issue at
2403 L<https://github.com/Perl/perl5/issues> explaining why you think it should be.
2405 Second problem: there must be a syntax so that the same subroutine
2406 declarations and calls can pass a structure as their first argument,
2407 or pass nothing. To solve this, the subroutines are named and
2408 declared in a particular way. Here's a typical start of a static
2409 function used within the Perl guts:
2412 S_incline(pTHX_ char *s)
2414 STATIC becomes "static" in C, and may be #define'd to nothing in some
2415 configurations in the future.
2417 A public function (i.e. part of the internal API, but not necessarily
2418 sanctioned for use in extensions) begins like this:
2421 Perl_sv_setiv(pTHX_ SV* dsv, IV num)
2423 C<pTHX_> is one of a number of macros (in F<perl.h>) that hide the
2424 details of the interpreter's context. THX stands for "thread", "this",
2425 or "thingy", as the case may be. (And no, George Lucas is not involved. :-)
2426 The first character could be 'p' for a B<p>rototype, 'a' for B<a>rgument,
2427 or 'd' for B<d>eclaration, so we have C<pTHX>, C<aTHX> and C<dTHX>, and
2430 =for apidoc Amnh||aTHX
2431 =for apidoc Amnh||aTHX_
2432 =for apidoc Amnh||dTHX
2433 =for apidoc Amnh||pTHX
2434 =for apidoc Amnh||pTHX_
2436 When Perl is built without options that set PERL_IMPLICIT_CONTEXT, there is no
2437 first argument containing the interpreter's context. The trailing underscore
2438 in the pTHX_ macro indicates that the macro expansion needs a comma
2439 after the context argument because other arguments follow it. If
2440 PERL_IMPLICIT_CONTEXT is not defined, pTHX_ will be ignored, and the
2441 subroutine is not prototyped to take the extra argument. The form of the
2442 macro without the trailing underscore is used when there are no additional
2445 When a core function calls another, it must pass the context. This
2446 is normally hidden via macros. Consider C<sv_setiv>. It expands into
2447 something like this:
2449 #ifdef PERL_IMPLICIT_CONTEXT
2450 #define sv_setiv(a,b) Perl_sv_setiv(aTHX_ a, b)
2451 /* can't do this for vararg functions, see below */
2453 #define sv_setiv Perl_sv_setiv
2456 This works well, and means that XS authors can gleefully write:
2460 and still have it work under all the modes Perl could have been
2463 This doesn't work so cleanly for varargs functions, though, as macros
2464 imply that the number of arguments is known in advance. Instead we
2465 either need to spell them out fully, passing C<aTHX_> as the first
2466 argument (the Perl core tends to do this with functions like
2467 Perl_warner), or use a context-free version.
2469 The context-free version of Perl_warner is called
2470 Perl_warner_nocontext, and does not take the extra argument. Instead
2471 it does dTHX; to get the context from thread-local storage. We
2472 C<#define warner Perl_warner_nocontext> so that extensions get source
2473 compatibility at the expense of performance. (Passing an arg is
2474 cheaper than grabbing it from thread-local storage.)
2476 You can ignore [pad]THXx when browsing the Perl headers/sources.
2477 Those are strictly for use within the core. Extensions and embedders
2478 need only be aware of [pad]THX.
2480 =head2 So what happened to dTHR?
2482 =for apidoc Amnh||dTHR
2484 C<dTHR> was introduced in perl 5.005 to support the older thread model.
2485 The older thread model now uses the C<THX> mechanism to pass context
2486 pointers around, so C<dTHR> is not useful any more. Perl 5.6.0 and
2487 later still have it for backward source compatibility, but it is defined
2490 =head2 How do I use all this in extensions?
2492 When Perl is built with PERL_IMPLICIT_CONTEXT, extensions that call
2493 any functions in the Perl API will need to pass the initial context
2494 argument somehow. The kicker is that you will need to write it in
2495 such a way that the extension still compiles when Perl hasn't been
2496 built with PERL_IMPLICIT_CONTEXT enabled.
2498 There are three ways to do this. First, the easy but inefficient way,
2499 which is also the default, in order to maintain source compatibility
2500 with extensions: whenever F<XSUB.h> is #included, it redefines the aTHX
2501 and aTHX_ macros to call a function that will return the context.
2502 Thus, something like:
2506 in your extension will translate to this when PERL_IMPLICIT_CONTEXT is
2509 Perl_sv_setiv(Perl_get_context(), sv, num);
2511 or to this otherwise:
2513 Perl_sv_setiv(sv, num);
2515 You don't have to do anything new in your extension to get this; since
2516 the Perl library provides Perl_get_context(), it will all just
2519 The second, more efficient way is to use the following template for
2522 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2527 STATIC void my_private_function(int arg1, int arg2);
2530 my_private_function(int arg1, int arg2)
2532 dTHX; /* fetch context */
2533 ... call many Perl API functions ...
2538 MODULE = Foo PACKAGE = Foo
2546 my_private_function(arg, 10);
2548 Note that the only two changes from the normal way of writing an
2549 extension is the addition of a C<#define PERL_NO_GET_CONTEXT> before
2550 including the Perl headers, followed by a C<dTHX;> declaration at
2551 the start of every function that will call the Perl API. (You'll
2552 know which functions need this, because the C compiler will complain
2553 that there's an undeclared identifier in those functions.) No changes
2554 are needed for the XSUBs themselves, because the XS() macro is
2555 correctly defined to pass in the implicit context if needed.
2557 The third, even more efficient way is to ape how it is done within
2561 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2566 /* pTHX_ only needed for functions that call Perl API */
2567 STATIC void my_private_function(pTHX_ int arg1, int arg2);
2570 my_private_function(pTHX_ int arg1, int arg2)
2572 /* dTHX; not needed here, because THX is an argument */
2573 ... call Perl API functions ...
2578 MODULE = Foo PACKAGE = Foo
2586 my_private_function(aTHX_ arg, 10);
2588 This implementation never has to fetch the context using a function
2589 call, since it is always passed as an extra argument. Depending on
2590 your needs for simplicity or efficiency, you may mix the previous
2591 two approaches freely.
2593 Never add a comma after C<pTHX> yourself--always use the form of the
2594 macro with the underscore for functions that take explicit arguments,
2595 or the form without the argument for functions with no explicit arguments.
2597 If one is compiling Perl with the C<-DPERL_GLOBAL_STRUCT> the C<dVAR>
2598 definition is needed if the Perl global variables (see F<perlvars.h>
2599 or F<globvar.sym>) are accessed in the function and C<dTHX> is not
2600 used (the C<dTHX> includes the C<dVAR> if necessary). One notices
2601 the need for C<dVAR> only with the said compile-time define, because
2602 otherwise the Perl global variables are visible as-is.
2604 =head2 Should I do anything special if I call perl from multiple threads?
2606 If you create interpreters in one thread and then proceed to call them in
2607 another, you need to make sure perl's own Thread Local Storage (TLS) slot is
2608 initialized correctly in each of those threads.
2610 The C<perl_alloc> and C<perl_clone> API functions will automatically set
2611 the TLS slot to the interpreter they created, so that there is no need to do
2612 anything special if the interpreter is always accessed in the same thread that
2613 created it, and that thread did not create or call any other interpreters
2614 afterwards. If that is not the case, you have to set the TLS slot of the
2615 thread before calling any functions in the Perl API on that particular
2616 interpreter. This is done by calling the C<PERL_SET_CONTEXT> macro in that
2617 thread as the first thing you do:
2619 /* do this before doing anything else with some_perl */
2620 PERL_SET_CONTEXT(some_perl);
2622 ... other Perl API calls on some_perl go here ...
