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).
41 The seven routines are:
46 SV* newSVpv(const char*, STRLEN);
47 SV* newSVpvn(const char*, STRLEN);
48 SV* newSVpvf(const char*, ...);
51 C<STRLEN> is an integer type (Size_t, usually defined as size_t in
52 F<config.h>) guaranteed to be large enough to represent the size of
53 any string that perl can handle.
55 In the unlikely case of a SV requiring more complex initialisation, you
56 can create an empty SV with newSV(len). If C<len> is 0 an empty SV of
57 type NULL is returned, else an SV of type PV is returned with len + 1 (for
58 the NUL) bytes of storage allocated, accessible via SvPVX. In both cases
59 the SV has the undef value.
61 SV *sv = newSV(0); /* no storage allocated */
62 SV *sv = newSV(10); /* 10 (+1) bytes of uninitialised storage
65 To change the value of an I<already-existing> SV, there are eight routines:
67 void sv_setiv(SV*, IV);
68 void sv_setuv(SV*, UV);
69 void sv_setnv(SV*, double);
70 void sv_setpv(SV*, const char*);
71 void sv_setpvn(SV*, const char*, STRLEN)
72 void sv_setpvf(SV*, const char*, ...);
73 void sv_vsetpvfn(SV*, const char*, STRLEN, va_list *,
75 void sv_setsv(SV*, SV*);
77 Notice that you can choose to specify the length of the string to be
78 assigned by using C<sv_setpvn>, C<newSVpvn>, or C<newSVpv>, or you may
79 allow Perl to calculate the length by using C<sv_setpv> or by specifying
80 0 as the second argument to C<newSVpv>. Be warned, though, that Perl will
81 determine the string's length by using C<strlen>, which depends on the
82 string terminating with a NUL character, and not otherwise containing
85 The arguments of C<sv_setpvf> are processed like C<sprintf>, and the
86 formatted output becomes the value.
88 C<sv_vsetpvfn> is an analogue of C<vsprintf>, but it allows you to specify
89 either a pointer to a variable argument list or the address and length of
90 an array of SVs. The last argument points to a boolean; on return, if that
91 boolean is true, then locale-specific information has been used to format
92 the string, and the string's contents are therefore untrustworthy (see
93 L<perlsec>). This pointer may be NULL if that information is not
94 important. Note that this function requires you to specify the length of
97 The C<sv_set*()> functions are not generic enough to operate on values
98 that have "magic". See L<Magic Virtual Tables> later in this document.
100 All SVs that contain strings should be terminated with a NUL character.
101 If it is not NUL-terminated there is a risk of
102 core dumps and corruptions from code which passes the string to C
103 functions or system calls which expect a NUL-terminated string.
104 Perl's own functions typically add a trailing NUL for this reason.
105 Nevertheless, you should be very careful when you pass a string stored
106 in an SV to a C function or system call.
108 To access the actual value that an SV points to, you can use the macros:
113 SvPV(SV*, STRLEN len)
116 which will automatically coerce the actual scalar type into an IV, UV, double,
119 In the C<SvPV> macro, the length of the string returned is placed into the
120 variable C<len> (this is a macro, so you do I<not> use C<&len>). If you do
121 not care what the length of the data is, use the C<SvPV_nolen> macro.
122 Historically the C<SvPV> macro with the global variable C<PL_na> has been
123 used in this case. But that can be quite inefficient because C<PL_na> must
124 be accessed in thread-local storage in threaded Perl. In any case, remember
125 that Perl allows arbitrary strings of data that may both contain NULs and
126 might not be terminated by a NUL.
128 Also remember that C doesn't allow you to safely say C<foo(SvPV(s, len),
129 len);>. It might work with your compiler, but it won't work for everyone.
130 Break this sort of statement up into separate assignments:
138 If you want to know if the scalar value is TRUE, you can use:
142 Although Perl will automatically grow strings for you, if you need to force
143 Perl to allocate more memory for your SV, you can use the macro
145 SvGROW(SV*, STRLEN newlen)
147 which will determine if more memory needs to be allocated. If so, it will
148 call the function C<sv_grow>. Note that C<SvGROW> can only increase, not
149 decrease, the allocated memory of an SV and that it does not automatically
150 add space for the trailing NUL byte (perl's own string functions typically do
151 C<SvGROW(sv, len + 1)>).
153 If you have an SV and want to know what kind of data Perl thinks is stored
154 in it, you can use the following macros to check the type of SV you have.
160 You can get and set the current length of the string stored in an SV with
161 the following macros:
164 SvCUR_set(SV*, I32 val)
166 You can also get a pointer to the end of the string stored in the SV
171 But note that these last three macros are valid only if C<SvPOK()> is true.
173 If you want to append something to the end of string stored in an C<SV*>,
174 you can use the following functions:
176 void sv_catpv(SV*, const char*);
177 void sv_catpvn(SV*, const char*, STRLEN);
178 void sv_catpvf(SV*, const char*, ...);
179 void sv_vcatpvfn(SV*, const char*, STRLEN, va_list *, SV **,
181 void sv_catsv(SV*, SV*);
183 The first function calculates the length of the string to be appended by
184 using C<strlen>. In the second, you specify the length of the string
185 yourself. The third function processes its arguments like C<sprintf> and
186 appends the formatted output. The fourth function works like C<vsprintf>.
187 You can specify the address and length of an array of SVs instead of the
188 va_list argument. The fifth function extends the string stored in the first
189 SV with the string stored in the second SV. It also forces the second SV
190 to be interpreted as a string.
192 The C<sv_cat*()> functions are not generic enough to operate on values that
193 have "magic". See L<Magic Virtual Tables> later in this document.
195 If you know the name of a scalar variable, you can get a pointer to its SV
196 by using the following:
198 SV* get_sv("package::varname", 0);
200 This returns NULL if the variable does not exist.
202 If you want to know if this variable (or any other SV) is actually C<defined>,
207 The scalar C<undef> value is stored in an SV instance called C<PL_sv_undef>.
209 Its address can be used whenever an C<SV*> is needed. Make sure that
210 you don't try to compare a random sv with C<&PL_sv_undef>. For example
211 when interfacing Perl code, it'll work correctly for:
215 But won't work when called as:
220 So to repeat always use SvOK() to check whether an sv is defined.
222 Also you have to be careful when using C<&PL_sv_undef> as a value in
223 AVs or HVs (see L<AVs, HVs and undefined values>).
225 There are also the two values C<PL_sv_yes> and C<PL_sv_no>, which contain
226 boolean TRUE and FALSE values, respectively. Like C<PL_sv_undef>, their
227 addresses can be used whenever an C<SV*> is needed.
229 Do not be fooled into thinking that C<(SV *) 0> is the same as C<&PL_sv_undef>.
233 if (I-am-to-return-a-real-value) {
234 sv = sv_2mortal(newSViv(42));
238 This code tries to return a new SV (which contains the value 42) if it should
239 return a real value, or undef otherwise. Instead it has returned a NULL
240 pointer which, somewhere down the line, will cause a segmentation violation,
241 bus error, or just weird results. Change the zero to C<&PL_sv_undef> in the
242 first line and all will be well.
244 To free an SV that you've created, call C<SvREFCNT_dec(SV*)>. Normally this
245 call is not necessary (see L<Reference Counts and Mortality>).
249 Perl provides the function C<sv_chop> to efficiently remove characters
250 from the beginning of a string; you give it an SV and a pointer to
251 somewhere inside the PV, and it discards everything before the
252 pointer. The efficiency comes by means of a little hack: instead of
253 actually removing the characters, C<sv_chop> sets the flag C<OOK>
254 (offset OK) to signal to other functions that the offset hack is in
255 effect, and it puts the number of bytes chopped off into the IV field
256 of the SV. It then moves the PV pointer (called C<SvPVX>) forward that
257 many bytes, and adjusts C<SvCUR> and C<SvLEN>.
259 Hence, at this point, the start of the buffer that we allocated lives
260 at C<SvPVX(sv) - SvIV(sv)> in memory and the PV pointer is pointing
261 into the middle of this allocated storage.
263 This is best demonstrated by example:
265 % ./perl -Ilib -MDevel::Peek -le '$a="12345"; $a=~s/.//; Dump($a)'
266 SV = PVIV(0x8128450) at 0x81340f0
268 FLAGS = (POK,OOK,pPOK)
270 PV = 0x8135781 ( "1" . ) "2345"\0
274 Here the number of bytes chopped off (1) is put into IV, and
275 C<Devel::Peek::Dump> helpfully reminds us that this is an offset. The
276 portion of the string between the "real" and the "fake" beginnings is
277 shown in parentheses, and the values of C<SvCUR> and C<SvLEN> reflect
278 the fake beginning, not the real one.
280 Something similar to the offset hack is performed on AVs to enable
281 efficient shifting and splicing off the beginning of the array; while
282 C<AvARRAY> points to the first element in the array that is visible from
283 Perl, C<AvALLOC> points to the real start of the C array. These are
284 usually the same, but a C<shift> operation can be carried out by
285 increasing C<AvARRAY> by one and decreasing C<AvFILL> and C<AvMAX>.
286 Again, the location of the real start of the C array only comes into
287 play when freeing the array. See C<av_shift> in F<av.c>.
289 =head2 What's Really Stored in an SV?
291 Recall that the usual method of determining the type of scalar you have is
292 to use C<Sv*OK> macros. Because a scalar can be both a number and a string,
293 usually these macros will always return TRUE and calling the C<Sv*V>
294 macros will do the appropriate conversion of string to integer/double or
295 integer/double to string.
297 If you I<really> need to know if you have an integer, double, or string
298 pointer in an SV, you can use the following three macros instead:
304 These will tell you if you truly have an integer, double, or string pointer
305 stored in your SV. The "p" stands for private.
307 There are various ways in which the private and public flags may differ.
308 For example, a tied SV may have a valid underlying value in the IV slot
309 (so SvIOKp is true), but the data should be accessed via the FETCH
310 routine rather than directly, so SvIOK is false. Another is when
311 numeric conversion has occurred and precision has been lost: only the
312 private flag is set on 'lossy' values. So when an NV is converted to an
313 IV with loss, SvIOKp, SvNOKp and SvNOK will be set, while SvIOK wont be.
315 In general, though, it's best to use the C<Sv*V> macros.
317 =head2 Working with AVs
319 There are two ways to create and load an AV. The first method creates an
324 The second method both creates the AV and initially populates it with SVs:
326 AV* av_make(I32 num, SV **ptr);
328 The second argument points to an array containing C<num> C<SV*>'s. Once the
329 AV has been created, the SVs can be destroyed, if so desired.
331 Once the AV has been created, the following operations are possible on it:
333 void av_push(AV*, SV*);
336 void av_unshift(AV*, I32 num);
338 These should be familiar operations, with the exception of C<av_unshift>.
339 This routine adds C<num> elements at the front of the array with the C<undef>
340 value. You must then use C<av_store> (described below) to assign values
341 to these new elements.
343 Here are some other functions:
346 SV** av_fetch(AV*, I32 key, I32 lval);
347 SV** av_store(AV*, I32 key, SV* val);
349 The C<av_len> function returns the highest index value in an array (just
350 like $#array in Perl). If the array is empty, -1 is returned. The
351 C<av_fetch> function returns the value at index C<key>, but if C<lval>
352 is non-zero, then C<av_fetch> will store an undef value at that index.
353 The C<av_store> function stores the value C<val> at index C<key>, and does
354 not increment the reference count of C<val>. Thus the caller is responsible
355 for taking care of that, and if C<av_store> returns NULL, the caller will
356 have to decrement the reference count to avoid a memory leak. Note that
357 C<av_fetch> and C<av_store> both return C<SV**>'s, not C<SV*>'s as their
364 void av_extend(AV*, I32 key);
366 The C<av_clear> function deletes all the elements in the AV* array, but
367 does not actually delete the array itself. The C<av_undef> function will
368 delete all the elements in the array plus the array itself. The
369 C<av_extend> function extends the array so that it contains at least C<key+1>
370 elements. If C<key+1> is less than the currently allocated length of the array,
371 then nothing is done.
373 If you know the name of an array variable, you can get a pointer to its AV
374 by using the following:
376 AV* get_av("package::varname", 0);
378 This returns NULL if the variable does not exist.
380 See L<Understanding the Magic of Tied Hashes and Arrays> for more
381 information on how to use the array access functions on tied arrays.
383 =head2 Working with HVs
385 To create an HV, you use the following routine:
389 Once the HV has been created, the following operations are possible on it:
391 SV** hv_store(HV*, const char* key, U32 klen, SV* val, U32 hash);
392 SV** hv_fetch(HV*, const char* key, U32 klen, I32 lval);
394 The C<klen> parameter is the length of the key being passed in (Note that
395 you cannot pass 0 in as a value of C<klen> to tell Perl to measure the
396 length of the key). The C<val> argument contains the SV pointer to the
397 scalar being stored, and C<hash> is the precomputed hash value (zero if
398 you want C<hv_store> to calculate it for you). The C<lval> parameter
399 indicates whether this fetch is actually a part of a store operation, in
400 which case a new undefined value will be added to the HV with the supplied
401 key and C<hv_fetch> will return as if the value had already existed.
403 Remember that C<hv_store> and C<hv_fetch> return C<SV**>'s and not just
404 C<SV*>. To access the scalar value, you must first dereference the return
405 value. However, you should check to make sure that the return value is
406 not NULL before dereferencing it.
