3 perlcall - Perl calling conventions from C
7 The purpose of this document is to show you how to call Perl subroutines
8 directly from C, i.e., how to write I<callbacks>.
10 Apart from discussing the C interface provided by Perl for writing
11 callbacks the document uses a series of examples to show how the
12 interface actually works in practice. In addition some techniques for
13 coding callbacks are covered.
15 Examples where callbacks are necessary include
19 =item * An Error Handler
21 You have created an XSUB interface to an application's C API.
23 A fairly common feature in applications is to allow you to define a C
24 function that will be called whenever something nasty occurs. What we
25 would like is to be able to specify a Perl subroutine that will be
28 =item * An Event Driven Program
30 The classic example of where callbacks are used is when writing an
31 event driven program like for an X windows application. In this case
32 you register functions to be called whenever specific events occur,
33 e.g., a mouse button is pressed, the cursor moves into a window or a
34 menu item is selected.
38 Although the techniques described here are applicable when embedding
39 Perl in a C program, this is not the primary goal of this document.
40 There are other details that must be considered and are specific to
41 embedding Perl. For details on embedding Perl in C refer to
44 Before you launch yourself head first into the rest of this document,
45 it would be a good idea to have read the following two documents -
46 L<perlxs> and L<perlguts>.
48 =head1 THE CALL_ FUNCTIONS
50 Although this stuff is easier to explain using examples, you first need
51 be aware of a few important definitions.
53 Perl has a number of C functions that allow you to call Perl
56 I32 call_sv(SV* sv, I32 flags);
57 I32 call_pv(char *subname, I32 flags);
58 I32 call_method(char *methname, I32 flags);
59 I32 call_argv(char *subname, I32 flags, register char **argv);
61 The key function is I<call_sv>. All the other functions are
62 fairly simple wrappers which make it easier to call Perl subroutines in
63 special cases. At the end of the day they will all call I<call_sv>
64 to invoke the Perl subroutine.
66 All the I<call_*> functions have a C<flags> parameter which is
67 used to pass a bit mask of options to Perl. This bit mask operates
68 identically for each of the functions. The settings available in the
69 bit mask are discussed in L<FLAG VALUES>.
71 Each of the functions will now be discussed in turn.
77 I<call_sv> takes two parameters, the first, C<sv>, is an SV*.
78 This allows you to specify the Perl subroutine to be called either as a
79 C string (which has first been converted to an SV) or a reference to a
80 subroutine. The section, I<Using call_sv>, shows how you can make
85 The function, I<call_pv>, is similar to I<call_sv> except it
86 expects its first parameter to be a C char* which identifies the Perl
87 subroutine you want to call, e.g., C<call_pv("fred", 0)>. If the
88 subroutine you want to call is in another package, just include the
89 package name in the string, e.g., C<"pkg::fred">.
93 The function I<call_method> is used to call a method from a Perl
94 class. The parameter C<methname> corresponds to the name of the method
95 to be called. Note that the class that the method belongs to is passed
96 on the Perl stack rather than in the parameter list. This class can be
97 either the name of the class (for a static method) or a reference to an
98 object (for a virtual method). See L<perlobj> for more information on
99 static and virtual methods and L<Using call_method> for an example
100 of using I<call_method>.
104 I<call_argv> calls the Perl subroutine specified by the C string
105 stored in the C<subname> parameter. It also takes the usual C<flags>
106 parameter. The final parameter, C<argv>, consists of a NULL terminated
107 list of C strings to be passed as parameters to the Perl subroutine.
108 See I<Using call_argv>.
112 All the functions return an integer. This is a count of the number of
113 items returned by the Perl subroutine. The actual items returned by the
114 subroutine are stored on the Perl stack.
116 As a general rule you should I<always> check the return value from
117 these functions. Even if you are expecting only a particular number of
118 values to be returned from the Perl subroutine, there is nothing to
119 stop someone from doing something unexpected--don't say you haven't
124 The C<flags> parameter in all the I<call_*> functions is a bit mask
125 which can consist of any combination of the symbols defined below,
131 Calls the Perl subroutine in a void context.
133 This flag has 2 effects:
139 It indicates to the subroutine being called that it is executing in
140 a void context (if it executes I<wantarray> the result will be the
145 It ensures that nothing is actually returned from the subroutine.
149 The value returned by the I<call_*> function indicates how many
150 items have been returned by the Perl subroutine - in this case it will
156 Calls the Perl subroutine in a scalar context. This is the default
157 context flag setting for all the I<call_*> functions.
159 This flag has 2 effects:
165 It indicates to the subroutine being called that it is executing in a
166 scalar context (if it executes I<wantarray> the result will be false).
170 It ensures that only a scalar is actually returned from the subroutine.
171 The subroutine can, of course, ignore the I<wantarray> and return a
172 list anyway. If so, then only the last element of the list will be
177 The value returned by the I<call_*> function indicates how many
178 items have been returned by the Perl subroutine - in this case it will
181 If 0, then you have specified the G_DISCARD flag.
183 If 1, then the item actually returned by the Perl subroutine will be
184 stored on the Perl stack - the section I<Returning a Scalar> shows how
185 to access this value on the stack. Remember that regardless of how
186 many items the Perl subroutine returns, only the last one will be
187 accessible from the stack - think of the case where only one value is
188 returned as being a list with only one element. Any other items that
189 were returned will not exist by the time control returns from the
190 I<call_*> function. The section I<Returning a list in a scalar
191 context> shows an example of this behavior.
196 Calls the Perl subroutine in a list context.
198 As with G_SCALAR, this flag has 2 effects:
204 It indicates to the subroutine being called that it is executing in a
205 list context (if it executes I<wantarray> the result will be true).
210 It ensures that all items returned from the subroutine will be
211 accessible when control returns from the I<call_*> function.