2624 =head2 Future Plans and PERL_IMPLICIT_SYS
2626 Just as PERL_IMPLICIT_CONTEXT provides a way to bundle up everything
2627 that the interpreter knows about itself and pass it around, so too are
2628 there plans to allow the interpreter to bundle up everything it knows
2629 about the environment it's running on. This is enabled with the
2630 PERL_IMPLICIT_SYS macro. Currently it only works with USE_ITHREADS on
2633 This allows the ability to provide an extra pointer (called the "host"
2634 environment) for all the system calls. This makes it possible for
2635 all the system stuff to maintain their own state, broken down into
2636 seven C structures. These are thin wrappers around the usual system
2637 calls (see F<win32/perllib.c>) for the default perl executable, but for a
2638 more ambitious host (like the one that would do fork() emulation) all
2639 the extra work needed to pretend that different interpreters are
2640 actually different "processes", would be done here.
2642 The Perl engine/interpreter and the host are orthogonal entities.
2643 There could be one or more interpreters in a process, and one or
2644 more "hosts", with free association between them.
2646 =head1 Internal Functions
2648 All of Perl's internal functions which will be exposed to the outside
2649 world are prefixed by C<Perl_> so that they will not conflict with XS
2650 functions or functions used in a program in which Perl is embedded.
2651 Similarly, all global variables begin with C<PL_>. (By convention,
2652 static functions start with C<S_>.)
2654 Inside the Perl core (C<PERL_CORE> defined), you can get at the functions
2655 either with or without the C<Perl_> prefix, thanks to a bunch of defines
2656 that live in F<embed.h>. Note that extension code should I<not> set
2657 C<PERL_CORE>; this exposes the full perl internals, and is likely to cause
2658 breakage of the XS in each new perl release.
2660 The file F<embed.h> is generated automatically from
2661 F<embed.pl> and F<embed.fnc>. F<embed.pl> also creates the prototyping
2662 header files for the internal functions, generates the documentation
2663 and a lot of other bits and pieces. It's important that when you add
2664 a new function to the core or change an existing one, you change the
2665 data in the table in F<embed.fnc> as well. Here's a sample entry from
2668 Apd |SV** |av_fetch |AV* ar|I32 key|I32 lval
2670 The first column is a set of flags, the second column the return type,
2671 the third column the name. Columns after that are the arguments.
2672 The flags are documented at the top of F<embed.fnc>.
2674 If you edit F<embed.pl> or F<embed.fnc>, you will need to run
2675 C<make regen_headers> to force a rebuild of F<embed.h> and other
2676 auto-generated files.
2678 =head2 Formatted Printing of IVs, UVs, and NVs
2680 If you are printing IVs, UVs, or NVS instead of the stdio(3) style
2681 formatting codes like C<%d>, C<%ld>, C<%f>, you should use the
2682 following macros for portability
2687 UVxf UV in hexadecimal
2692 =for apidoc Amnh||IVdf
2693 =for apidoc Amnh||UVuf
2694 =for apidoc Amnh||UVof
2695 =for apidoc Amnh||UVxf
2696 =for apidoc Amnh||NVef
2697 =for apidoc Amnh||NVff
2698 =for apidoc Amnh||NVgf
2700 These will take care of 64-bit integers and long doubles.
2703 printf("IV is %" IVdf "\n", iv);
2705 The C<IVdf> will expand to whatever is the correct format for the IVs.
2706 Note that the spaces are required around the format in case the code is
2707 compiled with C++, to maintain compliance with its standard.
2709 Note that there are different "long doubles": Perl will use
2710 whatever the compiler has.
2712 If you are printing addresses of pointers, use %p or UVxf combined
2715 =head2 Formatted Printing of SVs
2717 The contents of SVs may be printed using the C<SVf> format, like so:
2719 Perl_croak(aTHX_ "This croaked because: %" SVf "\n", SvfARG(err_msg))
2721 where C<err_msg> is an SV.
2723 =for apidoc Amnh||SVf
2724 =for apidoc Amh||SVfARG|SV *sv
2726 Not all scalar types are printable. Simple values certainly are: one of
2727 IV, UV, NV, or PV. Also, if the SV is a reference to some value,
2728 either it will be dereferenced and the value printed, or information
2729 about the type of that value and its address are displayed. The results
2730 of printing any other type of SV are undefined and likely to lead to an
2731 interpreter crash. NVs are printed using a C<%g>-ish format.
2733 Note that the spaces are required around the C<SVf> in case the code is
2734 compiled with C++, to maintain compliance with its standard.
2736 Note that any filehandle being printed to under UTF-8 must be expecting
2737 UTF-8 in order to get good results and avoid Wide-character warnings.
2738 One way to do this for typical filehandles is to invoke perl with the
2739 C<-C>> parameter. (See L<perlrun/-C [numberE<sol>list]>.
2741 You can use this to concatenate two scalars:
2743 SV *var1 = get_sv("var1", GV_ADD);
2744 SV *var2 = get_sv("var2", GV_ADD);
2745 SV *var3 = newSVpvf("var1=%" SVf " and var2=%" SVf,
2746 SVfARG(var1), SVfARG(var2));
2748 =head2 Formatted Printing of Strings
2750 If you just want the bytes printed in a 7bit NUL-terminated string, you can
2751 just use C<%s> (assuming they are all really only 7bit). But if there is a
2752 possibility the value will be encoded as UTF-8 or contains bytes above
2753 C<0x7F> (and therefore 8bit), you should instead use the C<UTF8f> format.
2754 And as its parameter, use the C<UTF8fARG()> macro:
2758 /* U+2018: \xE2\x80\x98 LEFT SINGLE QUOTATION MARK
2759 U+2019: \xE2\x80\x99 RIGHT SINGLE QUOTATION MARK */
2761 msg = "\xE2\x80\x98Uses fancy quotes\xE2\x80\x99";
2763 msg = "'Uses simple quotes'";
2765 Perl_croak(aTHX_ "The message is: %" UTF8f "\n",
2766 UTF8fARG(can_utf8, strlen(msg), msg));
2768 The first parameter to C<UTF8fARG> is a boolean: 1 if the string is in
2769 UTF-8; 0 if string is in native byte encoding (Latin1).
2770 The second parameter is the number of bytes in the string to print.
2771 And the third and final parameter is a pointer to the first byte in the
2774 Note that any filehandle being printed to under UTF-8 must be expecting
2775 UTF-8 in order to get good results and avoid Wide-character warnings.
2776 One way to do this for typical filehandles is to invoke perl with the
2777 C<-C>> parameter. (See L<perlrun/-C [numberE<sol>list]>.
2779 =head2 Formatted Printing of C<Size_t> and C<SSize_t>
2781 The most general way to do this is to cast them to a UV or IV, and
2783 L<previous section|/Formatted Printing of IVs, UVs, and NVs>.
2785 But if you're using C<PerlIO_printf()>, it's less typing and visual
2786 clutter to use the C<%z> length modifier (for I<siZe>):
2788 PerlIO_printf("STRLEN is %zu\n", len);
2790 This modifier is not portable, so its use should be restricted to
2793 =head2 Pointer-To-Integer and Integer-To-Pointer
2795 Because pointer size does not necessarily equal integer size,
2796 use the follow macros to do it right.
2801 INT2PTR(pointertotype, integer)
2803 =for apidoc Amh|void *|INT2PTR|type|int value
2804 =for apidoc Amh|UV|PTR2UV|void *
2805 =for apidoc Amh|IV|PTR2IV|void *
2806 =for apidoc Amh|NV|PTR2NV|void *
2811 SV *sv = INT2PTR(SV*, iv);
2818 =head2 Exception Handling
2820 There are a couple of macros to do very basic exception handling in XS
2821 modules. You have to define C<NO_XSLOCKS> before including F<XSUB.h> to
2822 be able to use these macros:
2827 You can use these macros if you call code that may croak, but you need
2828 to do some cleanup before giving control back to Perl. For example:
2830 dXCPT; /* set up necessary variables */
2833 code_that_may_croak();
2838 /* do cleanup here */
2842 Note that you always have to rethrow an exception that has been
2843 caught. Using these macros, it is not possible to just catch the
2844 exception and ignore it. If you have to ignore the exception, you
2845 have to use the C<call_*> function.