408 The first of these two functions checks if a hash table entry exists, and the
411 bool hv_exists(HV*, const char* key, U32 klen);
412 SV* hv_delete(HV*, const char* key, U32 klen, I32 flags);
414 If C<flags> does not include the C<G_DISCARD> flag then C<hv_delete> will
415 create and return a mortal copy of the deleted value.
417 And more miscellaneous functions:
422 Like their AV counterparts, C<hv_clear> deletes all the entries in the hash
423 table but does not actually delete the hash table. The C<hv_undef> deletes
424 both the entries and the hash table itself.
426 Perl keeps the actual data in a linked list of structures with a typedef of HE.
427 These contain the actual key and value pointers (plus extra administrative
428 overhead). The key is a string pointer; the value is an C<SV*>. However,
429 once you have an C<HE*>, to get the actual key and value, use the routines
432 I32 hv_iterinit(HV*);
433 /* Prepares starting point to traverse hash table */
434 HE* hv_iternext(HV*);
435 /* Get the next entry, and return a pointer to a
436 structure that has both the key and value */
437 char* hv_iterkey(HE* entry, I32* retlen);
438 /* Get the key from an HE structure and also return
439 the length of the key string */
440 SV* hv_iterval(HV*, HE* entry);
441 /* Return an SV pointer to the value of the HE
443 SV* hv_iternextsv(HV*, char** key, I32* retlen);
444 /* This convenience routine combines hv_iternext,
445 hv_iterkey, and hv_iterval. The key and retlen
446 arguments are return values for the key and its
447 length. The value is returned in the SV* argument */
449 If you know the name of a hash variable, you can get a pointer to its HV
450 by using the following:
452 HV* get_hv("package::varname", 0);
454 This returns NULL if the variable does not exist.
456 The hash algorithm is defined in the C<PERL_HASH(hash, key, klen)> macro:
460 hash = (hash * 33) + *key++;
461 hash = hash + (hash >> 5); /* after 5.6 */
463 The last step was added in version 5.6 to improve distribution of
464 lower bits in the resulting hash value.
466 See L<Understanding the Magic of Tied Hashes and Arrays> for more
467 information on how to use the hash access functions on tied hashes.
469 =head2 Hash API Extensions
471 Beginning with version 5.004, the following functions are also supported:
473 HE* hv_fetch_ent (HV* tb, SV* key, I32 lval, U32 hash);
474 HE* hv_store_ent (HV* tb, SV* key, SV* val, U32 hash);
476 bool hv_exists_ent (HV* tb, SV* key, U32 hash);
477 SV* hv_delete_ent (HV* tb, SV* key, I32 flags, U32 hash);
479 SV* hv_iterkeysv (HE* entry);
481 Note that these functions take C<SV*> keys, which simplifies writing
482 of extension code that deals with hash structures. These functions
483 also allow passing of C<SV*> keys to C<tie> functions without forcing
484 you to stringify the keys (unlike the previous set of functions).
486 They also return and accept whole hash entries (C<HE*>), making their
487 use more efficient (since the hash number for a particular string
488 doesn't have to be recomputed every time). See L<perlapi> for detailed
491 The following macros must always be used to access the contents of hash
492 entries. Note that the arguments to these macros must be simple
493 variables, since they may get evaluated more than once. See
494 L<perlapi> for detailed descriptions of these macros.
496 HePV(HE* he, STRLEN len)
500 HeSVKEY_force(HE* he)
501 HeSVKEY_set(HE* he, SV* sv)
503 These two lower level macros are defined, but must only be used when
504 dealing with keys that are not C<SV*>s:
509 Note that both C<hv_store> and C<hv_store_ent> do not increment the
510 reference count of the stored C<val>, which is the caller's responsibility.
511 If these functions return a NULL value, the caller will usually have to
512 decrement the reference count of C<val> to avoid a memory leak.
514 =head2 AVs, HVs and undefined values
516 Sometimes you have to store undefined values in AVs or HVs. Although
517 this may be a rare case, it can be tricky. That's because you're
518 used to using C<&PL_sv_undef> if you need an undefined SV.
520 For example, intuition tells you that this XS code:
523 av_store( av, 0, &PL_sv_undef );
525 is equivalent to this Perl code:
530 Unfortunately, this isn't true. AVs use C<&PL_sv_undef> as a marker
531 for indicating that an array element has not yet been initialized.
532 Thus, C<exists $av[0]> would be true for the above Perl code, but
533 false for the array generated by the XS code.
535 Other problems can occur when storing C<&PL_sv_undef> in HVs:
537 hv_store( hv, "key", 3, &PL_sv_undef, 0 );
539 This will indeed make the value C<undef>, but if you try to modify
540 the value of C<key>, you'll get the following error:
542 Modification of non-creatable hash value attempted
544 In perl 5.8.0, C<&PL_sv_undef> was also used to mark placeholders
545 in restricted hashes. This caused such hash entries not to appear
546 when iterating over the hash or when checking for the keys
547 with the C<hv_exists> function.
549 You can run into similar problems when you store C<&PL_sv_yes> or
550 C<&PL_sv_no> into AVs or HVs. Trying to modify such elements
551 will give you the following error:
553 Modification of a read-only value attempted
555 To make a long story short, you can use the special variables
556 C<&PL_sv_undef>, C<&PL_sv_yes> and C<&PL_sv_no> with AVs and
557 HVs, but you have to make sure you know what you're doing.
559 Generally, if you want to store an undefined value in an AV
560 or HV, you should not use C<&PL_sv_undef>, but rather create a
561 new undefined value using the C<newSV> function, for example:
563 av_store( av, 42, newSV(0) );
564 hv_store( hv, "foo", 3, newSV(0), 0 );
568 References are a special type of scalar that point to other data types
569 (including other references).
571 To create a reference, use either of the following functions:
573 SV* newRV_inc((SV*) thing);
574 SV* newRV_noinc((SV*) thing);
576 The C<thing> argument can be any of an C<SV*>, C<AV*>, or C<HV*>. The
577 functions are identical except that C<newRV_inc> increments the reference
578 count of the C<thing>, while C<newRV_noinc> does not. For historical
579 reasons, C<newRV> is a synonym for C<newRV_inc>.
581 Once you have a reference, you can use the following macro to dereference
586 then call the appropriate routines, casting the returned C<SV*> to either an
587 C<AV*> or C<HV*>, if required.
589 To determine if an SV is a reference, you can use the following macro:
593 To discover what type of value the reference refers to, use the following
594 macro and then check the return value.
598 The most useful types that will be returned are:
607 SVt_PVGV Glob (possibly a file handle)
608 SVt_PVMG Blessed or Magical Scalar
610 See the F<sv.h> header file for more details.
612 =head2 Blessed References and Class Objects
614 References are also used to support object-oriented programming. In perl's
615 OO lexicon, an object is simply a reference that has been blessed into a
616 package (or class). Once blessed, the programmer may now use the reference
617 to access the various methods in the class.
619 A reference can be blessed into a package with the following function:
621 SV* sv_bless(SV* sv, HV* stash);
623 The C<sv> argument must be a reference value. The C<stash> argument
624 specifies which class the reference will belong to. See
625 L<Stashes and Globs> for information on converting class names into stashes.
627 /* Still under construction */
629 The following function upgrades rv to reference if not already one.
630 Creates a new SV for rv to point to. If C<classname> is non-null, the SV
631 is blessed into the specified class. SV is returned.
633 SV* newSVrv(SV* rv, const char* classname);
635 The following three functions copy integer, unsigned integer or double
636 into an SV whose reference is C<rv>. SV is blessed if C<classname> is
639 SV* sv_setref_iv(SV* rv, const char* classname, IV iv);
640 SV* sv_setref_uv(SV* rv, const char* classname, UV uv);
641 SV* sv_setref_nv(SV* rv, const char* classname, NV iv);
643 The following function copies the pointer value (I<the address, not the
644 string!>) into an SV whose reference is rv. SV is blessed if C<classname>
647 SV* sv_setref_pv(SV* rv, const char* classname, void* pv);
649 The following function copies a string into an SV whose reference is C<rv>.
650 Set length to 0 to let Perl calculate the string length. SV is blessed if
651 C<classname> is non-null.
653 SV* sv_setref_pvn(SV* rv, const char* classname, char* pv,
656 The following function tests whether the SV is blessed into the specified
657 class. It does not check inheritance relationships.
659 int sv_isa(SV* sv, const char* name);
661 The following function tests whether the SV is a reference to a blessed object.
663 int sv_isobject(SV* sv);
665 The following function tests whether the SV is derived from the specified
666 class. SV can be either a reference to a blessed object or a string
667 containing a class name. This is the function implementing the
668 C<UNIVERSAL::isa> functionality.
670 bool sv_derived_from(SV* sv, const char* name);
672 To check if you've got an object derived from a specific class you have
675 if (sv_isobject(sv) && sv_derived_from(sv, class)) { ... }
677 =head2 Creating New Variables
679 To create a new Perl variable with an undef value which can be accessed from
680 your Perl script, use the following routines, depending on the variable type.
682 SV* get_sv("package::varname", GV_ADD);
683 AV* get_av("package::varname", GV_ADD);
684 HV* get_hv("package::varname", GV_ADD);
686 Notice the use of GV_ADD as the second parameter. The new variable can now
687 be set, using the routines appropriate to the data type.
689 There are additional macros whose values may be bitwise OR'ed with the
690 C<GV_ADD> argument to enable certain extra features. Those bits are:
696 Marks the variable as multiply defined, thus preventing the:
698 Name <varname> used only once: possible typo
706 Had to create <varname> unexpectedly
708 if the variable did not exist before the function was called.
712 If you do not specify a package name, the variable is created in the current
715 =head2 Reference Counts and Mortality
717 Perl uses a reference count-driven garbage collection mechanism. SVs,
718 AVs, or HVs (xV for short in the following) start their life with a
719 reference count of 1. If the reference count of an xV ever drops to 0,
720 then it will be destroyed and its memory made available for reuse.
722 This normally doesn't happen at the Perl level unless a variable is
723 undef'ed or the last variable holding a reference to it is changed or
724 overwritten. At the internal level, however, reference counts can be
725 manipulated with the following macros:
727 int SvREFCNT(SV* sv);
728 SV* SvREFCNT_inc(SV* sv);
729 void SvREFCNT_dec(SV* sv);
731 However, there is one other function which manipulates the reference
732 count of its argument. The C<newRV_inc> function, you will recall,
733 creates a reference to the specified argument. As a side effect,
734 it increments the argument's reference count. If this is not what
735 you want, use C<newRV_noinc> instead.
737 For example, imagine you want to return a reference from an XSUB function.
738 Inside the XSUB routine, you create an SV which initially has a reference
739 count of one. Then you call C<newRV_inc>, passing it the just-created SV.
740 This returns the reference as a new SV, but the reference count of the
741 SV you passed to C<newRV_inc> has been incremented to two. Now you
742 return the reference from the XSUB routine and forget about the SV.
743 But Perl hasn't! Whenever the returned reference is destroyed, the
744 reference count of the original SV is decreased to one and nothing happens.
745 The SV will hang around without any way to access it until Perl itself
746 terminates. This is a memory leak.
748 The correct procedure, then, is to use C<newRV_noinc> instead of
749 C<newRV_inc>. Then, if and when the last reference is destroyed,
750 the reference count of the SV will go to zero and it will be destroyed,
751 stopping any memory leak.
753 There are some convenience functions available that can help with the
754 destruction of xVs. These functions introduce the concept of "mortality".
755 An xV that is mortal has had its reference count marked to be decremented,
756 but not actually decremented, until "a short time later". Generally the
757 term "short time later" means a single Perl statement, such as a call to
758 an XSUB function. The actual determinant for when mortal xVs have their
759 reference count decremented depends on two macros, SAVETMPS and FREETMPS.
760 See L<perlcall> and L<perlxs> for more details on these macros.
762 "Mortalization" then is at its simplest a deferred C<SvREFCNT_dec>.
763 However, if you mortalize a variable twice, the reference count will
764 later be decremented twice.
766 "Mortal" SVs are mainly used for SVs that are placed on perl's stack.
767 For example an SV which is created just to pass a number to a called sub
768 is made mortal to have it cleaned up automatically when it's popped off
769 the stack. Similarly, results returned by XSUBs (which are pushed on the
770 stack) are often made mortal.
772 To create a mortal variable, use the functions:
776 SV* sv_mortalcopy(SV*)
778 The first call creates a mortal SV (with no value), the second converts an existing
779 SV to a mortal SV (and thus defers a call to C<SvREFCNT_dec>), and the
780 third creates a mortal copy of an existing SV.
781 Because C<sv_newmortal> gives the new SV no value, it must normally be given one
782 via C<sv_setpv>, C<sv_setiv>, etc. :
784 SV *tmp = sv_newmortal();
785 sv_setiv(tmp, an_integer);
787 As that is multiple C statements it is quite common so see this idiom instead:
789 SV *tmp = sv_2mortal(newSViv(an_integer));
792 You should be careful about creating mortal variables. Strange things
793 can happen if you make the same value mortal within multiple contexts,
794 or if you make a variable mortal multiple times. Thinking of "Mortalization"
795 as deferred C<SvREFCNT_dec> should help to minimize such problems.