215 The value returned by the I<call_*> function indicates how many
216 items have been returned by the Perl subroutine.
218 If 0, then you have specified the G_DISCARD flag.
220 If not 0, then it will be a count of the number of items returned by
221 the subroutine. These items will be stored on the Perl stack. The
222 section I<Returning a list of values> gives an example of using the
223 G_ARRAY flag and the mechanics of accessing the returned items from the
228 By default, the I<call_*> functions place the items returned from
229 by the Perl subroutine on the stack. If you are not interested in
230 these items, then setting this flag will make Perl get rid of them
231 automatically for you. Note that it is still possible to indicate a
232 context to the Perl subroutine by using either G_SCALAR or G_ARRAY.
234 If you do not set this flag then it is I<very> important that you make
235 sure that any temporaries (i.e., parameters passed to the Perl
236 subroutine and values returned from the subroutine) are disposed of
237 yourself. The section I<Returning a Scalar> gives details of how to
238 dispose of these temporaries explicitly and the section I<Using Perl to
239 dispose of temporaries> discusses the specific circumstances where you
240 can ignore the problem and let Perl deal with it for you.
244 Whenever a Perl subroutine is called using one of the I<call_*>
245 functions, it is assumed by default that parameters are to be passed to
246 the subroutine. If you are not passing any parameters to the Perl
247 subroutine, you can save a bit of time by setting this flag. It has
248 the effect of not creating the C<@_> array for the Perl subroutine.
250 Although the functionality provided by this flag may seem
251 straightforward, it should be used only if there is a good reason to do
252 so. The reason for being cautious is that even if you have specified
253 the G_NOARGS flag, it is still possible for the Perl subroutine that
254 has been called to think that you have passed it parameters.
256 In fact, what can happen is that the Perl subroutine you have called
257 can access the C<@_> array from a previous Perl subroutine. This will
258 occur when the code that is executing the I<call_*> function has
259 itself been called from another Perl subroutine. The code below
274 What has happened is that C<fred> accesses the C<@_> array which
280 It is possible for the Perl subroutine you are calling to terminate
281 abnormally, e.g., by calling I<die> explicitly or by not actually
282 existing. By default, when either of these events occurs, the
283 process will terminate immediately. If you want to trap this
284 type of event, specify the G_EVAL flag. It will put an I<eval { }>
285 around the subroutine call.
287 Whenever control returns from the I<call_*> function you need to
288 check the C<$@> variable as you would in a normal Perl script.
290 The value returned from the I<call_*> function is dependent on
291 what other flags have been specified and whether an error has
292 occurred. Here are all the different cases that can occur:
298 If the I<call_*> function returns normally, then the value
299 returned is as specified in the previous sections.
303 If G_DISCARD is specified, the return value will always be 0.
307 If G_ARRAY is specified I<and> an error has occurred, the return value
312 If G_SCALAR is specified I<and> an error has occurred, the return value
313 will be 1 and the value on the top of the stack will be I<undef>. This
314 means that if you have already detected the error by checking C<$@> and
315 you want the program to continue, you must remember to pop the I<undef>
320 See I<Using G_EVAL> for details on using G_EVAL.
324 Using the G_EVAL flag described above will always set C<$@>: clearing
325 it if there was no error, and setting it to describe the error if there
326 was an error in the called code. This is what you want if your intention
327 is to handle possible errors, but sometimes you just want to trap errors
328 and stop them interfering with the rest of the program.
330 This scenario will mostly be applicable to code that is meant to be called
331 from within destructors, asynchronous callbacks, and signal handlers.
332 In such situations, where the code being called has little relation to the
333 surrounding dynamic context, the main program needs to be insulated from
334 errors in the called code, even if they can't be handled intelligently.
335 It may also be useful to do this with code for C<__DIE__> or C<__WARN__>
336 hooks, and C<tie> functions.
338 The G_KEEPERR flag is meant to be used in conjunction with G_EVAL in
339 I<call_*> functions that are used to implement such code. This flag
340 has no effect when G_EVAL is not used.
342 When G_KEEPERR is used, any error in the called code will terminate the
343 call as usual, and the error will not propagate beyond the call (as usual
344 for G_EVAL), but it will not go into C<$@>. Instead the error will be
345 converted into a warning, prefixed with the string "\t(in cleanup)".
346 This can be disabled using C<no warnings 'misc'>. If there is no error,
347 C<$@> will not be cleared.
349 The G_KEEPERR flag was introduced in Perl version 5.002.
351 See I<Using G_KEEPERR> for an example of a situation that warrants the
354 =head2 Determining the Context
356 As mentioned above, you can determine the context of the currently
357 executing subroutine in Perl with I<wantarray>. The equivalent test
358 can be made in C by using the C<GIMME_V> macro, which returns
359 C<G_ARRAY> if you have been called in a list context, C<G_SCALAR> if
360 in a scalar context, or C<G_VOID> if in a void context (i.e. the
361 return value will not be used). An older version of this macro is
362 called C<GIMME>; in a void context it returns C<G_SCALAR> instead of
363 C<G_VOID>. An example of using the C<GIMME_V> macro is shown in
364 section I<Using GIMME_V>.
368 Enough of the definition talk, let's have a few examples.
370 Perl provides many macros to assist in accessing the Perl stack.
371 Wherever possible, these macros should always be used when interfacing
372 to Perl internals. We hope this should make the code less vulnerable
373 to any changes made to Perl in the future.
375 Another point worth noting is that in the first series of examples I
376 have made use of only the I<call_pv> function. This has been done
377 to keep the code simpler and ease you into the topic. Wherever
378 possible, if the choice is between using I<call_pv> and
379 I<call_sv>, you should always try to use I<call_sv>. See
380 I<Using call_sv> for details.
382 =head2 No Parameters, Nothing returned
384 This first trivial example will call a Perl subroutine, I<PrintUID>, to
385 print out the UID of the process.