2847 The advantage of using the above macros is that you don't have
2848 to setup an extra function for C<call_*>, and that using these
2849 macros is faster than using C<call_*>.
2851 =head2 Source Documentation
2853 There's an effort going on to document the internal functions and
2854 automatically produce reference manuals from them -- L<perlapi> is one
2855 such manual which details all the functions which are available to XS
2856 writers. L<perlintern> is the autogenerated manual for the functions
2857 which are not part of the API and are supposedly for internal use only.
2861 The following is an example and shouldn't be read as a real apidoc line
2863 Source documentation is created by putting POD comments into the C
2867 =for apidoc sv_setiv
2869 Copies an integer into the given SV. Does not handle 'set' magic. See
2870 L<perlapi/sv_setiv_mg>.
2875 Please try and supply some documentation if you add functions to the
2878 =head2 Backwards compatibility
2880 The Perl API changes over time. New functions are
2881 added or the interfaces of existing functions are
2882 changed. The C<Devel::PPPort> module tries to
2883 provide compatibility code for some of these changes, so XS writers don't
2884 have to code it themselves when supporting multiple versions of Perl.
2886 C<Devel::PPPort> generates a C header file F<ppport.h> that can also
2887 be run as a Perl script. To generate F<ppport.h>, run:
2889 perl -MDevel::PPPort -eDevel::PPPort::WriteFile
2891 Besides checking existing XS code, the script can also be used to retrieve
2892 compatibility information for various API calls using the C<--api-info>
2893 command line switch. For example:
2895 % perl ppport.h --api-info=sv_magicext
2897 For details, see C<perldoc ppport.h>.
2899 =head1 Unicode Support
2901 Perl 5.6.0 introduced Unicode support. It's important for porters and XS
2902 writers to understand this support and make sure that the code they
2903 write does not corrupt Unicode data.
2905 =head2 What B<is> Unicode, anyway?
2907 In the olden, less enlightened times, we all used to use ASCII. Most of
2908 us did, anyway. The big problem with ASCII is that it's American. Well,
2909 no, that's not actually the problem; the problem is that it's not
2910 particularly useful for people who don't use the Roman alphabet. What
2911 used to happen was that particular languages would stick their own
2912 alphabet in the upper range of the sequence, between 128 and 255. Of
2913 course, we then ended up with plenty of variants that weren't quite
2914 ASCII, and the whole point of it being a standard was lost.
2916 Worse still, if you've got a language like Chinese or
2917 Japanese that has hundreds or thousands of characters, then you really
2918 can't fit them into a mere 256, so they had to forget about ASCII
2919 altogether, and build their own systems using pairs of numbers to refer
2922 To fix this, some people formed Unicode, Inc. and
2923 produced a new character set containing all the characters you can
2924 possibly think of and more. There are several ways of representing these
2925 characters, and the one Perl uses is called UTF-8. UTF-8 uses
2926 a variable number of bytes to represent a character. You can learn more
2927 about Unicode and Perl's Unicode model in L<perlunicode>.
2929 (On EBCDIC platforms, Perl uses instead UTF-EBCDIC, which is a form of
2930 UTF-8 adapted for EBCDIC platforms. Below, we just talk about UTF-8.
2931 UTF-EBCDIC is like UTF-8, but the details are different. The macros
2932 hide the differences from you, just remember that the particular numbers
2933 and bit patterns presented below will differ in UTF-EBCDIC.)
2935 =head2 How can I recognise a UTF-8 string?
2937 You can't. This is because UTF-8 data is stored in bytes just like
2938 non-UTF-8 data. The Unicode character 200, (C<0xC8> for you hex types)
2939 capital E with a grave accent, is represented by the two bytes
2940 C<v196.172>. Unfortunately, the non-Unicode string C<chr(196).chr(172)>
2941 has that byte sequence as well. So you can't tell just by looking -- this
2942 is what makes Unicode input an interesting problem.
2944 In general, you either have to know what you're dealing with, or you
2945 have to guess. The API function C<is_utf8_string> can help; it'll tell
2946 you if a string contains only valid UTF-8 characters, and the chances
2947 of a non-UTF-8 string looking like valid UTF-8 become very small very
2948 quickly with increasing string length. On a character-by-character
2949 basis, C<isUTF8_CHAR>
2950 will tell you whether the current character in a string is valid UTF-8.
2952 =head2 How does UTF-8 represent Unicode characters?
2954 As mentioned above, UTF-8 uses a variable number of bytes to store a
2955 character. Characters with values 0...127 are stored in one
2956 byte, just like good ol' ASCII. Character 128 is stored as
2957 C<v194.128>; this continues up to character 191, which is
2958 C<v194.191>. Now we've run out of bits (191 is binary
2959 C<10111111>) so we move on; character 192 is C<v195.128>. And
2960 so it goes on, moving to three bytes at character 2048.
2961 L<perlunicode/Unicode Encodings> has pictures of how this works.
2963 Assuming you know you're dealing with a UTF-8 string, you can find out
2964 how long the first character in it is with the C<UTF8SKIP> macro:
2966 char *utf = "\305\233\340\240\201";
2969 len = UTF8SKIP(utf); /* len is 2 here */
2971 len = UTF8SKIP(utf); /* len is 3 here */
2973 Another way to skip over characters in a UTF-8 string is to use
2974 C<utf8_hop>, which takes a string and a number of characters to skip
2975 over. You're on your own about bounds checking, though, so don't use it
2978 All bytes in a multi-byte UTF-8 character will have the high bit set,
2979 so you can test if you need to do something special with this
2980 character like this (the C<UTF8_IS_INVARIANT()> is a macro that tests
2981 whether the byte is encoded as a single byte even in UTF-8):
2983 U8 *utf; /* Initialize this to point to the beginning of the
2984 sequence to convert */
2985 U8 *utf_end; /* Initialize this to 1 beyond the end of the sequence
2986 pointed to by 'utf' */
2987 UV uv; /* Returned code point; note: a UV, not a U8, not a
2989 STRLEN len; /* Returned length of character in bytes */
2991 if (!UTF8_IS_INVARIANT(*utf))
2992 /* Must treat this as UTF-8 */
2993 uv = utf8_to_uvchr_buf(utf, utf_end, &len);
2995 /* OK to treat this character as a byte */
2998 You can also see in that example that we use C<utf8_to_uvchr_buf> to get the
2999 value of the character; the inverse function C<uvchr_to_utf8> is available
3000 for putting a UV into UTF-8:
3002 if (!UVCHR_IS_INVARIANT(uv))
3003 /* Must treat this as UTF8 */
3004 utf8 = uvchr_to_utf8(utf8, uv);
3006 /* OK to treat this character as a byte */
3009 You B<must> convert characters to UVs using the above functions if
3010 you're ever in a situation where you have to match UTF-8 and non-UTF-8
3011 characters. You may not skip over UTF-8 characters in this case. If you
3012 do this, you'll lose the ability to match hi-bit non-UTF-8 characters;
3013 for instance, if your UTF-8 string contains C<v196.172>, and you skip
3014 that character, you can never match a C<chr(200)> in a non-UTF-8 string.
3017 (Note that we don't have to test for invariant characters in the
3018 examples above. The functions work on any well-formed UTF-8 input.
3019 It's just that its faster to avoid the function overhead when it's not
3022 =head2 How does Perl store UTF-8 strings?