796 For example if you are passing an SV which you I<know> has a high enough REFCNT
797 to survive its use on the stack you need not do any mortalization.
798 If you are not sure then doing an C<SvREFCNT_inc> and C<sv_2mortal>, or
799 making a C<sv_mortalcopy> is safer.
801 The mortal routines are not just for SVs; AVs and HVs can be
802 made mortal by passing their address (type-casted to C<SV*>) to the
803 C<sv_2mortal> or C<sv_mortalcopy> routines.
805 =head2 Stashes and Globs
807 A B<stash> is a hash that contains all variables that are defined
808 within a package. Each key of the stash is a symbol
809 name (shared by all the different types of objects that have the same
810 name), and each value in the hash table is a GV (Glob Value). This GV
811 in turn contains references to the various objects of that name,
812 including (but not limited to) the following:
821 There is a single stash called C<PL_defstash> that holds the items that exist
822 in the C<main> package. To get at the items in other packages, append the
823 string "::" to the package name. The items in the C<Foo> package are in
824 the stash C<Foo::> in PL_defstash. The items in the C<Bar::Baz> package are
825 in the stash C<Baz::> in C<Bar::>'s stash.
827 To get the stash pointer for a particular package, use the function:
829 HV* gv_stashpv(const char* name, I32 flags)
830 HV* gv_stashsv(SV*, I32 flags)
832 The first function takes a literal string, the second uses the string stored
833 in the SV. Remember that a stash is just a hash table, so you get back an
834 C<HV*>. The C<flags> flag will create a new package if it is set to GV_ADD.
836 The name that C<gv_stash*v> wants is the name of the package whose symbol table
837 you want. The default package is called C<main>. If you have multiply nested
838 packages, pass their names to C<gv_stash*v>, separated by C<::> as in the Perl
841 Alternately, if you have an SV that is a blessed reference, you can find
842 out the stash pointer by using:
844 HV* SvSTASH(SvRV(SV*));
846 then use the following to get the package name itself:
848 char* HvNAME(HV* stash);
850 If you need to bless or re-bless an object you can use the following
853 SV* sv_bless(SV*, HV* stash)
855 where the first argument, an C<SV*>, must be a reference, and the second
856 argument is a stash. The returned C<SV*> can now be used in the same way
859 For more information on references and blessings, consult L<perlref>.
861 =head2 Double-Typed SVs
863 Scalar variables normally contain only one type of value, an integer,
864 double, pointer, or reference. Perl will automatically convert the
865 actual scalar data from the stored type into the requested type.
867 Some scalar variables contain more than one type of scalar data. For
868 example, the variable C<$!> contains either the numeric value of C<errno>
869 or its string equivalent from either C<strerror> or C<sys_errlist[]>.
871 To force multiple data values into an SV, you must do two things: use the
872 C<sv_set*v> routines to add the additional scalar type, then set a flag
873 so that Perl will believe it contains more than one type of data. The
874 four macros to set the flags are:
881 The particular macro you must use depends on which C<sv_set*v> routine
882 you called first. This is because every C<sv_set*v> routine turns on
883 only the bit for the particular type of data being set, and turns off
886 For example, to create a new Perl variable called "dberror" that contains
887 both the numeric and descriptive string error values, you could use the
891 extern char *dberror_list;
893 SV* sv = get_sv("dberror", GV_ADD);
894 sv_setiv(sv, (IV) dberror);
895 sv_setpv(sv, dberror_list[dberror]);
898 If the order of C<sv_setiv> and C<sv_setpv> had been reversed, then the
899 macro C<SvPOK_on> would need to be called instead of C<SvIOK_on>.
901 =head2 Magic Variables
903 [This section still under construction. Ignore everything here. Post no
904 bills. Everything not permitted is forbidden.]
906 Any SV may be magical, that is, it has special features that a normal
907 SV does not have. These features are stored in the SV structure in a
908 linked list of C<struct magic>'s, typedef'ed to C<MAGIC>.
921 Note this is current as of patchlevel 0, and could change at any time.
923 =head2 Assigning Magic
925 Perl adds magic to an SV using the sv_magic function:
927 void sv_magic(SV* sv, SV* obj, int how, const char* name, I32 namlen);
929 The C<sv> argument is a pointer to the SV that is to acquire a new magical
932 If C<sv> is not already magical, Perl uses the C<SvUPGRADE> macro to
933 convert C<sv> to type C<SVt_PVMG>. Perl then continues by adding new magic
934 to the beginning of the linked list of magical features. Any prior entry
935 of the same type of magic is deleted. Note that this can be overridden,
936 and multiple instances of the same type of magic can be associated with an
939 The C<name> and C<namlen> arguments are used to associate a string with
940 the magic, typically the name of a variable. C<namlen> is stored in the
941 C<mg_len> field and if C<name> is non-null then either a C<savepvn> copy of
942 C<name> or C<name> itself is stored in the C<mg_ptr> field, depending on
943 whether C<namlen> is greater than zero or equal to zero respectively. As a
944 special case, if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
945 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
947 The sv_magic function uses C<how> to determine which, if any, predefined
948 "Magic Virtual Table" should be assigned to the C<mg_virtual> field.
949 See the L<Magic Virtual Tables> section below. The C<how> argument is also
950 stored in the C<mg_type> field. The value of C<how> should be chosen
951 from the set of macros C<PERL_MAGIC_foo> found in F<perl.h>. Note that before
952 these macros were added, Perl internals used to directly use character
953 literals, so you may occasionally come across old code or documentation
954 referring to 'U' magic rather than C<PERL_MAGIC_uvar> for example.
956 The C<obj> argument is stored in the C<mg_obj> field of the C<MAGIC>
957 structure. If it is not the same as the C<sv> argument, the reference
958 count of the C<obj> object is incremented. If it is the same, or if
959 the C<how> argument is C<PERL_MAGIC_arylen>, or if it is a NULL pointer,
960 then C<obj> is merely stored, without the reference count being incremented.
962 See also C<sv_magicext> in L<perlapi> for a more flexible way to add magic
965 There is also a function to add magic to an C<HV>:
967 void hv_magic(HV *hv, GV *gv, int how);
969 This simply calls C<sv_magic> and coerces the C<gv> argument into an C<SV>.
971 To remove the magic from an SV, call the function sv_unmagic:
973 int sv_unmagic(SV *sv, int type);
975 The C<type> argument should be equal to the C<how> value when the C<SV>
976 was initially made magical.
978 However, note that C<sv_unmagic> removes all magic of a certain C<type> from the
979 C<SV>. If you want to remove only certain magic of a C<type> based on the magic
980 virtual table, use C<sv_unmagicext> instead:
982 int sv_unmagicext(SV *sv, int type, MGVTBL *vtbl);
984 =head2 Magic Virtual Tables
986 The C<mg_virtual> field in the C<MAGIC> structure is a pointer to an
987 C<MGVTBL>, which is a structure of function pointers and stands for
988 "Magic Virtual Table" to handle the various operations that might be
989 applied to that variable.
991 The C<MGVTBL> has five (or sometimes eight) pointers to the following
994 int (*svt_get)(SV* sv, MAGIC* mg);
995 int (*svt_set)(SV* sv, MAGIC* mg);
996 U32 (*svt_len)(SV* sv, MAGIC* mg);
997 int (*svt_clear)(SV* sv, MAGIC* mg);
998 int (*svt_free)(SV* sv, MAGIC* mg);
1000 int (*svt_copy)(SV *sv, MAGIC* mg, SV *nsv,
1001 const char *name, I32 namlen);
1002 int (*svt_dup)(MAGIC *mg, CLONE_PARAMS *param);
1003 int (*svt_local)(SV *nsv, MAGIC *mg);
1006 This MGVTBL structure is set at compile-time in F<perl.h> and there are
1007 currently 32 types. These different structures contain pointers to various
1008 routines that perform additional actions depending on which function is
1011 Function pointer Action taken
1012 ---------------- ------------
1013 svt_get Do something before the value of the SV is
1015 svt_set Do something after the SV is assigned a value.
1016 svt_len Report on the SV's length.
1017 svt_clear Clear something the SV represents.
1018 svt_free Free any extra storage associated with the SV.
1020 svt_copy copy tied variable magic to a tied element
1021 svt_dup duplicate a magic structure during thread cloning
1022 svt_local copy magic to local value during 'local'
1024 For instance, the MGVTBL structure called C<vtbl_sv> (which corresponds
1025 to an C<mg_type> of C<PERL_MAGIC_sv>) contains:
1027 { magic_get, magic_set, magic_len, 0, 0 }
1029 Thus, when an SV is determined to be magical and of type C<PERL_MAGIC_sv>,
1030 if a get operation is being performed, the routine C<magic_get> is
1031 called. All the various routines for the various magical types begin
1032 with C<magic_>. NOTE: the magic routines are not considered part of
1033 the Perl API, and may not be exported by the Perl library.
1035 The last three slots are a recent addition, and for source code
1036 compatibility they are only checked for if one of the three flags
1037 MGf_COPY, MGf_DUP or MGf_LOCAL is set in mg_flags. This means that most
1038 code can continue declaring a vtable as a 5-element value. These three are
1039 currently used exclusively by the threading code, and are highly subject
1042 The current kinds of Magic Virtual Tables are:
1045 This table is generated by regen/mg_vtable.pl. Any changes made here
1048 =for mg_vtable.pl begin
1051 (old-style char and macro) MGVTBL Type of magic
1052 -------------------------- ------ -------------
1053 \0 PERL_MAGIC_sv vtbl_sv Special scalar variable
1054 # PERL_MAGIC_arylen vtbl_arylen Array length ($#ary)
1055 % PERL_MAGIC_rhash (none) extra data for restricted
1057 . PERL_MAGIC_pos vtbl_pos pos() lvalue
1058 : PERL_MAGIC_symtab (none) extra data for symbol
1060 < PERL_MAGIC_backref vtbl_backref for weak ref data
1061 @ PERL_MAGIC_arylen_p (none) to move arylen out of
1063 A PERL_MAGIC_overload vtbl_amagic %OVERLOAD hash
1064 a PERL_MAGIC_overload_elem vtbl_amagicelem %OVERLOAD hash element
1065 B PERL_MAGIC_bm vtbl_regexp Boyer-Moore
1066 (fast string search)
1067 c PERL_MAGIC_overload_table vtbl_ovrld Holds overload table
1069 D PERL_MAGIC_regdata vtbl_regdata Regex match position data
1071 d PERL_MAGIC_regdatum vtbl_regdatum Regex match position data
1073 E PERL_MAGIC_env vtbl_env %ENV hash
1074 e PERL_MAGIC_envelem vtbl_envelem %ENV hash element
1075 f PERL_MAGIC_fm vtbl_regdata Formline
1077 G PERL_MAGIC_study vtbl_regexp study()ed string
1078 g PERL_MAGIC_regex_global vtbl_mglob m//g target
1079 H PERL_MAGIC_hints vtbl_hints %^H hash
1080 h PERL_MAGIC_hintselem vtbl_hintselem %^H hash element
1081 I PERL_MAGIC_isa vtbl_isa @ISA array
1082 i PERL_MAGIC_isaelem vtbl_isaelem @ISA array element
1083 k PERL_MAGIC_nkeys vtbl_nkeys scalar(keys()) lvalue
1084 L PERL_MAGIC_dbfile (none) Debugger %_<filename
1085 l PERL_MAGIC_dbline vtbl_dbline Debugger %_<filename
1087 N PERL_MAGIC_shared (none) Shared between threads
1088 n PERL_MAGIC_shared_scalar (none) Shared between threads
1089 o PERL_MAGIC_collxfrm vtbl_collxfrm Locale transformation
1090 P PERL_MAGIC_tied vtbl_pack Tied array or hash
1091 p PERL_MAGIC_tiedelem vtbl_packelem Tied array or hash element
1092 q PERL_MAGIC_tiedscalar vtbl_packelem Tied scalar or handle
1093 r PERL_MAGIC_qr vtbl_regexp precompiled qr// regex
1094 S PERL_MAGIC_sig (none) %SIG hash
1095 s PERL_MAGIC_sigelem vtbl_sigelem %SIG hash element
1096 t PERL_MAGIC_taint vtbl_taint Taintedness
1097 U PERL_MAGIC_uvar vtbl_uvar Available for use by
1099 u PERL_MAGIC_uvar_elem (none) Reserved for use by
1101 V PERL_MAGIC_vstring vtbl_vstring SV was vstring literal
1102 v PERL_MAGIC_vec vtbl_vec vec() lvalue
1103 w PERL_MAGIC_utf8 vtbl_utf8 Cached UTF-8 information
1104 x PERL_MAGIC_substr vtbl_substr substr() lvalue
1105 y PERL_MAGIC_defelem vtbl_defelem Shadow "foreach" iterator
1106 variable / smart parameter
1108 ] PERL_MAGIC_checkcall vtbl_checkcall inlining/mutation of call
1110 ~ PERL_MAGIC_ext (none) Available for use by
1113 =for mg_vtable.pl end
1115 When an uppercase and lowercase letter both exist in the table, then the
1116 uppercase letter is typically used to represent some kind of composite type
1117 (a list or a hash), and the lowercase letter is used to represent an element
1118 of that composite type. Some internals code makes use of this case
1119 relationship. However, 'v' and 'V' (vec and v-string) are in no way related.
1121 The C<PERL_MAGIC_ext> and C<PERL_MAGIC_uvar> magic types are defined
1122 specifically for use by extensions and will not be used by perl itself.