392 and here is a C function to call it
400 call_pv("PrintUID", G_DISCARD|G_NOARGS);
405 A few points to note about this example.
411 Ignore C<dSP> and C<PUSHMARK(SP)> for now. They will be discussed in
416 We aren't passing any parameters to I<PrintUID> so G_NOARGS can be
421 We aren't interested in anything returned from I<PrintUID>, so
422 G_DISCARD is specified. Even if I<PrintUID> was changed to
423 return some value(s), having specified G_DISCARD will mean that they
424 will be wiped by the time control returns from I<call_pv>.
428 As I<call_pv> is being used, the Perl subroutine is specified as a
429 C string. In this case the subroutine name has been 'hard-wired' into the
434 Because we specified G_DISCARD, it is not necessary to check the value
435 returned from I<call_pv>. It will always be 0.
439 =head2 Passing Parameters
441 Now let's make a slightly more complex example. This time we want to
442 call a Perl subroutine, C<LeftString>, which will take 2 parameters--a
443 string ($s) and an integer ($n). The subroutine will simply
444 print the first $n characters of the string.
446 So the Perl subroutine would look like this
451 print substr($s, 0, $n), "\n";
454 The C function required to call I<LeftString> would look like this.
457 call_LeftString(a, b)
467 XPUSHs(sv_2mortal(newSVpv(a, 0)));
468 XPUSHs(sv_2mortal(newSViv(b)));
471 call_pv("LeftString", G_DISCARD);
477 Here are a few notes on the C function I<call_LeftString>.
483 Parameters are passed to the Perl subroutine using the Perl stack.
484 This is the purpose of the code beginning with the line C<dSP> and
485 ending with the line C<PUTBACK>. The C<dSP> declares a local copy
486 of the stack pointer. This local copy should B<always> be accessed
491 If you are going to put something onto the Perl stack, you need to know
492 where to put it. This is the purpose of the macro C<dSP>--it declares
493 and initializes a I<local> copy of the Perl stack pointer.
495 All the other macros which will be used in this example require you to
496 have used this macro.
498 The exception to this rule is if you are calling a Perl subroutine
499 directly from an XSUB function. In this case it is not necessary to
500 use the C<dSP> macro explicitly--it will be declared for you
505 Any parameters to be pushed onto the stack should be bracketed by the
506 C<PUSHMARK> and C<PUTBACK> macros. The purpose of these two macros, in
507 this context, is to count the number of parameters you are
508 pushing automatically. Then whenever Perl is creating the C<@_> array for the
509 subroutine, it knows how big to make it.
511 The C<PUSHMARK> macro tells Perl to make a mental note of the current
512 stack pointer. Even if you aren't passing any parameters (like the
513 example shown in the section I<No Parameters, Nothing returned>) you
514 must still call the C<PUSHMARK> macro before you can call any of the
515 I<call_*> functions--Perl still needs to know that there are no
518 The C<PUTBACK> macro sets the global copy of the stack pointer to be
519 the same as our local copy. If we didn't do this I<call_pv>
520 wouldn't know where the two parameters we pushed were--remember that
521 up to now all the stack pointer manipulation we have done is with our
522 local copy, I<not> the global copy.
526 Next, we come to XPUSHs. This is where the parameters actually get
527 pushed onto the stack. In this case we are pushing a string and an
530 See L<perlguts/"XSUBs and the Argument Stack"> for details
531 on how the XPUSH macros work.
535 Because we created temporary values (by means of sv_2mortal() calls)
536 we will have to tidy up the Perl stack and dispose of mortal SVs.
538 This is the purpose of
543 at the start of the function, and
548 at the end. The C<ENTER>/C<SAVETMPS> pair creates a boundary for any
549 temporaries we create. This means that the temporaries we get rid of
550 will be limited to those which were created after these calls.
552 The C<FREETMPS>/C<LEAVE> pair will get rid of any values returned by
553 the Perl subroutine (see next example), plus it will also dump the
554 mortal SVs we have created. Having C<ENTER>/C<SAVETMPS> at the
555 beginning of the code makes sure that no other mortals are destroyed.
557 Think of these macros as working a bit like using C<{> and C<}> in Perl
558 to limit the scope of local variables.
560 See the section I<Using Perl to dispose of temporaries> for details of
561 an alternative to using these macros.
565 Finally, I<LeftString> can now be called via the I<call_pv> function.
566 The only flag specified this time is G_DISCARD. Because we are passing
567 2 parameters to the Perl subroutine this time, we have not specified
572 =head2 Returning a Scalar
574 Now for an example of dealing with the items returned from a Perl
577 Here is a Perl subroutine, I<Adder>, that takes 2 integer parameters
578 and simply returns their sum.
586 Because we are now concerned with the return value from I<Adder>, the C
587 function required to call it is now a bit more complex.
601 XPUSHs(sv_2mortal(newSViv(a)));
602 XPUSHs(sv_2mortal(newSViv(b)));
605 count = call_pv("Adder", G_SCALAR);
610 croak("Big trouble\n");
612 printf ("The sum of %d and %d is %d\n", a, b, POPi);
619 Points to note this time are
625 The only flag specified this time was G_SCALAR. That means the C<@_>
626 array will be created and that the value returned by I<Adder> will
627 still exist after the call to I<call_pv>.
631 The purpose of the macro C<SPAGAIN> is to refresh the local copy of the
632 stack pointer. This is necessary because it is possible that the memory
633 allocated to the Perl stack has been reallocated whilst in the
636 If you are making use of the Perl stack pointer in your code you must
637 always refresh the local copy using SPAGAIN whenever you make use
638 of the I<call_*> functions or any other Perl internal function.