3024 Currently, Perl deals with UTF-8 strings and non-UTF-8 strings
3025 slightly differently. A flag in the SV, C<SVf_UTF8>, indicates that the
3026 string is internally encoded as UTF-8. Without it, the byte value is the
3027 codepoint number and vice versa. This flag is only meaningful if the SV
3028 is C<SvPOK> or immediately after stringification via C<SvPV> or a
3029 similar macro. You can check and manipulate this flag with the
3036 This flag has an important effect on Perl's treatment of the string: if
3037 UTF-8 data is not properly distinguished, regular expressions,
3038 C<length>, C<substr> and other string handling operations will have
3039 undesirable (wrong) results.
3041 The problem comes when you have, for instance, a string that isn't
3042 flagged as UTF-8, and contains a byte sequence that could be UTF-8 --
3043 especially when combining non-UTF-8 and UTF-8 strings.
3045 Never forget that the C<SVf_UTF8> flag is separate from the PV value; you
3046 need to be sure you don't accidentally knock it off while you're
3047 manipulating SVs. More specifically, you cannot expect to do this:
3056 nsv = newSVpvn(p, len);
3058 The C<char*> string does not tell you the whole story, and you can't
3059 copy or reconstruct an SV just by copying the string value. Check if the
3060 old SV has the UTF8 flag set (I<after> the C<SvPV> call), and act
3064 is_utf8 = SvUTF8(sv);
3065 frobnicate(p, is_utf8);
3066 nsv = newSVpvn(p, len);
3070 In the above, your C<frobnicate> function has been changed to be made
3071 aware of whether or not it's dealing with UTF-8 data, so that it can
3072 handle the string appropriately.
3074 Since just passing an SV to an XS function and copying the data of
3075 the SV is not enough to copy the UTF8 flags, even less right is just
3076 passing a S<C<char *>> to an XS function.
3078 For full generality, use the L<C<DO_UTF8>|perlapi/DO_UTF8> macro to see if the
3079 string in an SV is to be I<treated> as UTF-8. This takes into account
3080 if the call to the XS function is being made from within the scope of
3081 L<S<C<use bytes>>|bytes>. If so, the underlying bytes that comprise the
3082 UTF-8 string are to be exposed, rather than the character they
3083 represent. But this pragma should only really be used for debugging and
3084 perhaps low-level testing at the byte level. Hence most XS code need
3085 not concern itself with this, but various areas of the perl core do need
3088 And this isn't the whole story. Starting in Perl v5.12, strings that
3089 aren't encoded in UTF-8 may also be treated as Unicode under various
3090 conditions (see L<perlunicode/ASCII Rules versus Unicode Rules>).
3091 This is only really a problem for characters whose ordinals are between
3092 128 and 255, and their behavior varies under ASCII versus Unicode rules
3093 in ways that your code cares about (see L<perlunicode/The "Unicode Bug">).
3094 There is no published API for dealing with this, as it is subject to
3095 change, but you can look at the code for C<pp_lc> in F<pp.c> for an
3096 example as to how it's currently done.
3098 =head2 How do I convert a string to UTF-8?
3100 If you're mixing UTF-8 and non-UTF-8 strings, it is necessary to upgrade
3101 the non-UTF-8 strings to UTF-8. If you've got an SV, the easiest way to do
3104 sv_utf8_upgrade(sv);
3106 However, you must not do this, for example:
3109 sv_utf8_upgrade(left);
3111 If you do this in a binary operator, you will actually change one of the
3112 strings that came into the operator, and, while it shouldn't be noticeable
3113 by the end user, it can cause problems in deficient code.
3115 Instead, C<bytes_to_utf8> will give you a UTF-8-encoded B<copy> of its
3116 string argument. This is useful for having the data available for
3117 comparisons and so on, without harming the original SV. There's also
3118 C<utf8_to_bytes> to go the other way, but naturally, this will fail if
3119 the string contains any characters above 255 that can't be represented
3122 =head2 How do I compare strings?
3124 L<perlapi/sv_cmp> and L<perlapi/sv_cmp_flags> do a lexigraphic
3125 comparison of two SV's, and handle UTF-8ness properly. Note, however,
3126 that Unicode specifies a much fancier mechanism for collation, available
3127 via the L<Unicode::Collate> module.
3129 To just compare two strings for equality/non-equality, you can just use
3130 L<C<memEQ()>|perlapi/memEQ> and L<C<memNE()>|perlapi/memEQ> as usual,
3131 except the strings must be both UTF-8 or not UTF-8 encoded.
3133 To compare two strings case-insensitively, use
3134 L<C<foldEQ_utf8()>|perlapi/foldEQ_utf8> (the strings don't have to have
3135 the same UTF-8ness).
3137 =head2 Is there anything else I need to know?
3139 Not really. Just remember these things:
3145 There's no way to tell if a S<C<char *>> or S<C<U8 *>> string is UTF-8
3146 or not. But you can tell if an SV is to be treated as UTF-8 by calling
3147 C<DO_UTF8> on it, after stringifying it with C<SvPV> or a similar
3148 macro. And, you can tell if SV is actually UTF-8 (even if it is not to
3149 be treated as such) by looking at its C<SvUTF8> flag (again after
3150 stringifying it). Don't forget to set the flag if something should be
3152 Treat the flag as part of the PV, even though it's not -- if you pass on
3153 the PV to somewhere, pass on the flag too.
3157 If a string is UTF-8, B<always> use C<utf8_to_uvchr_buf> to get at the value,
3158 unless C<UTF8_IS_INVARIANT(*s)> in which case you can use C<*s>.
3162 When writing a character UV to a UTF-8 string, B<always> use
3163 C<uvchr_to_utf8>, unless C<UVCHR_IS_INVARIANT(uv))> in which case
3164 you can use C<*s = uv>.
3168 Mixing UTF-8 and non-UTF-8 strings is
3169 tricky. Use C<bytes_to_utf8> to get
3170 a new string which is UTF-8 encoded, and then combine them.
3174 =head1 Custom Operators
3176 Custom operator support is an experimental feature that allows you to
3177 define your own ops. This is primarily to allow the building of
3178 interpreters for other languages in the Perl core, but it also allows
3179 optimizations through the creation of "macro-ops" (ops which perform the
3180 functions of multiple ops which are usually executed together, such as
3181 C<gvsv, gvsv, add>.)
3183 This feature is implemented as a new op type, C<OP_CUSTOM>. The Perl
3184 core does not "know" anything special about this op type, and so it will
3185 not be involved in any optimizations. This also means that you can
3186 define your custom ops to be any op structure -- unary, binary, list and
3189 It's important to know what custom operators won't do for you. They
3190 won't let you add new syntax to Perl, directly. They won't even let you
3191 add new keywords, directly. In fact, they won't change the way Perl
3192 compiles a program at all. You have to do those changes yourself, after
3193 Perl has compiled the program. You do this either by manipulating the op
3194 tree using a C<CHECK> block and the C<B::Generate> module, or by adding
3195 a custom peephole optimizer with the C<optimize> module.
3197 When you do this, you replace ordinary Perl ops with custom ops by
3198 creating ops with the type C<OP_CUSTOM> and the C<op_ppaddr> of your own
3199 PP function. This should be defined in XS code, and should look like
3200 the PP ops in C<pp_*.c>. You are responsible for ensuring that your op
3201 takes the appropriate number of values from the stack, and you are
3202 responsible for adding stack marks if necessary.
3204 You should also "register" your op with the Perl interpreter so that it
3205 can produce sensible error and warning messages. Since it is possible to
3206 have multiple custom ops within the one "logical" op type C<OP_CUSTOM>,
3207 Perl uses the value of C<< o->op_ppaddr >> to determine which custom op
3208 it is dealing with. You should create an C<XOP> structure for each
3209 ppaddr you use, set the properties of the custom op with
3210 C<XopENTRY_set>, and register the structure against the ppaddr using
3211 C<Perl_custom_op_register>. A trivial example might look like:
3214 static OP *my_pp(pTHX);
3217 XopENTRY_set(&my_xop, xop_name, "myxop");
3218 XopENTRY_set(&my_xop, xop_desc, "Useless custom op");
3219 Perl_custom_op_register(aTHX_ my_pp, &my_xop);
3221 The available fields in the structure are:
3227 A short name for your op. This will be included in some error messages,
3228 and will also be returned as C<< $op->name >> by the L<B|B> module, so
3229 it will appear in the output of module like L<B::Concise|B::Concise>.