1123 Extensions can use C<PERL_MAGIC_ext> magic to 'attach' private information
1124 to variables (typically objects). This is especially useful because
1125 there is no way for normal perl code to corrupt this private information
1126 (unlike using extra elements of a hash object).
1128 Similarly, C<PERL_MAGIC_uvar> magic can be used much like tie() to call a
1129 C function any time a scalar's value is used or changed. The C<MAGIC>'s
1130 C<mg_ptr> field points to a C<ufuncs> structure:
1133 I32 (*uf_val)(pTHX_ IV, SV*);
1134 I32 (*uf_set)(pTHX_ IV, SV*);
1138 When the SV is read from or written to, the C<uf_val> or C<uf_set>
1139 function will be called with C<uf_index> as the first arg and a pointer to
1140 the SV as the second. A simple example of how to add C<PERL_MAGIC_uvar>
1141 magic is shown below. Note that the ufuncs structure is copied by
1142 sv_magic, so you can safely allocate it on the stack.
1150 uf.uf_val = &my_get_fn;
1151 uf.uf_set = &my_set_fn;
1153 sv_magic(sv, 0, PERL_MAGIC_uvar, (char*)&uf, sizeof(uf));
1155 Attaching C<PERL_MAGIC_uvar> to arrays is permissible but has no effect.
1157 For hashes there is a specialized hook that gives control over hash
1158 keys (but not values). This hook calls C<PERL_MAGIC_uvar> 'get' magic
1159 if the "set" function in the C<ufuncs> structure is NULL. The hook
1160 is activated whenever the hash is accessed with a key specified as
1161 an C<SV> through the functions C<hv_store_ent>, C<hv_fetch_ent>,
1162 C<hv_delete_ent>, and C<hv_exists_ent>. Accessing the key as a string
1163 through the functions without the C<..._ent> suffix circumvents the
1164 hook. See L<Hash::Util::FieldHash/GUTS> for a detailed description.
1166 Note that because multiple extensions may be using C<PERL_MAGIC_ext>
1167 or C<PERL_MAGIC_uvar> magic, it is important for extensions to take
1168 extra care to avoid conflict. Typically only using the magic on
1169 objects blessed into the same class as the extension is sufficient.
1170 For C<PERL_MAGIC_ext> magic, it is usually a good idea to define an
1171 C<MGVTBL>, even if all its fields will be C<0>, so that individual
1172 C<MAGIC> pointers can be identified as a particular kind of magic
1173 using their magic virtual table. C<mg_findext> provides an easy way
1176 STATIC MGVTBL my_vtbl = { 0, 0, 0, 0, 0, 0, 0, 0 };
1179 if ((mg = mg_findext(sv, PERL_MAGIC_ext, &my_vtbl))) {
1180 /* this is really ours, not another module's PERL_MAGIC_ext */
1181 my_priv_data_t *priv = (my_priv_data_t *)mg->mg_ptr;
1185 Also note that the C<sv_set*()> and C<sv_cat*()> functions described
1186 earlier do B<not> invoke 'set' magic on their targets. This must
1187 be done by the user either by calling the C<SvSETMAGIC()> macro after
1188 calling these functions, or by using one of the C<sv_set*_mg()> or
1189 C<sv_cat*_mg()> functions. Similarly, generic C code must call the
1190 C<SvGETMAGIC()> macro to invoke any 'get' magic if they use an SV
1191 obtained from external sources in functions that don't handle magic.
1192 See L<perlapi> for a description of these functions.
1193 For example, calls to the C<sv_cat*()> functions typically need to be
1194 followed by C<SvSETMAGIC()>, but they don't need a prior C<SvGETMAGIC()>
1195 since their implementation handles 'get' magic.
1197 =head2 Finding Magic
1199 MAGIC *mg_find(SV *sv, int type); /* Finds the magic pointer of that
1202 This routine returns a pointer to a C<MAGIC> structure stored in the SV.
1203 If the SV does not have that magical feature, C<NULL> is returned. If the
1204 SV has multiple instances of that magical feature, the first one will be
1205 returned. C<mg_findext> can be used to find a C<MAGIC> structure of an SV
1206 based on both its magic type and its magic virtual table:
1208 MAGIC *mg_findext(SV *sv, int type, MGVTBL *vtbl);
1210 Also, if the SV passed to C<mg_find> or C<mg_findext> is not of type
1211 SVt_PVMG, Perl may core dump.
1213 int mg_copy(SV* sv, SV* nsv, const char* key, STRLEN klen);
1215 This routine checks to see what types of magic C<sv> has. If the mg_type
1216 field is an uppercase letter, then the mg_obj is copied to C<nsv>, but
1217 the mg_type field is changed to be the lowercase letter.
1219 =head2 Understanding the Magic of Tied Hashes and Arrays
1221 Tied hashes and arrays are magical beasts of the C<PERL_MAGIC_tied>
1224 WARNING: As of the 5.004 release, proper usage of the array and hash
1225 access functions requires understanding a few caveats. Some
1226 of these caveats are actually considered bugs in the API, to be fixed
1227 in later releases, and are bracketed with [MAYCHANGE] below. If
1228 you find yourself actually applying such information in this section, be
1229 aware that the behavior may change in the future, umm, without warning.
1231 The perl tie function associates a variable with an object that implements
1232 the various GET, SET, etc methods. To perform the equivalent of the perl
1233 tie function from an XSUB, you must mimic this behaviour. The code below
1234 carries out the necessary steps - firstly it creates a new hash, and then
1235 creates a second hash which it blesses into the class which will implement
1236 the tie methods. Lastly it ties the two hashes together, and returns a
1237 reference to the new tied hash. Note that the code below does NOT call the
1238 TIEHASH method in the MyTie class -
1239 see L<Calling Perl Routines from within C Programs> for details on how
1250 tie = newRV_noinc((SV*)newHV());
1251 stash = gv_stashpv("MyTie", GV_ADD);
1252 sv_bless(tie, stash);
1253 hv_magic(hash, (GV*)tie, PERL_MAGIC_tied);
1254 RETVAL = newRV_noinc(hash);
1258 The C<av_store> function, when given a tied array argument, merely
1259 copies the magic of the array onto the value to be "stored", using
1260 C<mg_copy>. It may also return NULL, indicating that the value did not
1261 actually need to be stored in the array. [MAYCHANGE] After a call to
1262 C<av_store> on a tied array, the caller will usually need to call
1263 C<mg_set(val)> to actually invoke the perl level "STORE" method on the
1264 TIEARRAY object. If C<av_store> did return NULL, a call to
1265 C<SvREFCNT_dec(val)> will also be usually necessary to avoid a memory
1268 The previous paragraph is applicable verbatim to tied hash access using the
1269 C<hv_store> and C<hv_store_ent> functions as well.
1271 C<av_fetch> and the corresponding hash functions C<hv_fetch> and
1272 C<hv_fetch_ent> actually return an undefined mortal value whose magic
1273 has been initialized using C<mg_copy>. Note the value so returned does not
1274 need to be deallocated, as it is already mortal. [MAYCHANGE] But you will
1275 need to call C<mg_get()> on the returned value in order to actually invoke
1276 the perl level "FETCH" method on the underlying TIE object. Similarly,
1277 you may also call C<mg_set()> on the return value after possibly assigning
1278 a suitable value to it using C<sv_setsv>, which will invoke the "STORE"
1279 method on the TIE object. [/MAYCHANGE]
1282 In other words, the array or hash fetch/store functions don't really
1283 fetch and store actual values in the case of tied arrays and hashes. They
1284 merely call C<mg_copy> to attach magic to the values that were meant to be
1285 "stored" or "fetched". Later calls to C<mg_get> and C<mg_set> actually
1286 do the job of invoking the TIE methods on the underlying objects. Thus
1287 the magic mechanism currently implements a kind of lazy access to arrays
1290 Currently (as of perl version 5.004), use of the hash and array access
1291 functions requires the user to be aware of whether they are operating on
1292 "normal" hashes and arrays, or on their tied variants. The API may be
1293 changed to provide more transparent access to both tied and normal data
1294 types in future versions.
1297 You would do well to understand that the TIEARRAY and TIEHASH interfaces
1298 are mere sugar to invoke some perl method calls while using the uniform hash
1299 and array syntax. The use of this sugar imposes some overhead (typically
1300 about two to four extra opcodes per FETCH/STORE operation, in addition to
1301 the creation of all the mortal variables required to invoke the methods).
1302 This overhead will be comparatively small if the TIE methods are themselves
1303 substantial, but if they are only a few statements long, the overhead
1304 will not be insignificant.
1306 =head2 Localizing changes
1308 Perl has a very handy construction
1315 This construction is I<approximately> equivalent to
1324 The biggest difference is that the first construction would
1325 reinstate the initial value of $var, irrespective of how control exits
1326 the block: C<goto>, C<return>, C<die>/C<eval>, etc. It is a little bit
1327 more efficient as well.
1329 There is a way to achieve a similar task from C via Perl API: create a
1330 I<pseudo-block>, and arrange for some changes to be automatically
1331 undone at the end of it, either explicit, or via a non-local exit (via
1332 die()). A I<block>-like construct is created by a pair of
1333 C<ENTER>/C<LEAVE> macros (see L<perlcall/"Returning a Scalar">).
1334 Such a construct may be created specially for some important localized
1335 task, or an existing one (like boundaries of enclosing Perl
1336 subroutine/block, or an existing pair for freeing TMPs) may be
1337 used. (In the second case the overhead of additional localization must
1338 be almost negligible.) Note that any XSUB is automatically enclosed in
1339 an C<ENTER>/C<LEAVE> pair.
1341 Inside such a I<pseudo-block> the following service is available:
1345 =item C<SAVEINT(int i)>
1347 =item C<SAVEIV(IV i)>
1349 =item C<SAVEI32(I32 i)>
1351 =item C<SAVELONG(long i)>
1353 These macros arrange things to restore the value of integer variable
1354 C<i> at the end of enclosing I<pseudo-block>.
1356 =item C<SAVESPTR(s)>
1358 =item C<SAVEPPTR(p)>
1360 These macros arrange things to restore the value of pointers C<s> and
1361 C<p>. C<s> must be a pointer of a type which survives conversion to
1362 C<SV*> and back, C<p> should be able to survive conversion to C<char*>
1365 =item C<SAVEFREESV(SV *sv)>
1367 The refcount of C<sv> would be decremented at the end of
1368 I<pseudo-block>. This is similar to C<sv_2mortal> in that it is also a
1369 mechanism for doing a delayed C<SvREFCNT_dec>. However, while C<sv_2mortal>
1370 extends the lifetime of C<sv> until the beginning of the next statement,
1371 C<SAVEFREESV> extends it until the end of the enclosing scope. These
1372 lifetimes can be wildly different.
1374 Also compare C<SAVEMORTALIZESV>.
1376 =item C<SAVEMORTALIZESV(SV *sv)>
1378 Just like C<SAVEFREESV>, but mortalizes C<sv> at the end of the current
1379 scope instead of decrementing its reference count. This usually has the
1380 effect of keeping C<sv> alive until the statement that called the currently
1381 live scope has finished executing.
1383 =item C<SAVEFREEOP(OP *op)>
1385 The C<OP *> is op_free()ed at the end of I<pseudo-block>.
1387 =item C<SAVEFREEPV(p)>
1389 The chunk of memory which is pointed to by C<p> is Safefree()ed at the
1390 end of I<pseudo-block>.
1392 =item C<SAVECLEARSV(SV *sv)>
1394 Clears a slot in the current scratchpad which corresponds to C<sv> at
1395 the end of I<pseudo-block>.
1397 =item C<SAVEDELETE(HV *hv, char *key, I32 length)>
1399 The key C<key> of C<hv> is deleted at the end of I<pseudo-block>. The
1400 string pointed to by C<key> is Safefree()ed. If one has a I<key> in
1401 short-lived storage, the corresponding string may be reallocated like
1404 SAVEDELETE(PL_defstash, savepv(tmpbuf), strlen(tmpbuf));
1406 =item C<SAVEDESTRUCTOR(DESTRUCTORFUNC_NOCONTEXT_t f, void *p)>
1408 At the end of I<pseudo-block> the function C<f> is called with the
1411 =item C<SAVEDESTRUCTOR_X(DESTRUCTORFUNC_t f, void *p)>
1413 At the end of I<pseudo-block> the function C<f> is called with the
1414 implicit context argument (if any), and C<p>.
1416 =item C<SAVESTACK_POS()>
1418 The current offset on the Perl internal stack (cf. C<SP>) is restored
1419 at the end of I<pseudo-block>.
1423 The following API list contains functions, thus one needs to
1424 provide pointers to the modifiable data explicitly (either C pointers,
1425 or Perlish C<GV *>s). Where the above macros take C<int>, a similar
1426 function takes C<int *>.