642 Although only a single value was expected to be returned from I<Adder>,
643 it is still good practice to check the return code from I<call_pv>
646 Expecting a single value is not quite the same as knowing that there
647 will be one. If someone modified I<Adder> to return a list and we
648 didn't check for that possibility and take appropriate action the Perl
649 stack would end up in an inconsistent state. That is something you
650 I<really> don't want to happen ever.
654 The C<POPi> macro is used here to pop the return value from the stack.
655 In this case we wanted an integer, so C<POPi> was used.
658 Here is the complete list of POP macros available, along with the types
669 The final C<PUTBACK> is used to leave the Perl stack in a consistent
670 state before exiting the function. This is necessary because when we
671 popped the return value from the stack with C<POPi> it updated only our
672 local copy of the stack pointer. Remember, C<PUTBACK> sets the global
673 stack pointer to be the same as our local copy.
678 =head2 Returning a list of values
680 Now, let's extend the previous example to return both the sum of the
681 parameters and the difference.
683 Here is the Perl subroutine
691 and this is the C function
694 call_AddSubtract(a, b)
705 XPUSHs(sv_2mortal(newSViv(a)));
706 XPUSHs(sv_2mortal(newSViv(b)));
709 count = call_pv("AddSubtract", G_ARRAY);
714 croak("Big trouble\n");
716 printf ("%d - %d = %d\n", a, b, POPi);
717 printf ("%d + %d = %d\n", a, b, POPi);
724 If I<call_AddSubtract> is called like this
726 call_AddSubtract(7, 4);
728 then here is the output
739 We wanted list context, so G_ARRAY was used.
743 Not surprisingly C<POPi> is used twice this time because we were
744 retrieving 2 values from the stack. The important thing to note is that
745 when using the C<POP*> macros they come off the stack in I<reverse>
750 =head2 Returning a list in a scalar context
752 Say the Perl subroutine in the previous section was called in a scalar
756 call_AddSubScalar(a, b)
768 XPUSHs(sv_2mortal(newSViv(a)));
769 XPUSHs(sv_2mortal(newSViv(b)));
772 count = call_pv("AddSubtract", G_SCALAR);
776 printf ("Items Returned = %d\n", count);
778 for (i = 1; i <= count; ++i)
779 printf ("Value %d = %d\n", i, POPi);
786 The other modification made is that I<call_AddSubScalar> will print the
787 number of items returned from the Perl subroutine and their value (for
788 simplicity it assumes that they are integer). So if
789 I<call_AddSubScalar> is called
791 call_AddSubScalar(7, 4);
793 then the output will be
798 In this case the main point to note is that only the last item in the
799 list is returned from the subroutine, I<AddSubtract> actually made it back to
800 I<call_AddSubScalar>.
803 =head2 Returning Data from Perl via the parameter list
805 It is also possible to return values directly via the parameter list -
806 whether it is actually desirable to do it is another matter entirely.
808 The Perl subroutine, I<Inc>, below takes 2 parameters and increments
817 and here is a C function to call it.
832 sva = sv_2mortal(newSViv(a));
833 svb = sv_2mortal(newSViv(b));
840 count = call_pv("Inc", G_DISCARD);
843 croak ("call_Inc: expected 0 values from 'Inc', got %d\n",
846 printf ("%d + 1 = %d\n", a, SvIV(sva));
847 printf ("%d + 1 = %d\n", b, SvIV(svb));
853 To be able to access the two parameters that were pushed onto the stack
854 after they return from I<call_pv> it is necessary to make a note
855 of their addresses--thus the two variables C<sva> and C<svb>.
857 The reason this is necessary is that the area of the Perl stack which
858 held them will very likely have been overwritten by something else by
859 the time control returns from I<call_pv>.
866 Now an example using G_EVAL. Below is a Perl subroutine which computes
867 the difference of its 2 parameters. If this would result in a negative
868 result, the subroutine calls I<die>.
874 die "death can be fatal\n" if $a < $b;
879 and some C to call it
893 XPUSHs(sv_2mortal(newSViv(a)));
894 XPUSHs(sv_2mortal(newSViv(b)));
897 count = call_pv("Subtract", G_EVAL|G_SCALAR);
901 /* Check the eval first */
904 printf ("Uh oh - %s\n", SvPV_nolen(ERRSV));
910 croak("call_Subtract: wanted 1 value from 'Subtract', got %d\n",
913 printf ("%d - %d = %d\n", a, b, POPi);
921 If I<call_Subtract> is called thus
925 the following will be printed
927 Uh oh - death can be fatal
935 We want to be able to catch the I<die> so we have used the G_EVAL
936 flag. Not specifying this flag would mean that the program would
937 terminate immediately at the I<die> statement in the subroutine
946 printf ("Uh oh - %s\n", SvPV_nolen(ERRSV));
950 is the direct equivalent of this bit of Perl
952 print "Uh oh - $@\n" if $@;
954 C<PL_errgv> is a perl global of type C<GV *> that points to the
955 symbol table entry containing the error. C<ERRSV> therefore
956 refers to the C equivalent of C<$@>.
960 Note that the stack is popped using C<POPs> in the block where
961 C<SvTRUE(ERRSV)> is true. This is necessary because whenever a
962 I<call_*> function invoked with G_EVAL|G_SCALAR returns an error,
963 the top of the stack holds the value I<undef>. Because we want the
964 program to continue after detecting this error, it is essential that
965 the stack is tidied up by removing the I<undef>.