3233 A short description of the function of the op.
3237 Which of the various C<*OP> structures this op uses. This should be one of
3238 the C<OA_*> constants from F<op.h>, namely
3258 =item OA_PVOP_OR_SVOP
3260 This should be interpreted as 'C<PVOP>' only. The C<_OR_SVOP> is because
3261 the only core C<PVOP>, C<OP_TRANS>, can sometimes be a C<SVOP> instead.
3269 The other C<OA_*> constants should not be used.
3273 This member is of type C<Perl_cpeep_t>, which expands to C<void
3274 (*Perl_cpeep_t)(aTHX_ OP *o, OP *oldop)>. If it is set, this function
3275 will be called from C<Perl_rpeep> when ops of this type are encountered
3276 by the peephole optimizer. I<o> is the OP that needs optimizing;
3277 I<oldop> is the previous OP optimized, whose C<op_next> points to I<o>.
3281 C<B::Generate> directly supports the creation of custom ops by name.
3285 Descriptions above occasionally refer to "the stack", but there are in fact
3286 many stack-like data structures within the perl interpreter. When otherwise
3287 unqualified, "the stack" usually refers to the value stack.
3289 The various stacks have different purposes, and operate in slightly different
3290 ways. Their differences are noted below.
3294 This stack stores the values that regular perl code is operating on, usually
3295 intermediate values of expressions within a statement. The stack itself is
3296 formed of an array of SV pointers.
3298 The base of this stack is pointed to by the interpreter variable
3299 C<PL_stack_base>, of type C<SV **>.
3301 The head of the stack is C<PL_stack_sp>, and points to the most
3302 recently-pushed item.
3304 Items are pushed to the stack by using the C<PUSHs()> macro or its variants
3305 described above; C<XPUSHs()>, C<mPUSHs()>, C<mXPUSHs()> and the typed
3306 versions. Note carefully that the non-C<X> versions of these macros do not
3307 check the size of the stack and assume it to be big enough. These must be
3308 paired with a suitable check of the stack's size, such as the C<EXTEND> macro
3309 to ensure it is large enough. For example
3317 This is slightly more performant than making four separate checks in four
3318 separate C<mXPUSHi()> calls.
3320 As a further performance optimisation, the various C<PUSH> macros all operate
3321 using a local variable C<SP>, rather than the interpreter-global variable
3322 C<PL_stack_sp>. This variable is declared by the C<dSP> macro - though it is
3323 normally implied by XSUBs and similar so it is rare you have to consider it
3324 directly. Once declared, the C<PUSH> macros will operate only on this local
3325 variable, so before invoking any other perl core functions you must use the
3326 C<PUTBACK> macro to return the value from the local C<SP> variable back to
3327 the interpreter variable. Similarly, after calling a perl core function which
3328 may have had reason to move the stack or push/pop values to it, you must use
3329 the C<SPAGAIN> macro which refreshes the local C<SP> value back from the
3332 Items are popped from the stack by using the C<POPs> macro or its typed
3333 versions, There is also a macro C<TOPs> that inspects the topmost item without
3336 Note specifically that SV pointers on the value stack do not contribute to the
3337 overall reference count of the xVs being referred to. If newly-created xVs are
3338 being pushed to the stack you must arrange for them to be destroyed at a
3339 suitable time; usually by using one of the C<mPUSH*> macros or C<sv_2mortal()>
3340 to mortalise the xV.
3344 The value stack stores individual perl scalar values as temporaries between
3345 expressions. Some perl expressions operate on entire lists; for that purpose
3346 we need to know where on the stack each list begins. This is the purpose of the
3349 The mark stack stores integers as I32 values, which are the height of the
3350 value stack at the time before the list began; thus the mark itself actually
3351 points to the value stack entry one before the list. The list itself starts at
3354 The base of this stack is pointed to by the interpreter variable
3355 C<PL_markstack>, of type C<I32 *>.
3357 The head of the stack is C<PL_markstack_ptr>, and points to the most
3358 recently-pushed item.
3360 Items are pushed to the stack by using the C<PUSHMARK()> macro. Even though
3361 the stack itself stores (value) stack indices as integers, the C<PUSHMARK>
3362 macro should be given a stack pointer directly; it will calculate the index
3363 offset by comparing to the C<PL_stack_sp> variable. Thus almost always the
3364 code to perform this is
3368 Items are popped from the stack by the C<POPMARK> macro. There is also a macro
3369 C<TOPMARK> that inspects the topmost item without removing it. These macros
3370 return I32 index values directly. There is also the C<dMARK> macro which
3371 declares a new SV double-pointer variable, called C<mark>, which points at the
3372 marked stack slot; this is the usual macro that C code will use when operating
3373 on lists given on the stack.
3375 As noted above, the C<mark> variable itself will point at the most recently
3376 pushed value on the value stack before the list begins, and so the list itself
3377 starts at C<mark + 1>. The values of the list may be iterated by code such as
3379 for(SV **svp = mark + 1; svp <= PL_stack_sp; svp++) {
3384 Note specifically in the case that the list is already empty, C<mark> will
3385 equal C<PL_stack_sp>.
3387 Because the C<mark> variable is converted to a pointer on the value stack,
3388 extra care must be taken if C<EXTEND> or any of the C<XPUSH> macros are
3389 invoked within the function, because the stack may need to be moved to
3390 extend it and so the existing pointer will now be invalid. If this may be a
3391 problem, a possible solution is to track the mark offset as an integer and
3392 track the mark itself later on after the stack had been moved.
3394 I32 markoff = POPMARK;
3398 SP **mark = PL_stack_base + markoff;
3400 =head2 Temporaries Stack
3402 As noted above, xV references on the main value stack do not contribute to the
3403 reference count of an xV, and so another mechanism is used to track when
3404 temporary values which live on the stack must be released. This is the job of
3405 the temporaries stack.
3407 The temporaries stack stores pointers to xVs whose reference counts will be
3410 The base of this stack is pointed to by the interpreter variable
3411 C<PL_tmps_stack>, of type C<SV **>.
3413 The head of the stack is indexed by C<PL_tmps_ix>, an integer which stores the
3414 index in the array of the most recently-pushed item.
3416 There is no public API to directly push items to the temporaries stack. Instead,
3417 the API function C<sv_2mortal()> is used to mortalize an xV, adding its
3418 address to the temporaries stack.
3420 Likewise, there is no public API to read values from the temporaries stack.
3421 Instead. the macros C<SAVETMPS> and C<FREETPMS> are used. The C<SAVETMPS>
3422 macro establishes the base levels of the temporaries stack, by capturing the
3423 current value of C<PL_tmps_ix> into C<PL_tmps_floor> and saving the previous
3424 value to the save stack. Thereafter, whenever C<FREETMPS> is invoked all of
3425 the temporaries that have been pushed since that level are reclaimed.
3427 While it is common to see these two macros in pairs within an C<ENTER>/
3428 C<LEAVE> pair, it is not necessary to match them. It is permitted to invoke
3429 C<FREETMPS> multiple times since the most recent C<SAVETMPS>; for example in a
3430 loop iterating over elements of a list. While you can invoke C<SAVETMPS>
3431 multiple times within a scope pair, it is unlikely to be useful. Subsequent
3432 invocations will move the temporaries floor further up, thus effectively
3433 trapping the existing temporaries to only be released at the end of the scope.