1430 =item C<SV* save_scalar(GV *gv)>
1432 Equivalent to Perl code C<local $gv>.
1434 =item C<AV* save_ary(GV *gv)>
1436 =item C<HV* save_hash(GV *gv)>
1438 Similar to C<save_scalar>, but localize C<@gv> and C<%gv>.
1440 =item C<void save_item(SV *item)>
1442 Duplicates the current value of C<SV>, on the exit from the current
1443 C<ENTER>/C<LEAVE> I<pseudo-block> will restore the value of C<SV>
1444 using the stored value. It doesn't handle magic. Use C<save_scalar> if
1447 =item C<void save_list(SV **sarg, I32 maxsarg)>
1449 A variant of C<save_item> which takes multiple arguments via an array
1450 C<sarg> of C<SV*> of length C<maxsarg>.
1452 =item C<SV* save_svref(SV **sptr)>
1454 Similar to C<save_scalar>, but will reinstate an C<SV *>.
1456 =item C<void save_aptr(AV **aptr)>
1458 =item C<void save_hptr(HV **hptr)>
1460 Similar to C<save_svref>, but localize C<AV *> and C<HV *>.
1464 The C<Alias> module implements localization of the basic types within the
1465 I<caller's scope>. People who are interested in how to localize things in
1466 the containing scope should take a look there too.
1470 =head2 XSUBs and the Argument Stack
1472 The XSUB mechanism is a simple way for Perl programs to access C subroutines.
1473 An XSUB routine will have a stack that contains the arguments from the Perl
1474 program, and a way to map from the Perl data structures to a C equivalent.
1476 The stack arguments are accessible through the C<ST(n)> macro, which returns
1477 the C<n>'th stack argument. Argument 0 is the first argument passed in the
1478 Perl subroutine call. These arguments are C<SV*>, and can be used anywhere
1481 Most of the time, output from the C routine can be handled through use of
1482 the RETVAL and OUTPUT directives. However, there are some cases where the
1483 argument stack is not already long enough to handle all the return values.
1484 An example is the POSIX tzname() call, which takes no arguments, but returns
1485 two, the local time zone's standard and summer time abbreviations.
1487 To handle this situation, the PPCODE directive is used and the stack is
1488 extended using the macro:
1492 where C<SP> is the macro that represents the local copy of the stack pointer,
1493 and C<num> is the number of elements the stack should be extended by.
1495 Now that there is room on the stack, values can be pushed on it using C<PUSHs>
1496 macro. The pushed values will often need to be "mortal" (See
1497 L</Reference Counts and Mortality>):
1499 PUSHs(sv_2mortal(newSViv(an_integer)))
1500 PUSHs(sv_2mortal(newSVuv(an_unsigned_integer)))
1501 PUSHs(sv_2mortal(newSVnv(a_double)))
1502 PUSHs(sv_2mortal(newSVpv("Some String",0)))
1503 /* Although the last example is better written as the more
1505 PUSHs(newSVpvs_flags("Some String", SVs_TEMP))
1507 And now the Perl program calling C<tzname>, the two values will be assigned
1510 ($standard_abbrev, $summer_abbrev) = POSIX::tzname;
1512 An alternate (and possibly simpler) method to pushing values on the stack is
1517 This macro automatically adjusts the stack for you, if needed. Thus, you
1518 do not need to call C<EXTEND> to extend the stack.
1520 Despite their suggestions in earlier versions of this document the macros
1521 C<(X)PUSH[iunp]> are I<not> suited to XSUBs which return multiple results.
1522 For that, either stick to the C<(X)PUSHs> macros shown above, or use the new
1523 C<m(X)PUSH[iunp]> macros instead; see L</Putting a C value on Perl stack>.
1525 For more information, consult L<perlxs> and L<perlxstut>.
1527 =head2 Autoloading with XSUBs
1529 If an AUTOLOAD routine is an XSUB, as with Perl subroutines, Perl puts the
1530 fully-qualified name of the autoloaded subroutine in the $AUTOLOAD variable
1531 of the XSUB's package.
1533 But it also puts the same information in certain fields of the XSUB itself:
1535 HV *stash = CvSTASH(cv);
1536 const char *subname = SvPVX(cv);
1537 STRLEN name_length = SvCUR(cv); /* in bytes */
1538 U32 is_utf8 = SvUTF8(cv);
1540 C<SvPVX(cv)> contains just the sub name itself, not including the package.
1541 For an AUTOLOAD routine in UNIVERSAL or one of its superclasses,
1542 C<CvSTASH(cv)> returns NULL during a method call on a nonexistent package.
1544 B<Note>: Setting $AUTOLOAD stopped working in 5.6.1, which did not support
1545 XS AUTOLOAD subs at all. Perl 5.8.0 introduced the use of fields in the
1546 XSUB itself. Perl 5.16.0 restored the setting of $AUTOLOAD. If you need
1547 to support 5.8-5.14, use the XSUB's fields.
1549 =head2 Calling Perl Routines from within C Programs
1551 There are four routines that can be used to call a Perl subroutine from
1552 within a C program. These four are:
1554 I32 call_sv(SV*, I32);
1555 I32 call_pv(const char*, I32);
1556 I32 call_method(const char*, I32);
1557 I32 call_argv(const char*, I32, register char**);
1559 The routine most often used is C<call_sv>. The C<SV*> argument
1560 contains either the name of the Perl subroutine to be called, or a
1561 reference to the subroutine. The second argument consists of flags
1562 that control the context in which the subroutine is called, whether
1563 or not the subroutine is being passed arguments, how errors should be
1564 trapped, and how to treat return values.
1566 All four routines return the number of arguments that the subroutine returned
1569 These routines used to be called C<perl_call_sv>, etc., before Perl v5.6.0,
1570 but those names are now deprecated; macros of the same name are provided for
1573 When using any of these routines (except C<call_argv>), the programmer
1574 must manipulate the Perl stack. These include the following macros and
1589 For a detailed description of calling conventions from C to Perl,
1590 consult L<perlcall>.
1592 =head2 Memory Allocation
1596 All memory meant to be used with the Perl API functions should be manipulated
1597 using the macros described in this section. The macros provide the necessary
1598 transparency between differences in the actual malloc implementation that is
1601 It is suggested that you enable the version of malloc that is distributed
1602 with Perl. It keeps pools of various sizes of unallocated memory in
1603 order to satisfy allocation requests more quickly. However, on some
1604 platforms, it may cause spurious malloc or free errors.
1606 The following three macros are used to initially allocate memory :
1608 Newx(pointer, number, type);
1609 Newxc(pointer, number, type, cast);
1610 Newxz(pointer, number, type);
1612 The first argument C<pointer> should be the name of a variable that will
1613 point to the newly allocated memory.
1615 The second and third arguments C<number> and C<type> specify how many of
1616 the specified type of data structure should be allocated. The argument
1617 C<type> is passed to C<sizeof>. The final argument to C<Newxc>, C<cast>,
1618 should be used if the C<pointer> argument is different from the C<type>
1621 Unlike the C<Newx> and C<Newxc> macros, the C<Newxz> macro calls C<memzero>
1622 to zero out all the newly allocated memory.
1626 Renew(pointer, number, type);
1627 Renewc(pointer, number, type, cast);
1630 These three macros are used to change a memory buffer size or to free a
1631 piece of memory no longer needed. The arguments to C<Renew> and C<Renewc>
1632 match those of C<New> and C<Newc> with the exception of not needing the
1633 "magic cookie" argument.
1637 Move(source, dest, number, type);
1638 Copy(source, dest, number, type);
1639 Zero(dest, number, type);
1641 These three macros are used to move, copy, or zero out previously allocated
1642 memory. The C<source> and C<dest> arguments point to the source and
1643 destination starting points. Perl will move, copy, or zero out C<number>
1644 instances of the size of the C<type> data structure (using the C<sizeof>
1649 The most recent development releases of Perl have been experimenting with
1650 removing Perl's dependency on the "normal" standard I/O suite and allowing
1651 other stdio implementations to be used. This involves creating a new
1652 abstraction layer that then calls whichever implementation of stdio Perl
1653 was compiled with. All XSUBs should now use the functions in the PerlIO
1654 abstraction layer and not make any assumptions about what kind of stdio
1657 For a complete description of the PerlIO abstraction, consult L<perlapio>.
1659 =head2 Putting a C value on Perl stack
1661 A lot of opcodes (this is an elementary operation in the internal perl
1662 stack machine) put an SV* on the stack. However, as an optimization
1663 the corresponding SV is (usually) not recreated each time. The opcodes
1664 reuse specially assigned SVs (I<target>s) which are (as a corollary)
1665 not constantly freed/created.
1667 Each of the targets is created only once (but see
1668 L<Scratchpads and recursion> below), and when an opcode needs to put
1669 an integer, a double, or a string on stack, it just sets the
1670 corresponding parts of its I<target> and puts the I<target> on stack.
1672 The macro to put this target on stack is C<PUSHTARG>, and it is
1673 directly used in some opcodes, as well as indirectly in zillions of
1674 others, which use it via C<(X)PUSH[iunp]>.
1676 Because the target is reused, you must be careful when pushing multiple
1677 values on the stack. The following code will not do what you think:
1682 This translates as "set C<TARG> to 10, push a pointer to C<TARG> onto
1683 the stack; set C<TARG> to 20, push a pointer to C<TARG> onto the stack".
1684 At the end of the operation, the stack does not contain the values 10
1685 and 20, but actually contains two pointers to C<TARG>, which we have set
1688 If you need to push multiple different values then you should either use
1689 the C<(X)PUSHs> macros, or else use the new C<m(X)PUSH[iunp]> macros,
1690 none of which make use of C<TARG>. The C<(X)PUSHs> macros simply push an
1691 SV* on the stack, which, as noted under L</XSUBs and the Argument Stack>,
1692 will often need to be "mortal". The new C<m(X)PUSH[iunp]> macros make
1693 this a little easier to achieve by creating a new mortal for you (via
1694 C<(X)PUSHmortal>), pushing that onto the stack (extending it if necessary
1695 in the case of the C<mXPUSH[iunp]> macros), and then setting its value.
1696 Thus, instead of writing this to "fix" the example above:
1698 XPUSHs(sv_2mortal(newSViv(10)))
1699 XPUSHs(sv_2mortal(newSViv(20)))
1701 you can simply write:
1706 On a related note, if you do use C<(X)PUSH[iunp]>, then you're going to
1707 need a C<dTARG> in your variable declarations so that the C<*PUSH*>
1708 macros can make use of the local variable C<TARG>. See also C<dTARGET>
1713 The question remains on when the SVs which are I<target>s for opcodes
1714 are created. The answer is that they are created when the current
1715 unit--a subroutine or a file (for opcodes for statements outside of
1716 subroutines)--is compiled. During this time a special anonymous Perl
1717 array is created, which is called a scratchpad for the current unit.
1719 A scratchpad keeps SVs which are lexicals for the current unit and are
1720 targets for opcodes. One can deduce that an SV lives on a scratchpad
1721 by looking on its flags: lexicals have C<SVs_PADMY> set, and
1722 I<target>s have C<SVs_PADTMP> set.
1724 The correspondence between OPs and I<target>s is not 1-to-1. Different
1725 OPs in the compile tree of the unit can use the same target, if this
1726 would not conflict with the expected life of the temporary.
1728 =head2 Scratchpads and recursion
1730 In fact it is not 100% true that a compiled unit contains a pointer to
1731 the scratchpad AV. In fact it contains a pointer to an AV of
1732 (initially) one element, and this element is the scratchpad AV. Why do
1733 we need an extra level of indirection?
1735 The answer is B<recursion>, and maybe B<threads>. Both
1736 these can create several execution pointers going into the same
1737 subroutine. For the subroutine-child not write over the temporaries
1738 for the subroutine-parent (lifespan of which covers the call to the
1739 child), the parent and the child should have different
1740 scratchpads. (I<And> the lexicals should be separate anyway!)
1742 So each subroutine is born with an array of scratchpads (of length 1).
1743 On each entry to the subroutine it is checked that the current
1744 depth of the recursion is not more than the length of this array, and
1745 if it is, new scratchpad is created and pushed into the array.
1747 The I<target>s on this scratchpad are C<undef>s, but they are already
1748 marked with correct flags.
1750 =head1 Compiled code
1754 Here we describe the internal form your code is converted to by
1755 Perl. Start with a simple example:
1759 This is converted to a tree similar to this one:
1767 (but slightly more complicated). This tree reflects the way Perl
1768 parsed your code, but has nothing to do with the execution order.
1769 There is an additional "thread" going through the nodes of the tree
1770 which shows the order of execution of the nodes. In our simplified
1771 example above it looks like:
1773 $b ---> $c ---> + ---> $a ---> assign-to
1775 But with the actual compile tree for C<$a = $b + $c> it is different:
1776 some nodes I<optimized away>. As a corollary, though the actual tree
1777 contains more nodes than our simplified example, the execution order
1778 is the same as in our example.
1780 =head2 Examining the tree
1782 If you have your perl compiled for debugging (usually done with
1783 C<-DDEBUGGING> on the C<Configure> command line), you may examine the
1784 compiled tree by specifying C<-Dx> on the Perl command line. The
1785 output takes several lines per node, and for C<$b+$c> it looks like
1790 FLAGS = (SCALAR,KIDS)
1792 TYPE = null ===> (4)
1794 FLAGS = (SCALAR,KIDS)
1796 3 TYPE = gvsv ===> 4
1802 TYPE = null ===> (5)
1804 FLAGS = (SCALAR,KIDS)
1806 4 TYPE = gvsv ===> 5
1812 This tree has 5 nodes (one per C<TYPE> specifier), only 3 of them are
1813 not optimized away (one per number in the left column). The immediate
1814 children of the given node correspond to C<{}> pairs on the same level
1815 of indentation, thus this listing corresponds to the tree:
1823 The execution order is indicated by C<===E<gt>> marks, thus it is C<3
1824 4 5 6> (node C<6> is not included into above listing), i.e.,
1825 C<gvsv gvsv add whatever>.
1827 Each of these nodes represents an op, a fundamental operation inside the
1828 Perl core. The code which implements each operation can be found in the
1829 F<pp*.c> files; the function which implements the op with type C<gvsv>
1830 is C<pp_gvsv>, and so on. As the tree above shows, different ops have
1831 different numbers of children: C<add> is a binary operator, as one would
1832 expect, and so has two children. To accommodate the various different
1833 numbers of children, there are various types of op data structure, and
1834 they link together in different ways.