970 =head2 Using G_KEEPERR
972 Consider this rather facetious example, where we have used an XS
973 version of the call_Subtract example above inside a destructor:
976 sub new { bless {}, $_[0] }
979 die "death can be fatal" if $a < $b;
982 sub DESTROY { call_Subtract(5, 4); }
983 sub foo { die "foo dies"; }
990 print "Saw: $@" if $@; # should be, but isn't
992 This example will fail to recognize that an error occurred inside the
993 C<eval {}>. Here's why: the call_Subtract code got executed while perl
994 was cleaning up temporaries when exiting the outer braced block, and because
995 call_Subtract is implemented with I<call_pv> using the G_EVAL
996 flag, it promptly reset C<$@>. This results in the failure of the
997 outermost test for C<$@>, and thereby the failure of the error trap.
999 Appending the G_KEEPERR flag, so that the I<call_pv> call in
1000 call_Subtract reads:
1002 count = call_pv("Subtract", G_EVAL|G_SCALAR|G_KEEPERR);
1004 will preserve the error and restore reliable error handling.
1006 =head2 Using call_sv
1008 In all the previous examples I have 'hard-wired' the name of the Perl
1009 subroutine to be called from C. Most of the time though, it is more
1010 convenient to be able to specify the name of the Perl subroutine from
1011 within the Perl script.
1013 Consider the Perl code below
1017 print "Hello there\n";
1022 Here is a snippet of XSUB which defines I<CallSubPV>.
1029 call_pv(name, G_DISCARD|G_NOARGS);
1031 That is fine as far as it goes. The thing is, the Perl subroutine
1032 can be specified as only a string. For Perl 4 this was adequate,
1033 but Perl 5 allows references to subroutines and anonymous subroutines.
1034 This is where I<call_sv> is useful.
1036 The code below for I<CallSubSV> is identical to I<CallSubPV> except
1037 that the C<name> parameter is now defined as an SV* and we use
1038 I<call_sv> instead of I<call_pv>.
1045 call_sv(name, G_DISCARD|G_NOARGS);
1047 Because we are using an SV to call I<fred> the following can all be used
1053 CallSubSV( sub { print "Hello there\n" } );
1055 As you can see, I<call_sv> gives you much greater flexibility in
1056 how you can specify the Perl subroutine.
1058 You should note that if it is necessary to store the SV (C<name> in the
1059 example above) which corresponds to the Perl subroutine so that it can
1060 be used later in the program, it not enough just to store a copy of the
1061 pointer to the SV. Say the code above had been like this
1063 static SV * rememberSub;
1075 call_sv(rememberSub, G_DISCARD|G_NOARGS);
1077 The reason this is wrong is that by the time you come to use the
1078 pointer C<rememberSub> in C<CallSavedSub1>, it may or may not still refer
1079 to the Perl subroutine that was recorded in C<SaveSub1>. This is
1080 particularly true for these cases
1085 SaveSub1( sub { print "Hello there\n" } );
1088 By the time each of the C<SaveSub1> statements above have been executed,
1089 the SV*s which corresponded to the parameters will no longer exist.
1090 Expect an error message from Perl of the form
1092 Can't use an undefined value as a subroutine reference at ...
1094 for each of the C<CallSavedSub1> lines.
1096 Similarly, with this code
1103 you can expect one of these messages (which you actually get is dependent on
1104 the version of Perl you are using)
1106 Not a CODE reference at ...
1107 Undefined subroutine &main::47 called ...
1109 The variable $ref may have referred to the subroutine C<fred>
1110 whenever the call to C<SaveSub1> was made but by the time
1111 C<CallSavedSub1> gets called it now holds the number C<47>. Because we
1112 saved only a pointer to the original SV in C<SaveSub1>, any changes to
1113 $ref will be tracked by the pointer C<rememberSub>. This means that
1114 whenever C<CallSavedSub1> gets called, it will attempt to execute the
1115 code which is referenced by the SV* C<rememberSub>. In this case
1116 though, it now refers to the integer C<47>, so expect Perl to complain
1119 A similar but more subtle problem is illustrated with this code
1126 This time whenever C<CallSavedSub1> get called it will execute the Perl
1127 subroutine C<joe> (assuming it exists) rather than C<fred> as was
1128 originally requested in the call to C<SaveSub1>.
1130 To get around these problems it is necessary to take a full copy of the
1131 SV. The code below shows C<SaveSub2> modified to do that
1133 static SV * keepSub = (SV*)NULL;
1139 /* Take a copy of the callback */
1140 if (keepSub == (SV*)NULL)
1141 /* First time, so create a new SV */
1142 keepSub = newSVsv(name);
1144 /* Been here before, so overwrite */
1145 SvSetSV(keepSub, name);
1151 call_sv(keepSub, G_DISCARD|G_NOARGS);
1153 To avoid creating a new SV every time C<SaveSub2> is called,
1154 the function first checks to see if it has been called before. If not,
1155 then space for a new SV is allocated and the reference to the Perl
1156 subroutine, C<name> is copied to the variable C<keepSub> in one
1157 operation using C<newSVsv>. Thereafter, whenever C<SaveSub2> is called
1158 the existing SV, C<keepSub>, is overwritten with the new value using
1161 =head2 Using call_argv
1163 Here is a Perl subroutine which prints whatever parameters are passed
1170 foreach (@list) { print "$_\n" }
1173 and here is an example of I<call_argv> which will call
1176 static char * words[] = {"alpha", "beta", "gamma", "delta", NULL};
1183 call_argv("PrintList", G_DISCARD, words);
1186 Note that it is not necessary to call C<PUSHMARK> in this instance.
1187 This is because I<call_argv> will do it for you.
1189 =head2 Using call_method
1191 Consider the following Perl code
1204 my ($self, $index) = @_;
1205 print "$index: $$self[$index]\n";
1211 print "This is Class $class version 1.0\n";
1215 It implements just a very simple class to manage an array. Apart from
1216 the constructor, C<new>, it declares methods, one static and one
1217 virtual. The static method, C<PrintID>, prints out simply the class
1218 name and a version number. The virtual method, C<Display>, prints out a
1219 single element of the array. Here is an all Perl example of using it.