3437 The save stack is used by perl to implement the C<local> keyword and other
3438 similar behaviours; any cleanup operations that need to be performed when
3439 leaving the current scope. Items pushed to this stack generally capture the
3440 current value of some internal variable or state, which will be restored when
3441 the scope is unwound due to leaving, C<return>, C<die>, C<goto> or other
3444 Whereas other perl internal stacks store individual items all of the same type
3445 (usually SV pointers or integers), the items pushed to the save stack are
3446 formed of many different types, having multiple fields to them. For example,
3447 the C<SAVEt_INT> type needs to store both the address of the C<int> variable
3448 to restore, and the value to restore it to. This information could have been
3449 stored using fields of a C<struct>, but would have to be large enough to store
3450 three pointers in the largest case, which would waste a lot of space in most
3451 of the smaller cases.
3453 Instead, the stack stores information in a variable-length encoding of C<ANY>
3454 structures. The final value pushed is stored in the C<UV> field which encodes
3455 the kind of item held by the preceeding items; the count and types of which
3456 will depend on what kind of item is being stored. The kind field is pushed
3457 last because that will be the first field to be popped when unwinding items
3460 The base of this stack is pointed to by the interpreter variable
3461 C<PL_savestack>, of type C<ANY *>.
3463 The head of the stack is indexed by C<PL_savestack_ix>, an integer which
3464 stores the index in the array at which the next item should be pushed. (Note
3465 that this is different to most other stacks, which reference the most
3466 recently-pushed item).
3468 Items are pushed to the save stack by using the various C<SAVE...()> macros.
3469 Many of these macros take a variable and store both its address and current
3470 value on the save stack, ensuring that value gets restored on scope exit.
3478 There are also a variety of other special-purpose macros which save particular
3479 types or values of interest. C<SAVETMPS> has already been mentioned above.
3480 Others include C<SAVEFREEPV> which arranges for a PV (i.e. a string buffer) to
3481 be freed, or C<SAVEDESTRUCTOR> which arranges for a given function pointer to
3482 be invoked on scope exit. A full list of such macros can be found in
3485 There is no public API for popping individual values or items from the save
3486 stack. Instead, via the scope stack, the C<ENTER> and C<LEAVE> pair form a way
3487 to start and stop nested scopes. Leaving a nested scope via C<LEAVE> will
3488 restore all of the saved values that had been pushed since the most recent
3493 As with the mark stack to the value stack, the scope stack forms a pair with
3494 the save stack. The scope stack stores the height of the save stack at which
3495 nested scopes begin, and allows the save stack to be unwound back to that
3496 point when the scope is left.
3498 When perl is built with debugging enabled, there is a second part to this
3499 stack storing human-readable string names describing the type of stack
3500 context. Each push operation saves the name as well as the height of the save
3501 stack, and each pop operation checks the topmost name with what is expected,
3502 causing an assertion failure if the name does not match.
3504 The base of this stack is pointed to by the interpreter variable
3505 C<PL_scopestack>, of type C<I32 *>. If enabled, the scope stack names are
3506 stored in a separate array pointed to by C<PL_scopestack_name>, of type
3509 The head of the stack is indexed by C<PL_scopestack_ix>, an integer which
3510 stores the index of the array or arrays at which the next item should be
3511 pushed. (Note that this is different to most other stacks, which reference the
3512 most recently-pushed item).
3514 Values are pushed to the scope stack using the C<ENTER> macro, which begins a
3515 new nested scope. Any items pushed to the save stack are then restored at the
3516 next nested invocation of the C<LEAVE> macro.
3518 =head1 Dynamic Scope and the Context Stack
3520 B<Note:> this section describes a non-public internal API that is subject
3521 to change without notice.
3523 =head2 Introduction to the context stack
3525 In Perl, dynamic scoping refers to the runtime nesting of things like
3526 subroutine calls, evals etc, as well as the entering and exiting of block
3527 scopes. For example, the restoring of a C<local>ised variable is
3528 determined by the dynamic scope.
3530 Perl tracks the dynamic scope by a data structure called the context
3531 stack, which is an array of C<PERL_CONTEXT> structures, and which is
3532 itself a big union for all the types of context. Whenever a new scope is
3533 entered (such as a block, a C<for> loop, or a subroutine call), a new
3534 context entry is pushed onto the stack. Similarly when leaving a block or
3535 returning from a subroutine call etc. a context is popped. Since the
3536 context stack represents the current dynamic scope, it can be searched.
3537 For example, C<next LABEL> searches back through the stack looking for a
3538 loop context that matches the label; C<return> pops contexts until it
3539 finds a sub or eval context or similar; C<caller> examines sub contexts on
3542 Each context entry is labelled with a context type, C<cx_type>. Typical
3543 context types are C<CXt_SUB>, C<CXt_EVAL> etc., as well as C<CXt_BLOCK>
3544 and C<CXt_NULL> which represent a basic scope (as pushed by C<pp_enter>)
3545 and a sort block. The type determines which part of the context union are
3548 The main division in the context struct is between a substitution scope
3549 (C<CXt_SUBST>) and block scopes, which are everything else. The former is
3550 just used while executing C<s///e>, and won't be discussed further
3553 All the block scope types share a common base, which corresponds to
3554 C<CXt_BLOCK>. This stores the old values of various scope-related
3555 variables like C<PL_curpm>, as well as information about the current
3556 scope, such as C<gimme>. On scope exit, the old variables are restored.
3558 Particular block scope types store extra per-type information. For
3559 example, C<CXt_SUB> stores the currently executing CV, while the various
3560 for loop types might hold the original loop variable SV. On scope exit,
3561 the per-type data is processed; for example the CV has its reference count
3562 decremented, and the original loop variable is restored.
3564 The macro C<cxstack> returns the base of the current context stack, while
3565 C<cxstack_ix> is the index of the current frame within that stack.
3567 In fact, the context stack is actually part of a stack-of-stacks system;
3568 whenever something unusual is done such as calling a C<DESTROY> or tie
3569 handler, a new stack is pushed, then popped at the end.
3571 Note that the API described here changed considerably in perl 5.24; prior
3572 to that, big macros like C<PUSHBLOCK> and C<POPSUB> were used; in 5.24
3573 they were replaced by the inline static functions described below. In
3574 addition, the ordering and detail of how these macros/function work
3575 changed in many ways, often subtly. In particular they didn't handle
3576 saving the savestack and temps stack positions, and required additional
3577 C<ENTER>, C<SAVETMPS> and C<LEAVE> compared to the new functions. The
3578 old-style macros will not be described further.
3581 =head2 Pushing contexts
3583 For pushing a new context, the two basic functions are
3584 C<cx = cx_pushblock()>, which pushes a new basic context block and returns
3585 its address, and a family of similar functions with names like
3586 C<cx_pushsub(cx)> which populate the additional type-dependent fields in
3587 the C<cx> struct. Note that C<CXt_NULL> and C<CXt_BLOCK> don't have their
3588 own push functions, as they don't store any data beyond that pushed by
3591 The fields of the context struct and the arguments to the C<cx_*>
3592 functions are subject to change between perl releases, representing
3593 whatever is convenient or efficient for that release.
3595 A typical context stack pushing can be found in C<pp_entersub>; the
3596 following shows a simplified and stripped-down example of a non-XS call,
3597 along with comments showing roughly what each function does.