1836 The simplest type of op structure is C<OP>: this has no children. Unary
1837 operators, C<UNOP>s, have one child, and this is pointed to by the
1838 C<op_first> field. Binary operators (C<BINOP>s) have not only an
1839 C<op_first> field but also an C<op_last> field. The most complex type of
1840 op is a C<LISTOP>, which has any number of children. In this case, the
1841 first child is pointed to by C<op_first> and the last child by
1842 C<op_last>. The children in between can be found by iteratively
1843 following the C<op_sibling> pointer from the first child to the last.
1845 There are also two other op types: a C<PMOP> holds a regular expression,
1846 and has no children, and a C<LOOP> may or may not have children. If the
1847 C<op_children> field is non-zero, it behaves like a C<LISTOP>. To
1848 complicate matters, if a C<UNOP> is actually a C<null> op after
1849 optimization (see L</Compile pass 2: context propagation>) it will still
1850 have children in accordance with its former type.
1852 Another way to examine the tree is to use a compiler back-end module, such
1855 =head2 Compile pass 1: check routines
1857 The tree is created by the compiler while I<yacc> code feeds it
1858 the constructions it recognizes. Since I<yacc> works bottom-up, so does
1859 the first pass of perl compilation.
1861 What makes this pass interesting for perl developers is that some
1862 optimization may be performed on this pass. This is optimization by
1863 so-called "check routines". The correspondence between node names
1864 and corresponding check routines is described in F<opcode.pl> (do not
1865 forget to run C<make regen_headers> if you modify this file).
1867 A check routine is called when the node is fully constructed except
1868 for the execution-order thread. Since at this time there are no
1869 back-links to the currently constructed node, one can do most any
1870 operation to the top-level node, including freeing it and/or creating
1871 new nodes above/below it.
1873 The check routine returns the node which should be inserted into the
1874 tree (if the top-level node was not modified, check routine returns
1877 By convention, check routines have names C<ck_*>. They are usually
1878 called from C<new*OP> subroutines (or C<convert>) (which in turn are
1879 called from F<perly.y>).
1881 =head2 Compile pass 1a: constant folding
1883 Immediately after the check routine is called the returned node is
1884 checked for being compile-time executable. If it is (the value is
1885 judged to be constant) it is immediately executed, and a I<constant>
1886 node with the "return value" of the corresponding subtree is
1887 substituted instead. The subtree is deleted.
1889 If constant folding was not performed, the execution-order thread is
1892 =head2 Compile pass 2: context propagation
1894 When a context for a part of compile tree is known, it is propagated
1895 down through the tree. At this time the context can have 5 values
1896 (instead of 2 for runtime context): void, boolean, scalar, list, and
1897 lvalue. In contrast with the pass 1 this pass is processed from top
1898 to bottom: a node's context determines the context for its children.
1900 Additional context-dependent optimizations are performed at this time.
1901 Since at this moment the compile tree contains back-references (via
1902 "thread" pointers), nodes cannot be free()d now. To allow
1903 optimized-away nodes at this stage, such nodes are null()ified instead
1904 of free()ing (i.e. their type is changed to OP_NULL).
1906 =head2 Compile pass 3: peephole optimization
1908 After the compile tree for a subroutine (or for an C<eval> or a file)
1909 is created, an additional pass over the code is performed. This pass
1910 is neither top-down or bottom-up, but in the execution order (with
1911 additional complications for conditionals). Optimizations performed
1912 at this stage are subject to the same restrictions as in the pass 2.
1914 Peephole optimizations are done by calling the function pointed to
1915 by the global variable C<PL_peepp>. By default, C<PL_peepp> just
1916 calls the function pointed to by the global variable C<PL_rpeepp>.
1917 By default, that performs some basic op fixups and optimisations along
1918 the execution-order op chain, and recursively calls C<PL_rpeepp> for
1919 each side chain of ops (resulting from conditionals). Extensions may
1920 provide additional optimisations or fixups, hooking into either the
1921 per-subroutine or recursive stage, like this:
1923 static peep_t prev_peepp;
1924 static void my_peep(pTHX_ OP *o)
1926 /* custom per-subroutine optimisation goes here */
1928 /* custom per-subroutine optimisation may also go here */
1931 prev_peepp = PL_peepp;
1934 static peep_t prev_rpeepp;
1935 static void my_rpeep(pTHX_ OP *o)
1938 for(; o; o = o->op_next) {
1939 /* custom per-op optimisation goes here */
1941 prev_rpeepp(orig_o);
1944 prev_rpeepp = PL_rpeepp;
1945 PL_rpeepp = my_rpeep;
1947 =head2 Pluggable runops
1949 The compile tree is executed in a runops function. There are two runops
1950 functions, in F<run.c> and in F<dump.c>. C<Perl_runops_debug> is used
1951 with DEBUGGING and C<Perl_runops_standard> is used otherwise. For fine
1952 control over the execution of the compile tree it is possible to provide
1953 your own runops function.
1955 It's probably best to copy one of the existing runops functions and
1956 change it to suit your needs. Then, in the BOOT section of your XS
1959 PL_runops = my_runops;
1961 This function should be as efficient as possible to keep your programs
1962 running as fast as possible.
1964 =head2 Compile-time scope hooks
1966 As of perl 5.14 it is possible to hook into the compile-time lexical
1967 scope mechanism using C<Perl_blockhook_register>. This is used like
1970 STATIC void my_start_hook(pTHX_ int full);
1971 STATIC BHK my_hooks;
1974 BhkENTRY_set(&my_hooks, bhk_start, my_start_hook);
1975 Perl_blockhook_register(aTHX_ &my_hooks);
1977 This will arrange to have C<my_start_hook> called at the start of
1978 compiling every lexical scope. The available hooks are:
1982 =item C<void bhk_start(pTHX_ int full)>
1984 This is called just after starting a new lexical scope. Note that Perl
1989 creates two scopes: the first starts at the C<(> and has C<full == 1>,
1990 the second starts at the C<{> and has C<full == 0>. Both end at the
1991 C<}>, so calls to C<start> and C<pre/post_end> will match. Anything
1992 pushed onto the save stack by this hook will be popped just before the
1993 scope ends (between the C<pre_> and C<post_end> hooks, in fact).
1995 =item C<void bhk_pre_end(pTHX_ OP **o)>
1997 This is called at the end of a lexical scope, just before unwinding the
1998 stack. I<o> is the root of the optree representing the scope; it is a
1999 double pointer so you can replace the OP if you need to.
2001 =item C<void bhk_post_end(pTHX_ OP **o)>
2003 This is called at the end of a lexical scope, just after unwinding the
2004 stack. I<o> is as above. Note that it is possible for calls to C<pre_>
2005 and C<post_end> to nest, if there is something on the save stack that
2008 =item C<void bhk_eval(pTHX_ OP *const o)>
2010 This is called just before starting to compile an C<eval STRING>, C<do
2011 FILE>, C<require> or C<use>, after the eval has been set up. I<o> is the
2012 OP that requested the eval, and will normally be an C<OP_ENTEREVAL>,
2013 C<OP_DOFILE> or C<OP_REQUIRE>.
2017 Once you have your hook functions, you need a C<BHK> structure to put
2018 them in. It's best to allocate it statically, since there is no way to
2019 free it once it's registered. The function pointers should be inserted
2020 into this structure using the C<BhkENTRY_set> macro, which will also set
2021 flags indicating which entries are valid. If you do need to allocate
2022 your C<BHK> dynamically for some reason, be sure to zero it before you
2025 Once registered, there is no mechanism to switch these hooks off, so if
2026 that is necessary you will need to do this yourself. An entry in C<%^H>
2027 is probably the best way, so the effect is lexically scoped; however it
2028 is also possible to use the C<BhkDISABLE> and C<BhkENABLE> macros to
2029 temporarily switch entries on and off. You should also be aware that
2030 generally speaking at least one scope will have opened before your
2031 extension is loaded, so you will see some C<pre/post_end> pairs that
2032 didn't have a matching C<start>.
2034 =head1 Examining internal data structures with the C<dump> functions
2036 To aid debugging, the source file F<dump.c> contains a number of
2037 functions which produce formatted output of internal data structures.
2039 The most commonly used of these functions is C<Perl_sv_dump>; it's used
2040 for dumping SVs, AVs, HVs, and CVs. The C<Devel::Peek> module calls
2041 C<sv_dump> to produce debugging output from Perl-space, so users of that
2042 module should already be familiar with its format.
2044 C<Perl_op_dump> can be used to dump an C<OP> structure or any of its
2045 derivatives, and produces output similar to C<perl -Dx>; in fact,
2046 C<Perl_dump_eval> will dump the main root of the code being evaluated,
2047 exactly like C<-Dx>.
2049 Other useful functions are C<Perl_dump_sub>, which turns a C<GV> into an
2050 op tree, C<Perl_dump_packsubs> which calls C<Perl_dump_sub> on all the
2051 subroutines in a package like so: (Thankfully, these are all xsubs, so
2052 there is no op tree)
2054 (gdb) print Perl_dump_packsubs(PL_defstash)
2056 SUB attributes::bootstrap = (xsub 0x811fedc 0)
2058 SUB UNIVERSAL::can = (xsub 0x811f50c 0)
2060 SUB UNIVERSAL::isa = (xsub 0x811f304 0)
2062 SUB UNIVERSAL::VERSION = (xsub 0x811f7ac 0)
2064 SUB DynaLoader::boot_DynaLoader = (xsub 0x805b188 0)
2066 and C<Perl_dump_all>, which dumps all the subroutines in the stash and
2067 the op tree of the main root.
2069 =head1 How multiple interpreters and concurrency are supported
2071 =head2 Background and PERL_IMPLICIT_CONTEXT
2073 The Perl interpreter can be regarded as a closed box: it has an API
2074 for feeding it code or otherwise making it do things, but it also has
2075 functions for its own use. This smells a lot like an object, and
2076 there are ways for you to build Perl so that you can have multiple
2077 interpreters, with one interpreter represented either as a C structure,
2078 or inside a thread-specific structure. These structures contain all
2079 the context, the state of that interpreter.
2081 One macro controls the major Perl build flavor: MULTIPLICITY. The
2082 MULTIPLICITY build has a C structure that packages all the interpreter
2083 state. With multiplicity-enabled perls, PERL_IMPLICIT_CONTEXT is also
2084 normally defined, and enables the support for passing in a "hidden" first
2085 argument that represents all three data structures. MULTIPLICITY makes
2086 multi-threaded perls possible (with the ithreads threading model, related
2087 to the macro USE_ITHREADS.)
2089 Two other "encapsulation" macros are the PERL_GLOBAL_STRUCT and
2090 PERL_GLOBAL_STRUCT_PRIVATE (the latter turns on the former, and the
2091 former turns on MULTIPLICITY.) The PERL_GLOBAL_STRUCT causes all the
2092 internal variables of Perl to be wrapped inside a single global struct,
2093 struct perl_vars, accessible as (globals) &PL_Vars or PL_VarsPtr or
2094 the function Perl_GetVars(). The PERL_GLOBAL_STRUCT_PRIVATE goes
2095 one step further, there is still a single struct (allocated in main()
2096 either from heap or from stack) but there are no global data symbols
2097 pointing to it. In either case the global struct should be initialised
2098 as the very first thing in main() using Perl_init_global_struct() and
2099 correspondingly tear it down after perl_free() using Perl_free_global_struct(),
2100 please see F<miniperlmain.c> for usage details. You may also need
2101 to use C<dVAR> in your coding to "declare the global variables"
2102 when you are using them. dTHX does this for you automatically.
2104 To see whether you have non-const data you can use a BSD-compatible C<nm>:
2106 nm libperl.a | grep -v ' [TURtr] '
2108 If this displays any C<D> or C<d> symbols, you have non-const data.
2110 For backward compatibility reasons defining just PERL_GLOBAL_STRUCT
2111 doesn't actually hide all symbols inside a big global struct: some
2112 PerlIO_xxx vtables are left visible. The PERL_GLOBAL_STRUCT_PRIVATE
2113 then hides everything (see how the PERLIO_FUNCS_DECL is used).
2115 All this obviously requires a way for the Perl internal functions to be
2116 either subroutines taking some kind of structure as the first
2117 argument, or subroutines taking nothing as the first argument. To
2118 enable these two very different ways of building the interpreter,
2119 the Perl source (as it does in so many other situations) makes heavy
2120 use of macros and subroutine naming conventions.
2122 First problem: deciding which functions will be public API functions and
2123 which will be private. All functions whose names begin C<S_> are private
2124 (think "S" for "secret" or "static"). All other functions begin with
2125 "Perl_", but just because a function begins with "Perl_" does not mean it is
2126 part of the API. (See L</Internal Functions>.) The easiest way to be B<sure> a
2127 function is part of the API is to find its entry in L<perlapi>.