1221 $a = Mine->new('red', 'green', 'blue');
1228 This is Class Mine version 1.0
1230 Calling a Perl method from C is fairly straightforward. The following
1237 a reference to the object for a virtual method or the name of the class
1238 for a static method.
1242 the name of the method.
1246 any other parameters specific to the method.
1250 Here is a simple XSUB which illustrates the mechanics of calling both
1251 the C<PrintID> and C<Display> methods from C.
1254 call_Method(ref, method, index)
1261 XPUSHs(sv_2mortal(newSViv(index)));
1264 call_method(method, G_DISCARD);
1267 call_PrintID(class, method)
1272 XPUSHs(sv_2mortal(newSVpv(class, 0)));
1275 call_method(method, G_DISCARD);
1278 So the methods C<PrintID> and C<Display> can be invoked like this
1280 $a = Mine->new('red', 'green', 'blue');
1281 call_Method($a, 'Display', 1);
1282 call_PrintID('Mine', 'PrintID');
1284 The only thing to note is that in both the static and virtual methods,
1285 the method name is not passed via the stack--it is used as the first
1286 parameter to I<call_method>.
1288 =head2 Using GIMME_V
1290 Here is a trivial XSUB which prints the context in which it is
1291 currently executing.
1296 I32 gimme = GIMME_V;
1297 if (gimme == G_VOID)
1298 printf ("Context is Void\n");
1299 else if (gimme == G_SCALAR)
1300 printf ("Context is Scalar\n");
1302 printf ("Context is Array\n");
1304 and here is some Perl to test it
1310 The output from that will be
1316 =head2 Using Perl to dispose of temporaries
1318 In the examples given to date, any temporaries created in the callback
1319 (i.e., parameters passed on the stack to the I<call_*> function or
1320 values returned via the stack) have been freed by one of these methods
1326 specifying the G_DISCARD flag with I<call_*>.
1330 explicitly disposed of using the C<ENTER>/C<SAVETMPS> -
1331 C<FREETMPS>/C<LEAVE> pairing.
1335 There is another method which can be used, namely letting Perl do it
1336 for you automatically whenever it regains control after the callback
1337 has terminated. This is done by simply not using the
1345 sequence in the callback (and not, of course, specifying the G_DISCARD
1348 If you are going to use this method you have to be aware of a possible
1349 memory leak which can arise under very specific circumstances. To
1350 explain these circumstances you need to know a bit about the flow of
1351 control between Perl and the callback routine.
1353 The examples given at the start of the document (an error handler and
1354 an event driven program) are typical of the two main sorts of flow
1355 control that you are likely to encounter with callbacks. There is a
1356 very important distinction between them, so pay attention.
1358 In the first example, an error handler, the flow of control could be as
1359 follows. You have created an interface to an external library.
1360 Control can reach the external library like this
1362 perl --> XSUB --> external library
1364 Whilst control is in the library, an error condition occurs. You have
1365 previously set up a Perl callback to handle this situation, so it will
1366 get executed. Once the callback has finished, control will drop back to
1367 Perl again. Here is what the flow of control will be like in that
1370 perl --> XSUB --> external library
1374 external library --> call_* --> perl
1376 perl <-- XSUB <-- external library <-- call_* <----+
1378 After processing of the error using I<call_*> is completed,
1379 control reverts back to Perl more or less immediately.
1381 In the diagram, the further right you go the more deeply nested the
1382 scope is. It is only when control is back with perl on the extreme
1383 left of the diagram that you will have dropped back to the enclosing
1384 scope and any temporaries you have left hanging around will be freed.
1386 In the second example, an event driven program, the flow of control
1387 will be more like this
1389 perl --> XSUB --> event handler
1391 event handler --> call_* --> perl
1393 event handler <-- call_* <----+
1395 event handler --> call_* --> perl
1397 event handler <-- call_* <----+
1399 event handler --> call_* --> perl
1401 event handler <-- call_* <----+
1403 In this case the flow of control can consist of only the repeated
1406 event handler --> call_* --> perl
1408 for practically the complete duration of the program. This means that
1409 control may I<never> drop back to the surrounding scope in Perl at the
1412 So what is the big problem? Well, if you are expecting Perl to tidy up
1413 those temporaries for you, you might be in for a long wait. For Perl
1414 to dispose of your temporaries, control must drop back to the
1415 enclosing scope at some stage. In the event driven scenario that may
1416 never happen. This means that as time goes on, your program will
1417 create more and more temporaries, none of which will ever be freed. As
1418 each of these temporaries consumes some memory your program will
1419 eventually consume all the available memory in your system--kapow!
1421 So here is the bottom line--if you are sure that control will revert
1422 back to the enclosing Perl scope fairly quickly after the end of your
1423 callback, then it isn't absolutely necessary to dispose explicitly of
1424 any temporaries you may have created. Mind you, if you are at all
1425 uncertain about what to do, it doesn't do any harm to tidy up anyway.
1428 =head2 Strategies for storing Callback Context Information
1431 Potentially one of the trickiest problems to overcome when designing a
1432 callback interface can be figuring out how to store the mapping between
1433 the C callback function and the Perl equivalent.
1435 To help understand why this can be a real problem first consider how a
1436 callback is set up in an all C environment. Typically a C API will
1437 provide a function to register a callback. This will expect a pointer
1438 to a function as one of its parameters. Below is a call to a
1439 hypothetical function C<register_fatal> which registers the C function
1440 to get called when a fatal error occurs.