3601 bool hasargs = cBOOL(PL_op->op_flags & OPf_STACKED);
3602 OP *retop = PL_op->op_next;
3603 I32 old_ss_ix = PL_savestack_ix;
3606 /* ... make mortal copies of stack args which are PADTMPs here ... */
3608 /* ... do any additional savestack pushes here ... */
3610 /* Now push a new context entry of type 'CXt_SUB'; initially just
3611 * doing the actions common to all block types: */
3613 cx = cx_pushblock(CXt_SUB, gimme, MARK, old_ss_ix);
3615 /* this does (approximately):
3616 CXINC; /* cxstack_ix++ (grow if necessary) */
3617 cx = CX_CUR(); /* and get the address of new frame */
3618 cx->cx_type = CXt_SUB;
3619 cx->blk_gimme = gimme;
3620 cx->blk_oldsp = MARK - PL_stack_base;
3621 cx->blk_oldsaveix = old_ss_ix;
3622 cx->blk_oldcop = PL_curcop;
3623 cx->blk_oldmarksp = PL_markstack_ptr - PL_markstack;
3624 cx->blk_oldscopesp = PL_scopestack_ix;
3625 cx->blk_oldpm = PL_curpm;
3626 cx->blk_old_tmpsfloor = PL_tmps_floor;
3628 PL_tmps_floor = PL_tmps_ix;
3632 /* then update the new context frame with subroutine-specific info,
3633 * such as the CV about to be executed: */
3635 cx_pushsub(cx, cv, retop, hasargs);
3637 /* this does (approximately):
3638 cx->blk_sub.cv = cv;
3639 cx->blk_sub.olddepth = CvDEPTH(cv);
3640 cx->blk_sub.prevcomppad = PL_comppad;
3641 cx->cx_type |= (hasargs) ? CXp_HASARGS : 0;
3642 cx->blk_sub.retop = retop;
3643 SvREFCNT_inc_simple_void_NN(cv);
3646 Note that C<cx_pushblock()> sets two new floors: for the args stack (to
3647 C<MARK>) and the temps stack (to C<PL_tmps_ix>). While executing at this
3648 scope level, every C<nextstate> (amongst others) will reset the args and
3649 tmps stack levels to these floors. Note that since C<cx_pushblock> uses
3650 the current value of C<PL_tmps_ix> rather than it being passed as an arg,
3651 this dictates at what point C<cx_pushblock> should be called. In
3652 particular, any new mortals which should be freed only on scope exit
3653 (rather than at the next C<nextstate>) should be created first.
3655 Most callers of C<cx_pushblock> simply set the new args stack floor to the
3656 top of the previous stack frame, but for C<CXt_LOOP_LIST> it stores the
3657 items being iterated over on the stack, and so sets C<blk_oldsp> to the
3658 top of these items instead. Note that, contrary to its name, C<blk_oldsp>
3659 doesn't always represent the value to restore C<PL_stack_sp> to on scope
3662 Note the early capture of C<PL_savestack_ix> to C<old_ss_ix>, which is
3663 later passed as an arg to C<cx_pushblock>. In the case of C<pp_entersub>,
3664 this is because, although most values needing saving are stored in fields
3665 of the context struct, an extra value needs saving only when the debugger
3666 is running, and it doesn't make sense to bloat the struct for this rare
3667 case. So instead it is saved on the savestack. Since this value gets
3668 calculated and saved before the context is pushed, it is necessary to pass
3669 the old value of C<PL_savestack_ix> to C<cx_pushblock>, to ensure that the
3670 saved value gets freed during scope exit. For most users of
3671 C<cx_pushblock>, where nothing needs pushing on the save stack,
3672 C<PL_savestack_ix> is just passed directly as an arg to C<cx_pushblock>.
3674 Note that where possible, values should be saved in the context struct
3675 rather than on the save stack; it's much faster that way.
3677 Normally C<cx_pushblock> should be immediately followed by the appropriate
3678 C<cx_pushfoo>, with nothing between them; this is because if code
3679 in-between could die (e.g. a warning upgraded to fatal), then the context
3680 stack unwinding code in C<dounwind> would see (in the example above) a
3681 C<CXt_SUB> context frame, but without all the subroutine-specific fields
3682 set, and crashes would soon ensue.
3684 Where the two must be separate, initially set the type to C<CXt_NULL> or
3685 C<CXt_BLOCK>, and later change it to C<CXt_foo> when doing the
3686 C<cx_pushfoo>. This is exactly what C<pp_enteriter> does, once it's
3687 determined which type of loop it's pushing.
3689 =head2 Popping contexts
3691 Contexts are popped using C<cx_popsub()> etc. and C<cx_popblock()>. Note
3692 however, that unlike C<cx_pushblock>, neither of these functions actually
3693 decrement the current context stack index; this is done separately using
3696 There are two main ways that contexts are popped. During normal execution
3697 as scopes are exited, functions like C<pp_leave>, C<pp_leaveloop> and
3698 C<pp_leavesub> process and pop just one context using C<cx_popfoo> and
3699 C<cx_popblock>. On the other hand, things like C<pp_return> and C<next>
3700 may have to pop back several scopes until a sub or loop context is found,
3701 and exceptions (such as C<die>) need to pop back contexts until an eval
3702 context is found. Both of these are accomplished by C<dounwind()>, which
3703 is capable of processing and popping all contexts above the target one.
3705 Here is a typical example of context popping, as found in C<pp_leavesub>
3706 (simplified slightly):
3715 gimme = cx->blk_gimme;
3716 oldsp = PL_stack_base + cx->blk_oldsp; /* last arg of previous frame */
3718 if (gimme == G_VOID)
3719 PL_stack_sp = oldsp;
3721 leave_adjust_stacks(oldsp, oldsp, gimme, 0);
3726 retop = cx->blk_sub.retop;
3731 The steps above are in a very specific order, designed to be the reverse
3732 order of when the context was pushed. The first thing to do is to copy
3733 and/or protect any any return arguments and free any temps in the current
3734 scope. Scope exits like an rvalue sub normally return a mortal copy of
3735 their return args (as opposed to lvalue subs). It is important to make
3736 this copy before the save stack is popped or variables are restored, or
3737 bad things like the following can happen:
3739 sub f { my $x =...; $x } # $x freed before we get to copy it
3740 sub f { /(...)/; $1 } # PL_curpm restored before $1 copied
3742 Although we wish to free any temps at the same time, we have to be careful
3743 not to free any temps which are keeping return args alive; nor to free the
3744 temps we have just created while mortal copying return args. Fortunately,
3745 C<leave_adjust_stacks()> is capable of making mortal copies of return args,
3746 shifting args down the stack, and only processing those entries on the
3747 temps stack that are safe to do so.
3749 In void context no args are returned, so it's more efficient to skip
3750 calling C<leave_adjust_stacks()>. Also in void context, a C<nextstate> op
3751 is likely to be imminently called which will do a C<FREETMPS>, so there's
3752 no need to do that either.
3754 The next step is to pop savestack entries: C<CX_LEAVE_SCOPE(cx)> is just
3755 defined as C<< LEAVE_SCOPE(cx->blk_oldsaveix) >>. Note that during the
3756 popping, it's possible for perl to call destructors, call C<STORE> to undo
3757 localisations of tied vars, and so on. Any of these can die or call
3758 C<exit()>. In this case, C<dounwind()> will be called, and the current
3759 context stack frame will be re-processed. Thus it is vital that all steps
3760 in popping a context are done in such a way to support reentrancy. The
3761 other alternative, of decrementing C<cxstack_ix> I<before> processing the
3762 frame, would lead to leaks and the like if something died halfway through,
3763 or overwriting of the current frame.
3765 C<CX_LEAVE_SCOPE> itself is safely re-entrant: if only half the savestack
3766 items have been popped before dying and getting trapped by eval, then the
3767 C<CX_LEAVE_SCOPE>s in C<dounwind> or C<pp_leaveeval> will continue where
3768 the first one left off.