2128 If it exists in L<perlapi>, it's part of the API. If it doesn't, and you
2129 think it should be (i.e., you need it for your extension), send mail via
2130 L<perlbug> explaining why you think it should be.
2132 Second problem: there must be a syntax so that the same subroutine
2133 declarations and calls can pass a structure as their first argument,
2134 or pass nothing. To solve this, the subroutines are named and
2135 declared in a particular way. Here's a typical start of a static
2136 function used within the Perl guts:
2139 S_incline(pTHX_ char *s)
2141 STATIC becomes "static" in C, and may be #define'd to nothing in some
2142 configurations in the future.
2144 A public function (i.e. part of the internal API, but not necessarily
2145 sanctioned for use in extensions) begins like this:
2148 Perl_sv_setiv(pTHX_ SV* dsv, IV num)
2150 C<pTHX_> is one of a number of macros (in F<perl.h>) that hide the
2151 details of the interpreter's context. THX stands for "thread", "this",
2152 or "thingy", as the case may be. (And no, George Lucas is not involved. :-)
2153 The first character could be 'p' for a B<p>rototype, 'a' for B<a>rgument,
2154 or 'd' for B<d>eclaration, so we have C<pTHX>, C<aTHX> and C<dTHX>, and
2157 When Perl is built without options that set PERL_IMPLICIT_CONTEXT, there is no
2158 first argument containing the interpreter's context. The trailing underscore
2159 in the pTHX_ macro indicates that the macro expansion needs a comma
2160 after the context argument because other arguments follow it. If
2161 PERL_IMPLICIT_CONTEXT is not defined, pTHX_ will be ignored, and the
2162 subroutine is not prototyped to take the extra argument. The form of the
2163 macro without the trailing underscore is used when there are no additional
2166 When a core function calls another, it must pass the context. This
2167 is normally hidden via macros. Consider C<sv_setiv>. It expands into
2168 something like this:
2170 #ifdef PERL_IMPLICIT_CONTEXT
2171 #define sv_setiv(a,b) Perl_sv_setiv(aTHX_ a, b)
2172 /* can't do this for vararg functions, see below */
2174 #define sv_setiv Perl_sv_setiv
2177 This works well, and means that XS authors can gleefully write:
2181 and still have it work under all the modes Perl could have been
2184 This doesn't work so cleanly for varargs functions, though, as macros
2185 imply that the number of arguments is known in advance. Instead we
2186 either need to spell them out fully, passing C<aTHX_> as the first
2187 argument (the Perl core tends to do this with functions like
2188 Perl_warner), or use a context-free version.
2190 The context-free version of Perl_warner is called
2191 Perl_warner_nocontext, and does not take the extra argument. Instead
2192 it does dTHX; to get the context from thread-local storage. We
2193 C<#define warner Perl_warner_nocontext> so that extensions get source
2194 compatibility at the expense of performance. (Passing an arg is
2195 cheaper than grabbing it from thread-local storage.)
2197 You can ignore [pad]THXx when browsing the Perl headers/sources.
2198 Those are strictly for use within the core. Extensions and embedders
2199 need only be aware of [pad]THX.
2201 =head2 So what happened to dTHR?
2203 C<dTHR> was introduced in perl 5.005 to support the older thread model.
2204 The older thread model now uses the C<THX> mechanism to pass context
2205 pointers around, so C<dTHR> is not useful any more. Perl 5.6.0 and
2206 later still have it for backward source compatibility, but it is defined
2209 =head2 How do I use all this in extensions?
2211 When Perl is built with PERL_IMPLICIT_CONTEXT, extensions that call
2212 any functions in the Perl API will need to pass the initial context
2213 argument somehow. The kicker is that you will need to write it in
2214 such a way that the extension still compiles when Perl hasn't been
2215 built with PERL_IMPLICIT_CONTEXT enabled.
2217 There are three ways to do this. First, the easy but inefficient way,
2218 which is also the default, in order to maintain source compatibility
2219 with extensions: whenever F<XSUB.h> is #included, it redefines the aTHX
2220 and aTHX_ macros to call a function that will return the context.
2221 Thus, something like:
2225 in your extension will translate to this when PERL_IMPLICIT_CONTEXT is
2228 Perl_sv_setiv(Perl_get_context(), sv, num);
2230 or to this otherwise:
2232 Perl_sv_setiv(sv, num);
2234 You don't have to do anything new in your extension to get this; since
2235 the Perl library provides Perl_get_context(), it will all just
2238 The second, more efficient way is to use the following template for
2241 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2246 STATIC void my_private_function(int arg1, int arg2);
2249 my_private_function(int arg1, int arg2)
2251 dTHX; /* fetch context */
2252 ... call many Perl API functions ...
2257 MODULE = Foo PACKAGE = Foo
2265 my_private_function(arg, 10);
2267 Note that the only two changes from the normal way of writing an
2268 extension is the addition of a C<#define PERL_NO_GET_CONTEXT> before
2269 including the Perl headers, followed by a C<dTHX;> declaration at
2270 the start of every function that will call the Perl API. (You'll
2271 know which functions need this, because the C compiler will complain
2272 that there's an undeclared identifier in those functions.) No changes
2273 are needed for the XSUBs themselves, because the XS() macro is
2274 correctly defined to pass in the implicit context if needed.
2276 The third, even more efficient way is to ape how it is done within
2280 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2285 /* pTHX_ only needed for functions that call Perl API */
2286 STATIC void my_private_function(pTHX_ int arg1, int arg2);
2289 my_private_function(pTHX_ int arg1, int arg2)
2291 /* dTHX; not needed here, because THX is an argument */
2292 ... call Perl API functions ...
2297 MODULE = Foo PACKAGE = Foo
2305 my_private_function(aTHX_ arg, 10);
2307 This implementation never has to fetch the context using a function
2308 call, since it is always passed as an extra argument. Depending on
2309 your needs for simplicity or efficiency, you may mix the previous
2310 two approaches freely.
2312 Never add a comma after C<pTHX> yourself--always use the form of the
2313 macro with the underscore for functions that take explicit arguments,
2314 or the form without the argument for functions with no explicit arguments.
2316 If one is compiling Perl with the C<-DPERL_GLOBAL_STRUCT> the C<dVAR>
2317 definition is needed if the Perl global variables (see F<perlvars.h>
2318 or F<globvar.sym>) are accessed in the function and C<dTHX> is not
2319 used (the C<dTHX> includes the C<dVAR> if necessary). One notices
2320 the need for C<dVAR> only with the said compile-time define, because
2321 otherwise the Perl global variables are visible as-is.
2323 =head2 Should I do anything special if I call perl from multiple threads?
2325 If you create interpreters in one thread and then proceed to call them in
2326 another, you need to make sure perl's own Thread Local Storage (TLS) slot is
2327 initialized correctly in each of those threads.
2329 The C<perl_alloc> and C<perl_clone> API functions will automatically set
2330 the TLS slot to the interpreter they created, so that there is no need to do
2331 anything special if the interpreter is always accessed in the same thread that
2332 created it, and that thread did not create or call any other interpreters
2333 afterwards. If that is not the case, you have to set the TLS slot of the
2334 thread before calling any functions in the Perl API on that particular
2335 interpreter. This is done by calling the C<PERL_SET_CONTEXT> macro in that
2336 thread as the first thing you do:
2338 /* do this before doing anything else with some_perl */
2339 PERL_SET_CONTEXT(some_perl);
2341 ... other Perl API calls on some_perl go here ...
2343 =head2 Future Plans and PERL_IMPLICIT_SYS
2345 Just as PERL_IMPLICIT_CONTEXT provides a way to bundle up everything
2346 that the interpreter knows about itself and pass it around, so too are
2347 there plans to allow the interpreter to bundle up everything it knows
2348 about the environment it's running on. This is enabled with the
2349 PERL_IMPLICIT_SYS macro. Currently it only works with USE_ITHREADS on
2352 This allows the ability to provide an extra pointer (called the "host"
2353 environment) for all the system calls. This makes it possible for
2354 all the system stuff to maintain their own state, broken down into
2355 seven C structures. These are thin wrappers around the usual system
2356 calls (see F<win32/perllib.c>) for the default perl executable, but for a
2357 more ambitious host (like the one that would do fork() emulation) all
2358 the extra work needed to pretend that different interpreters are
2359 actually different "processes", would be done here.
2361 The Perl engine/interpreter and the host are orthogonal entities.
2362 There could be one or more interpreters in a process, and one or
2363 more "hosts", with free association between them.
2365 =head1 Internal Functions
2367 All of Perl's internal functions which will be exposed to the outside
2368 world are prefixed by C<Perl_> so that they will not conflict with XS
2369 functions or functions used in a program in which Perl is embedded.
2370 Similarly, all global variables begin with C<PL_>. (By convention,
2371 static functions start with C<S_>.)
2373 Inside the Perl core (C<PERL_CORE> defined), you can get at the functions
2374 either with or without the C<Perl_> prefix, thanks to a bunch of defines
2375 that live in F<embed.h>. Note that extension code should I<not> set
2376 C<PERL_CORE>; this exposes the full perl internals, and is likely to cause
2377 breakage of the XS in each new perl release.
2379 The file F<embed.h> is generated automatically from
2380 F<embed.pl> and F<embed.fnc>. F<embed.pl> also creates the prototyping
2381 header files for the internal functions, generates the documentation
2382 and a lot of other bits and pieces. It's important that when you add
2383 a new function to the core or change an existing one, you change the
2384 data in the table in F<embed.fnc> as well. Here's a sample entry from
2387 Apd |SV** |av_fetch |AV* ar|I32 key|I32 lval
2389 The second column is the return type, the third column the name. Columns
2390 after that are the arguments. The first column is a set of flags:
2396 This function is a part of the public API. All such functions should also
2397 have 'd', very few do not.
2401 This function has a C<Perl_> prefix; i.e. it is defined as
2406 This function has documentation using the C<apidoc> feature which we'll
2407 look at in a second. Some functions have 'd' but not 'A'; docs are good.
2411 Other available flags are:
2417 This is a static function and is defined as C<STATIC S_whatever>, and
2418 usually called within the sources as C<whatever(...)>.
2422 This does not need an interpreter context, so the definition has no
2423 C<pTHX>, and it follows that callers don't use C<aTHX>. (See
2424 L</Background and PERL_IMPLICIT_CONTEXT>.)
2428 This function never returns; C<croak>, C<exit> and friends.
2432 This function takes a variable number of arguments, C<printf> style.
2433 The argument list should end with C<...>, like this:
2435 Afprd |void |croak |const char* pat|...
2439 This function is part of the experimental development API, and may change
2440 or disappear without notice.
2444 This function should not have a compatibility macro to define, say,
2445 C<Perl_parse> to C<parse>. It must be called as C<Perl_parse>.
2449 This function isn't exported out of the Perl core.
2453 This is implemented as a macro.
2457 This function is explicitly exported.
2461 This function is visible to extensions included in the Perl core.
2465 Binary backward compatibility; this function is a macro but also has
2466 a C<Perl_> implementation (which is exported).
2470 See the comments at the top of C<embed.fnc> for others.
2474 If you edit F<embed.pl> or F<embed.fnc>, you will need to run
2475 C<make regen_headers> to force a rebuild of F<embed.h> and other
2476 auto-generated files.
2478 =head2 Formatted Printing of IVs, UVs, and NVs
2480 If you are printing IVs, UVs, or NVS instead of the stdio(3) style
2481 formatting codes like C<%d>, C<%ld>, C<%f>, you should use the
2482 following macros for portability
2487 UVxf UV in hexadecimal
2492 These will take care of 64-bit integers and long doubles.
2495 printf("IV is %"IVdf"\n", iv);
2497 The IVdf will expand to whatever is the correct format for the IVs.
2499 If you are printing addresses of pointers, use UVxf combined
2500 with PTR2UV(), do not use %lx or %p.
2502 =head2 Pointer-To-Integer and Integer-To-Pointer
2504 Because pointer size does not necessarily equal integer size,
2505 use the follow macros to do it right.
2510 INT2PTR(pointertotype, integer)
2515 SV *sv = INT2PTR(SV*, iv);
2522 =head2 Exception Handling
2524 There are a couple of macros to do very basic exception handling in XS
2525 modules. You have to define C<NO_XSLOCKS> before including F<XSUB.h> to
2526 be able to use these macros:
2531 You can use these macros if you call code that may croak, but you need
2532 to do some cleanup before giving control back to Perl. For example:
2534 dXCPT; /* set up necessary variables */
2537 code_that_may_croak();
2542 /* do cleanup here */
2546 Note that you always have to rethrow an exception that has been
2547 caught. Using these macros, it is not possible to just catch the
2548 exception and ignore it. If you have to ignore the exception, you
2549 have to use the C<call_*> function.
2551 The advantage of using the above macros is that you don't have
2552 to setup an extra function for C<call_*>, and that using these
2553 macros is faster than using C<call_*>.
2555 =head2 Source Documentation
2557 There's an effort going on to document the internal functions and
2558 automatically produce reference manuals from them - L<perlapi> is one
2559 such manual which details all the functions which are available to XS
2560 writers. L<perlintern> is the autogenerated manual for the functions
2561 which are not part of the API and are supposedly for internal use only.