1442 register_fatal(cb1);
1444 The single parameter C<cb1> is a pointer to a function, so you must
1445 have defined C<cb1> in your code, say something like this
1450 printf ("Fatal Error\n");
1454 Now change that to call a Perl subroutine instead
1456 static SV * callback = (SV*)NULL;
1465 /* Call the Perl sub to process the callback */
1466 call_sv(callback, G_DISCARD);
1474 /* Remember the Perl sub */
1475 if (callback == (SV*)NULL)
1476 callback = newSVsv(fn);
1478 SvSetSV(callback, fn);
1480 /* register the callback with the external library */
1481 register_fatal(cb1);
1483 where the Perl equivalent of C<register_fatal> and the callback it
1484 registers, C<pcb1>, might look like this
1486 # Register the sub pcb1
1487 register_fatal(\&pcb1);
1491 die "I'm dying...\n";
1494 The mapping between the C callback and the Perl equivalent is stored in
1495 the global variable C<callback>.
1497 This will be adequate if you ever need to have only one callback
1498 registered at any time. An example could be an error handler like the
1499 code sketched out above. Remember though, repeated calls to
1500 C<register_fatal> will replace the previously registered callback
1501 function with the new one.
1503 Say for example you want to interface to a library which allows asynchronous
1504 file i/o. In this case you may be able to register a callback whenever
1505 a read operation has completed. To be of any use we want to be able to
1506 call separate Perl subroutines for each file that is opened. As it
1507 stands, the error handler example above would not be adequate as it
1508 allows only a single callback to be defined at any time. What we
1509 require is a means of storing the mapping between the opened file and
1510 the Perl subroutine we want to be called for that file.
1512 Say the i/o library has a function C<asynch_read> which associates a C
1513 function C<ProcessRead> with a file handle C<fh>--this assumes that it
1514 has also provided some routine to open the file and so obtain the file
1517 asynch_read(fh, ProcessRead)
1519 This may expect the C I<ProcessRead> function of this form
1522 ProcessRead(fh, buffer)
1529 To provide a Perl interface to this library we need to be able to map
1530 between the C<fh> parameter and the Perl subroutine we want called. A
1531 hash is a convenient mechanism for storing this mapping. The code
1532 below shows a possible implementation
1534 static HV * Mapping = (HV*)NULL;
1537 asynch_read(fh, callback)
1541 /* If the hash doesn't already exist, create it */
1542 if (Mapping == (HV*)NULL)
1545 /* Save the fh -> callback mapping */
1546 hv_store(Mapping, (char*)&fh, sizeof(fh), newSVsv(callback), 0);
1548 /* Register with the C Library */
1549 asynch_read(fh, asynch_read_if);
1551 and C<asynch_read_if> could look like this
1554 asynch_read_if(fh, buffer)
1561 /* Get the callback associated with fh */
1562 sv = hv_fetch(Mapping, (char*)&fh , sizeof(fh), FALSE);
1563 if (sv == (SV**)NULL)
1564 croak("Internal error...\n");
1567 XPUSHs(sv_2mortal(newSViv(fh)));
1568 XPUSHs(sv_2mortal(newSVpv(buffer, 0)));
1571 /* Call the Perl sub */
1572 call_sv(*sv, G_DISCARD);
1575 For completeness, here is C<asynch_close>. This shows how to remove
1576 the entry from the hash C<Mapping>.
1582 /* Remove the entry from the hash */
1583 (void) hv_delete(Mapping, (char*)&fh, sizeof(fh), G_DISCARD);
1585 /* Now call the real asynch_close */
1588 So the Perl interface would look like this
1592 my($handle, $buffer) = @_;
1595 # Register the Perl callback
1596 asynch_read($fh, \&callback1);
1600 The mapping between the C callback and Perl is stored in the global
1601 hash C<Mapping> this time. Using a hash has the distinct advantage that
1602 it allows an unlimited number of callbacks to be registered.
1604 What if the interface provided by the C callback doesn't contain a
1605 parameter which allows the file handle to Perl subroutine mapping? Say
1606 in the asynchronous i/o package, the callback function gets passed only
1607 the C<buffer> parameter like this
1616 Without the file handle there is no straightforward way to map from the
1617 C callback to the Perl subroutine.
1619 In this case a possible way around this problem is to predefine a
1620 series of C functions to act as the interface to Perl, thus
1623 #define NULL_HANDLE -1
1624 typedef void (*FnMap)();
1636 static struct MapStruct Map [MAX_CB] =
1638 { fn1, NULL, NULL_HANDLE },
1639 { fn2, NULL, NULL_HANDLE },
1640 { fn3, NULL, NULL_HANDLE }
1651 XPUSHs(sv_2mortal(newSVpv(buffer, 0)));
1654 /* Call the Perl sub */
1655 call_sv(Map[index].PerlSub, G_DISCARD);
1680 array_asynch_read(fh, callback)
1685 int null_index = MAX_CB;
1687 /* Find the same handle or an empty entry */
1688 for (index = 0; index < MAX_CB; ++index)
1690 if (Map[index].Handle == fh)
1693 if (Map[index].Handle == NULL_HANDLE)
1697 if (index == MAX_CB && null_index == MAX_CB)
1698 croak ("Too many callback functions registered\n");
1700 if (index == MAX_CB)
1703 /* Save the file handle */
1704 Map[index].Handle = fh;
1706 /* Remember the Perl sub */
1707 if (Map[index].PerlSub == (SV*)NULL)
1708 Map[index].PerlSub = newSVsv(callback);
1710 SvSetSV(Map[index].PerlSub, callback);
1712 asynch_read(fh, Map[index].Function);
1715 array_asynch_close(fh)
1720 /* Find the file handle */
1721 for (index = 0; index < MAX_CB; ++ index)
1722 if (Map[index].Handle == fh)
1725 if (index == MAX_CB)
1726 croak ("could not close fh %d\n", fh);
1728 Map[index].Handle = NULL_HANDLE;
1729 SvREFCNT_dec(Map[index].PerlSub);
1730 Map[index].PerlSub = (SV*)NULL;
1734 In this case the functions C<fn1>, C<fn2>, and C<fn3> are used to
1735 remember the Perl subroutine to be called. Each of the functions holds
1736 a separate hard-wired index which is used in the function C<Pcb> to
1737 access the C<Map> array and actually call the Perl subroutine.