3770 The next step is the type-specific context processing; in this case
3771 C<cx_popsub>. In part, this looks like:
3773 cv = cx->blk_sub.cv;
3774 CvDEPTH(cv) = cx->blk_sub.olddepth;
3775 cx->blk_sub.cv = NULL;
3778 where its processing the just-executed CV. Note that before it decrements
3779 the CV's reference count, it nulls the C<blk_sub.cv>. This means that if
3780 it re-enters, the CV won't be freed twice. It also means that you can't
3781 rely on such type-specific fields having useful values after the return
3784 Next, C<cx_popblock> restores all the various interpreter vars to their
3785 previous values or previous high water marks; it expands to:
3787 PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp;
3788 PL_scopestack_ix = cx->blk_oldscopesp;
3789 PL_curpm = cx->blk_oldpm;
3790 PL_curcop = cx->blk_oldcop;
3791 PL_tmps_floor = cx->blk_old_tmpsfloor;
3793 Note that it I<doesn't> restore C<PL_stack_sp>; as mentioned earlier,
3794 which value to restore it to depends on the context type (specifically
3795 C<for (list) {}>), and what args (if any) it returns; and that will
3796 already have been sorted out earlier by C<leave_adjust_stacks()>.
3798 Finally, the context stack pointer is actually decremented by C<CX_POP(cx)>.
3799 After this point, it's possible that that the current context frame could
3800 be overwritten by other contexts being pushed. Although things like ties
3801 and C<DESTROY> are supposed to work within a new context stack, it's best
3802 not to assume this. Indeed on debugging builds, C<CX_POP(cx)> deliberately
3803 sets C<cx> to null to detect code that is still relying on the field
3804 values in that context frame. Note in the C<pp_leavesub()> example above,
3805 we grab C<blk_sub.retop> I<before> calling C<CX_POP>.
3807 =head2 Redoing contexts
3809 Finally, there is C<cx_topblock(cx)>, which acts like a super-C<nextstate>
3810 as regards to resetting various vars to their base values. It is used in
3811 places like C<pp_next>, C<pp_redo> and C<pp_goto> where rather than
3812 exiting a scope, we want to re-initialise the scope. As well as resetting
3813 C<PL_stack_sp> like C<nextstate>, it also resets C<PL_markstack_ptr>,
3814 C<PL_scopestack_ix> and C<PL_curpm>. Note that it doesn't do a
3818 =head1 Slab-based operator allocation
3820 B<Note:> this section describes a non-public internal API that is subject
3821 to change without notice.
3823 Perl's internal error-handling mechanisms implement C<die> (and its internal
3824 equivalents) using longjmp. If this occurs during lexing, parsing or
3825 compilation, we must ensure that any ops allocated as part of the compilation
3826 process are freed. (Older Perl versions did not adequately handle this
3827 situation: when failing a parse, they would leak ops that were stored in
3828 C C<auto> variables and not linked anywhere else.)
3830 To handle this situation, Perl uses I<op slabs> that are attached to the
3831 currently-compiling CV. A slab is a chunk of allocated memory. New ops are
3832 allocated as regions of the slab. If the slab fills up, a new one is created
3833 (and linked from the previous one). When an error occurs and the CV is freed,
3834 any ops remaining are freed.
3836 Each op is preceded by two pointers: one points to the next op in the slab, and
3837 the other points to the slab that owns it. The next-op pointer is needed so
3838 that Perl can iterate over a slab and free all its ops. (Op structures are of
3839 different sizes, so the slab's ops can't merely be treated as a dense array.)
3840 The slab pointer is needed for accessing a reference count on the slab: when
3841 the last op on a slab is freed, the slab itself is freed.
3843 The slab allocator puts the ops at the end of the slab first. This will tend to
3844 allocate the leaves of the op tree first, and the layout will therefore
3845 hopefully be cache-friendly. In addition, this means that there's no need to
3846 store the size of the slab (see below on why slabs vary in size), because Perl
3847 can follow pointers to find the last op.
3849 It might seem possible eliminate slab reference counts altogether, by having
3850 all ops implicitly attached to C<PL_compcv> when allocated and freed when the
3851 CV is freed. That would also allow C<op_free> to skip C<FreeOp> altogether, and
3852 thus free ops faster. But that doesn't work in those cases where ops need to
3853 survive beyond their CVs, such as re-evals.
3855 The CV also has to have a reference count on the slab. Sometimes the first op
3856 created is immediately freed. If the reference count of the slab reaches 0,
3857 then it will be freed with the CV still pointing to it.
3859 CVs use the C<CVf_SLABBED> flag to indicate that the CV has a reference count
3860 on the slab. When this flag is set, the slab is accessible via C<CvSTART> when
3861 C<CvROOT> is not set, or by subtracting two pointers C<(2*sizeof(I32 *))> from
3862 C<CvROOT> when it is set. The alternative to this approach of sneaking the slab
3863 into C<CvSTART> during compilation would be to enlarge the C<xpvcv> struct by
3864 another pointer. But that would make all CVs larger, even though slab-based op
3865 freeing is typically of benefit only for programs that make significant use of
3868 When the C<CVf_SLABBED> flag is set, the CV takes responsibility for freeing
3869 the slab. If C<CvROOT> is not set when the CV is freed or undeffed, it is
3870 assumed that a compilation error has occurred, so the op slab is traversed and
3871 all the ops are freed.
3873 Under normal circumstances, the CV forgets about its slab (decrementing the
3874 reference count) when the root is attached. So the slab reference counting that
3875 happens when ops are freed takes care of freeing the slab. In some cases, the
3876 CV is told to forget about the slab (C<cv_forget_slab>) precisely so that the
3877 ops can survive after the CV is done away with.
3879 Forgetting the slab when the root is attached is not strictly necessary, but
3880 avoids potential problems with C<CvROOT> being written over. There is code all
3881 over the place, both in core and on CPAN, that does things with C<CvROOT>, so
3882 forgetting the slab makes things more robust and avoids potential problems.
3884 Since the CV takes ownership of its slab when flagged, that flag is never
3885 copied when a CV is cloned, as one CV could free a slab that another CV still
3886 points to, since forced freeing of ops ignores the reference count (but asserts
3887 that it looks right).
3889 To avoid slab fragmentation, freed ops are marked as freed and attached to the
3890 slab's freed chain (an idea stolen from DBM::Deep). Those freed ops are reused
3891 when possible. Not reusing freed ops would be simpler, but it would result in
3892 significantly higher memory usage for programs with large C<if (DEBUG) {...}>
3895 C<SAVEFREEOP> is slightly problematic under this scheme. Sometimes it can cause
3896 an op to be freed after its CV. If the CV has forcibly freed the ops on its
3897 slab and the slab itself, then we will be fiddling with a freed slab. Making
3898 C<SAVEFREEOP> a no-op doesn't help, as sometimes an op can be savefreed when
3899 there is no compilation error, so the op would never be freed. It holds
3900 a reference count on the slab, so the whole slab would leak. So C<SAVEFREEOP>
3901 now sets a special flag on the op (C<< ->op_savefree >>). The forced freeing of
3902 ops after a compilation error won't free any ops thus marked.
3904 Since many pieces of code create tiny subroutines consisting of only a few ops,
3905 and since a huge slab would be quite a bit of baggage for those to carry
3906 around, the first slab is always very small. To avoid allocating too many
3907 slabs for a single CV, each subsequent slab is twice the size of the previous.
3909 Smartmatch expects to be able to allocate an op at run time, run it, and then
3910 throw it away. For that to work the op is simply malloced when PL_compcv hasn't
3911 been set up. So all slab-allocated ops are marked as such (C<< ->op_slabbed >>),
3912 to distinguish them from malloced ops.
3917 Until May 1997, this document was maintained by Jeff Okamoto
3918 E<lt>okamoto@corp.hp.comE<gt>. It is now maintained as part of Perl
3919 itself by the Perl 5 Porters E<lt>perl5-porters@perl.orgE<gt>.
3921 With lots of help and suggestions from Dean Roehrich, Malcolm Beattie,
3922 Andreas Koenig, Paul Hudson, Ilya Zakharevich, Paul Marquess, Neil
3923 Bowers, Matthew Green, Tim Bunce, Spider Boardman, Ulrich Pfeifer,
3924 Stephen McCamant, and Gurusamy Sarathy.
3928 L<perlapi>, L<perlintern>, L<perlxs>, L<perlembed>