2563 Source documentation is created by putting POD comments into the C
2567 =for apidoc sv_setiv
2569 Copies an integer into the given SV. Does not handle 'set' magic. See
2575 Please try and supply some documentation if you add functions to the
2578 =head2 Backwards compatibility
2580 The Perl API changes over time. New functions are added or the interfaces
2581 of existing functions are changed. The C<Devel::PPPort> module tries to
2582 provide compatibility code for some of these changes, so XS writers don't
2583 have to code it themselves when supporting multiple versions of Perl.
2585 C<Devel::PPPort> generates a C header file F<ppport.h> that can also
2586 be run as a Perl script. To generate F<ppport.h>, run:
2588 perl -MDevel::PPPort -eDevel::PPPort::WriteFile
2590 Besides checking existing XS code, the script can also be used to retrieve
2591 compatibility information for various API calls using the C<--api-info>
2592 command line switch. For example:
2594 % perl ppport.h --api-info=sv_magicext
2596 For details, see C<perldoc ppport.h>.
2598 =head1 Unicode Support
2600 Perl 5.6.0 introduced Unicode support. It's important for porters and XS
2601 writers to understand this support and make sure that the code they
2602 write does not corrupt Unicode data.
2604 =head2 What B<is> Unicode, anyway?
2606 In the olden, less enlightened times, we all used to use ASCII. Most of
2607 us did, anyway. The big problem with ASCII is that it's American. Well,
2608 no, that's not actually the problem; the problem is that it's not
2609 particularly useful for people who don't use the Roman alphabet. What
2610 used to happen was that particular languages would stick their own
2611 alphabet in the upper range of the sequence, between 128 and 255. Of
2612 course, we then ended up with plenty of variants that weren't quite
2613 ASCII, and the whole point of it being a standard was lost.
2615 Worse still, if you've got a language like Chinese or
2616 Japanese that has hundreds or thousands of characters, then you really
2617 can't fit them into a mere 256, so they had to forget about ASCII
2618 altogether, and build their own systems using pairs of numbers to refer
2621 To fix this, some people formed Unicode, Inc. and
2622 produced a new character set containing all the characters you can
2623 possibly think of and more. There are several ways of representing these
2624 characters, and the one Perl uses is called UTF-8. UTF-8 uses
2625 a variable number of bytes to represent a character. You can learn more
2626 about Unicode and Perl's Unicode model in L<perlunicode>.
2628 =head2 How can I recognise a UTF-8 string?
2630 You can't. This is because UTF-8 data is stored in bytes just like
2631 non-UTF-8 data. The Unicode character 200, (C<0xC8> for you hex types)
2632 capital E with a grave accent, is represented by the two bytes
2633 C<v196.172>. Unfortunately, the non-Unicode string C<chr(196).chr(172)>
2634 has that byte sequence as well. So you can't tell just by looking - this
2635 is what makes Unicode input an interesting problem.
2637 In general, you either have to know what you're dealing with, or you
2638 have to guess. The API function C<is_utf8_string> can help; it'll tell
2639 you if a string contains only valid UTF-8 characters. However, it can't
2640 do the work for you. On a character-by-character basis, C<is_utf8_char>
2641 will tell you whether the current character in a string is valid UTF-8.
2643 =head2 How does UTF-8 represent Unicode characters?
2645 As mentioned above, UTF-8 uses a variable number of bytes to store a
2646 character. Characters with values 0...127 are stored in one byte, just
2647 like good ol' ASCII. Character 128 is stored as C<v194.128>; this
2648 continues up to character 191, which is C<v194.191>. Now we've run out of
2649 bits (191 is binary C<10111111>) so we move on; 192 is C<v195.128>. And
2650 so it goes on, moving to three bytes at character 2048.
2652 Assuming you know you're dealing with a UTF-8 string, you can find out
2653 how long the first character in it is with the C<UTF8SKIP> macro:
2655 char *utf = "\305\233\340\240\201";
2658 len = UTF8SKIP(utf); /* len is 2 here */
2660 len = UTF8SKIP(utf); /* len is 3 here */
2662 Another way to skip over characters in a UTF-8 string is to use
2663 C<utf8_hop>, which takes a string and a number of characters to skip
2664 over. You're on your own about bounds checking, though, so don't use it
2667 All bytes in a multi-byte UTF-8 character will have the high bit set,
2668 so you can test if you need to do something special with this
2669 character like this (the UTF8_IS_INVARIANT() is a macro that tests
2670 whether the byte can be encoded as a single byte even in UTF-8):
2673 U8 *utf_end; /* 1 beyond buffer pointed to by utf */
2674 UV uv; /* Note: a UV, not a U8, not a char */
2675 STRLEN len; /* length of character in bytes */
2677 if (!UTF8_IS_INVARIANT(*utf))
2678 /* Must treat this as UTF-8 */
2679 uv = utf8_to_uvchr_buf(utf, utf_end, &len);
2681 /* OK to treat this character as a byte */
2684 You can also see in that example that we use C<utf8_to_uvchr_buf> to get the
2685 value of the character; the inverse function C<uvchr_to_utf8> is available
2686 for putting a UV into UTF-8:
2688 if (!UTF8_IS_INVARIANT(uv))
2689 /* Must treat this as UTF8 */
2690 utf8 = uvchr_to_utf8(utf8, uv);
2692 /* OK to treat this character as a byte */
2695 You B<must> convert characters to UVs using the above functions if
2696 you're ever in a situation where you have to match UTF-8 and non-UTF-8
2697 characters. You may not skip over UTF-8 characters in this case. If you
2698 do this, you'll lose the ability to match hi-bit non-UTF-8 characters;
2699 for instance, if your UTF-8 string contains C<v196.172>, and you skip
2700 that character, you can never match a C<chr(200)> in a non-UTF-8 string.
2703 =head2 How does Perl store UTF-8 strings?
2705 Currently, Perl deals with Unicode strings and non-Unicode strings
2706 slightly differently. A flag in the SV, C<SVf_UTF8>, indicates that the
2707 string is internally encoded as UTF-8. Without it, the byte value is the
2708 codepoint number and vice versa (in other words, the string is encoded
2709 as iso-8859-1, but C<use feature 'unicode_strings'> is needed to get iso-8859-1
2710 semantics). You can check and manipulate this flag with the
2717 This flag has an important effect on Perl's treatment of the string: if
2718 Unicode data is not properly distinguished, regular expressions,
2719 C<length>, C<substr> and other string handling operations will have
2720 undesirable results.
2722 The problem comes when you have, for instance, a string that isn't
2723 flagged as UTF-8, and contains a byte sequence that could be UTF-8 -
2724 especially when combining non-UTF-8 and UTF-8 strings.
2726 Never forget that the C<SVf_UTF8> flag is separate to the PV value; you
2727 need be sure you don't accidentally knock it off while you're
2728 manipulating SVs. More specifically, you cannot expect to do this:
2737 nsv = newSVpvn(p, len);
2739 The C<char*> string does not tell you the whole story, and you can't
2740 copy or reconstruct an SV just by copying the string value. Check if the
2741 old SV has the UTF8 flag set, and act accordingly:
2745 nsv = newSVpvn(p, len);
2749 In fact, your C<frobnicate> function should be made aware of whether or
2750 not it's dealing with UTF-8 data, so that it can handle the string
2753 Since just passing an SV to an XS function and copying the data of
2754 the SV is not enough to copy the UTF8 flags, even less right is just
2755 passing a C<char *> to an XS function.
2757 =head2 How do I convert a string to UTF-8?
2759 If you're mixing UTF-8 and non-UTF-8 strings, it is necessary to upgrade
2760 one of the strings to UTF-8. If you've got an SV, the easiest way to do
2763 sv_utf8_upgrade(sv);
2765 However, you must not do this, for example:
2768 sv_utf8_upgrade(left);
2770 If you do this in a binary operator, you will actually change one of the
2771 strings that came into the operator, and, while it shouldn't be noticeable
2772 by the end user, it can cause problems in deficient code.
2774 Instead, C<bytes_to_utf8> will give you a UTF-8-encoded B<copy> of its
2775 string argument. This is useful for having the data available for
2776 comparisons and so on, without harming the original SV. There's also
2777 C<utf8_to_bytes> to go the other way, but naturally, this will fail if
2778 the string contains any characters above 255 that can't be represented
2781 =head2 Is there anything else I need to know?
2783 Not really. Just remember these things:
2789 There's no way to tell if a string is UTF-8 or not. You can tell if an SV
2790 is UTF-8 by looking at its C<SvUTF8> flag. Don't forget to set the flag if
2791 something should be UTF-8. Treat the flag as part of the PV, even though
2792 it's not - if you pass on the PV to somewhere, pass on the flag too.
2796 If a string is UTF-8, B<always> use C<utf8_to_uvchr_buf> to get at the value,
2797 unless C<UTF8_IS_INVARIANT(*s)> in which case you can use C<*s>.
2801 When writing a character C<uv> to a UTF-8 string, B<always> use
2802 C<uvchr_to_utf8>, unless C<UTF8_IS_INVARIANT(uv))> in which case
2803 you can use C<*s = uv>.
2807 Mixing UTF-8 and non-UTF-8 strings is tricky. Use C<bytes_to_utf8> to get
2808 a new string which is UTF-8 encoded, and then combine them.
2812 =head1 Custom Operators
2814 Custom operator support is a new experimental feature that allows you to
2815 define your own ops. This is primarily to allow the building of
2816 interpreters for other languages in the Perl core, but it also allows
2817 optimizations through the creation of "macro-ops" (ops which perform the
2818 functions of multiple ops which are usually executed together, such as
2819 C<gvsv, gvsv, add>.)
2821 This feature is implemented as a new op type, C<OP_CUSTOM>. The Perl
2822 core does not "know" anything special about this op type, and so it will
2823 not be involved in any optimizations. This also means that you can
2824 define your custom ops to be any op structure - unary, binary, list and
2827 It's important to know what custom operators won't do for you. They
2828 won't let you add new syntax to Perl, directly. They won't even let you
2829 add new keywords, directly. In fact, they won't change the way Perl
2830 compiles a program at all. You have to do those changes yourself, after
2831 Perl has compiled the program. You do this either by manipulating the op
2832 tree using a C<CHECK> block and the C<B::Generate> module, or by adding
2833 a custom peephole optimizer with the C<optimize> module.
2835 When you do this, you replace ordinary Perl ops with custom ops by
2836 creating ops with the type C<OP_CUSTOM> and the C<pp_addr> of your own
2837 PP function. This should be defined in XS code, and should look like
2838 the PP ops in C<pp_*.c>. You are responsible for ensuring that your op
2839 takes the appropriate number of values from the stack, and you are
2840 responsible for adding stack marks if necessary.
2842 You should also "register" your op with the Perl interpreter so that it
2843 can produce sensible error and warning messages. Since it is possible to
2844 have multiple custom ops within the one "logical" op type C<OP_CUSTOM>,
2845 Perl uses the value of C<< o->op_ppaddr >> to determine which custom op
2846 it is dealing with. You should create an C<XOP> structure for each
2847 ppaddr you use, set the properties of the custom op with
2848 C<XopENTRY_set>, and register the structure against the ppaddr using
2849 C<Perl_custom_op_register>. A trivial example might look like:
2852 static OP *my_pp(pTHX);
2855 XopENTRY_set(&my_xop, xop_name, "myxop");
2856 XopENTRY_set(&my_xop, xop_desc, "Useless custom op");
2857 Perl_custom_op_register(aTHX_ my_pp, &my_xop);
2859 The available fields in the structure are:
2865 A short name for your op. This will be included in some error messages,
2866 and will also be returned as C<< $op->name >> by the L<B|B> module, so
2867 it will appear in the output of module like L<B::Concise|B::Concise>.
2871 A short description of the function of the op.
2875 Which of the various C<*OP> structures this op uses. This should be one of
2876 the C<OA_*> constants from F<op.h>, namely
2896 =item OA_PVOP_OR_SVOP
2898 This should be interpreted as 'C<PVOP>' only. The C<_OR_SVOP> is because
2899 the only core C<PVOP>, C<OP_TRANS>, can sometimes be a C<SVOP> instead.
2907 The other C<OA_*> constants should not be used.
2911 This member is of type C<Perl_cpeep_t>, which expands to C<void
2912 (*Perl_cpeep_t)(aTHX_ OP *o, OP *oldop)>. If it is set, this function
2913 will be called from C<Perl_rpeep> when ops of this type are encountered
2914 by the peephole optimizer. I<o> is the OP that needs optimizing;
2915 I<oldop> is the previous OP optimized, whose C<op_next> points to I<o>.
2919 C<B::Generate> directly supports the creation of custom ops by name.
2923 Until May 1997, this document was maintained by Jeff Okamoto
2924 E<lt>okamoto@corp.hp.comE<gt>. It is now maintained as part of Perl
2925 itself by the Perl 5 Porters E<lt>perl5-porters@perl.orgE<gt>.
2927 With lots of help and suggestions from Dean Roehrich, Malcolm Beattie,
2928 Andreas Koenig, Paul Hudson, Ilya Zakharevich, Paul Marquess, Neil
2929 Bowers, Matthew Green, Tim Bunce, Spider Boardman, Ulrich Pfeifer,
2930 Stephen McCamant, and Gurusamy Sarathy.
2934 L<perlapi>, L<perlintern>, L<perlxs>, L<perlembed>