1739 There are some obvious disadvantages with this technique.
1741 Firstly, the code is considerably more complex than with the previous
1744 Secondly, there is a hard-wired limit (in this case 3) to the number of
1745 callbacks that can exist simultaneously. The only way to increase the
1746 limit is by modifying the code to add more functions and then
1747 recompiling. None the less, as long as the number of functions is
1748 chosen with some care, it is still a workable solution and in some
1749 cases is the only one available.
1751 To summarize, here are a number of possible methods for you to consider
1752 for storing the mapping between C and the Perl callback
1756 =item 1. Ignore the problem - Allow only 1 callback
1758 For a lot of situations, like interfacing to an error handler, this may
1759 be a perfectly adequate solution.
1761 =item 2. Create a sequence of callbacks - hard wired limit
1763 If it is impossible to tell from the parameters passed back from the C
1764 callback what the context is, then you may need to create a sequence of C
1765 callback interface functions, and store pointers to each in an array.
1767 =item 3. Use a parameter to map to the Perl callback
1769 A hash is an ideal mechanism to store the mapping between C and Perl.
1774 =head2 Alternate Stack Manipulation
1777 Although I have made use of only the C<POP*> macros to access values
1778 returned from Perl subroutines, it is also possible to bypass these
1779 macros and read the stack using the C<ST> macro (See L<perlxs> for a
1780 full description of the C<ST> macro).
1782 Most of the time the C<POP*> macros should be adequate, the main
1783 problem with them is that they force you to process the returned values
1784 in sequence. This may not be the most suitable way to process the
1785 values in some cases. What we want is to be able to access the stack in
1786 a random order. The C<ST> macro as used when coding an XSUB is ideal
1789 The code below is the example given in the section I<Returning a list
1790 of values> recoded to use C<ST> instead of C<POP*>.
1793 call_AddSubtract2(a, b)
1805 XPUSHs(sv_2mortal(newSViv(a)));
1806 XPUSHs(sv_2mortal(newSViv(b)));
1809 count = call_pv("AddSubtract", G_ARRAY);
1813 ax = (SP - PL_stack_base) + 1;
1816 croak("Big trouble\n");
1818 printf ("%d + %d = %d\n", a, b, SvIV(ST(0)));
1819 printf ("%d - %d = %d\n", a, b, SvIV(ST(1)));
1832 Notice that it was necessary to define the variable C<ax>. This is
1833 because the C<ST> macro expects it to exist. If we were in an XSUB it
1834 would not be necessary to define C<ax> as it is already defined for
1843 ax = (SP - PL_stack_base) + 1;
1845 sets the stack up so that we can use the C<ST> macro.
1849 Unlike the original coding of this example, the returned
1850 values are not accessed in reverse order. So C<ST(0)> refers to the
1851 first value returned by the Perl subroutine and C<ST(count-1)>
1856 =head2 Creating and calling an anonymous subroutine in C
1858 As we've already shown, C<call_sv> can be used to invoke an
1859 anonymous subroutine. However, our example showed a Perl script
1860 invoking an XSUB to perform this operation. Let's see how it can be
1861 done inside our C code:
1865 SV *cvrv = eval_pv("sub { print 'You will not find me cluttering any namespace!' }", TRUE);
1869 call_sv(cvrv, G_VOID|G_NOARGS);
1871 C<eval_pv> is used to compile the anonymous subroutine, which
1872 will be the return value as well (read more about C<eval_pv> in
1873 L<perlapi/eval_pv>). Once this code reference is in hand, it
1874 can be mixed in with all the previous examples we've shown.
1876 =head1 LIGHTWEIGHT CALLBACKS
1878 Sometimes you need to invoke the same subroutine repeatedly.
1879 This usually happens with a function that acts on a list of
1880 values, such as Perl's built-in sort(). You can pass a
1881 comparison function to sort(), which will then be invoked
1882 for every pair of values that needs to be compared. The first()
1883 and reduce() functions from L<List::Util> follow a similar
1886 In this case it is possible to speed up the routine (often
1887 quite substantially) by using the lightweight callback API.
1888 The idea is that the calling context only needs to be
1889 created and destroyed once, and the sub can be called
1890 arbitrarily many times in between.
1892 It is usual to pass parameters using global variables (typically
1893 $_ for one parameter, or $a and $b for two parameters) rather
1894 than via @_. (It is possible to use the @_ mechanism if you know
1895 what you're doing, though there is as yet no supported API for
1896 it. It's also inherently slower.)
1898 The pattern of macro calls is like this:
1900 dMULTICALL; /* Declare local variables */
1901 I32 gimme = G_SCALAR; /* context of the call: G_SCALAR,
1902 * G_LIST, or G_VOID */
1904 PUSH_MULTICALL(cv); /* Set up the context for calling cv,
1905 and set local vars appropriately */
1908 /* set the value(s) af your parameter variables */
1909 MULTICALL; /* Make the actual call */
1912 POP_MULTICALL; /* Tear down the calling context */
1914 For some concrete examples, see the implementation of the
1915 first() and reduce() functions of List::Util 1.18. There you
1916 will also find a header file that emulates the multicall API
1917 on older versions of perl.
1921 L<perlxs>, L<perlguts>, L<perlembed>
1927 Special thanks to the following people who assisted in the creation of
1930 Jeff Okamoto, Tim Bunce, Nick Gianniotis, Steve Kelem, Gurusamy Sarathy
1935 Version 1.3, 14th Apr 1997