3 # Copyright (c) 1996, 1997, 1998 Malcolm Beattie
5 # You may distribute under the terms of either the GNU General Public
6 # License or the Artistic License, as specified in the README file.
12 @B::ISA = qw(Exporter);
14 # walkoptree_slow comes from B.pm (you are there),
15 # walkoptree comes from B.xs
21 # Our BOOT code needs $VERSION set, and will append to @EXPORT_OK.
22 # Want our constants loaded before the compiler meets OPf_KIDS below, as
23 # the combination of having the constant stay a Proxy Constant Subroutine
24 # and its value being inlined saves a little over .5K
30 push @B::EXPORT_OK, (qw(minus_c ppname save_BEGINs
31 class peekop cast_I32 cstring cchar hash threadsv_names
32 main_root main_start main_cv svref_2object opnumber
33 sub_generation amagic_generation perlstring
34 walkoptree_slow walkoptree walkoptree_exec walksymtable
35 parents comppadlist sv_undef compile_stats timing_info
36 begin_av init_av check_av end_av regex_padav dowarn
37 defstash curstash warnhook diehook inc_gv @optype
39 ), $] > 5.009 && 'unitcheck_av');
41 @B::SV::ISA = 'B::OBJECT';
42 @B::NULL::ISA = 'B::SV';
43 @B::PV::ISA = 'B::SV';
44 @B::IV::ISA = 'B::SV';
45 @B::NV::ISA = 'B::SV';
46 # RV is eliminated with 5.11.0, but effectively is a specialisation of IV now.
47 @B::RV::ISA = $] >= 5.011 ? 'B::IV' : 'B::SV';
48 @B::PVIV::ISA = qw(B::PV B::IV);
49 @B::PVNV::ISA = qw(B::PVIV B::NV);
50 @B::PVMG::ISA = 'B::PVNV';
51 @B::REGEXP::ISA = 'B::PVMG' if $] >= 5.011;
52 # Change in the inheritance hierarchy post 5.9.0
53 @B::PVLV::ISA = $] > 5.009 ? 'B::GV' : 'B::PVMG';
54 # BM is eliminated post 5.9.5, but effectively is a specialisation of GV now.
55 @B::BM::ISA = $] > 5.009005 ? 'B::GV' : 'B::PVMG';
56 @B::AV::ISA = 'B::PVMG';
57 @B::GV::ISA = 'B::PVMG';
58 @B::HV::ISA = 'B::PVMG';
59 @B::CV::ISA = 'B::PVMG';
60 @B::IO::ISA = 'B::PVMG';
61 @B::FM::ISA = 'B::CV';
63 @B::OP::ISA = 'B::OBJECT';
64 @B::UNOP::ISA = 'B::OP';
65 @B::BINOP::ISA = 'B::UNOP';
66 @B::LOGOP::ISA = 'B::UNOP';
67 @B::LISTOP::ISA = 'B::BINOP';
68 @B::SVOP::ISA = 'B::OP';
69 @B::PADOP::ISA = 'B::OP';
70 @B::PVOP::ISA = 'B::OP';
71 @B::LOOP::ISA = 'B::LISTOP';
72 @B::PMOP::ISA = 'B::LISTOP';
73 @B::COP::ISA = 'B::OP';
75 @B::SPECIAL::ISA = 'B::OBJECT';
77 @B::optype = qw(OP UNOP BINOP LOGOP LISTOP PMOP SVOP PADOP PVOP LOOP COP);
78 # bytecode.pl contained the following comment:
79 # Nullsv *must* come first in the following so that the condition
80 # ($$sv == 0) can continue to be used to test (sv == Nullsv).
81 @B::specialsv_name = qw(Nullsv &PL_sv_undef &PL_sv_yes &PL_sv_no
82 (SV*)pWARN_ALL (SV*)pWARN_NONE (SV*)pWARN_STD);
85 # Stop "-w" from complaining about the lack of a real B::OBJECT class
90 my $name = (shift())->NAME;
92 # The regex below corresponds to the isCONTROLVAR macro
95 $name =~ s/^([\cA-\cZ\c\\c[\c]\c?\c_\c^])/"^".
96 chr( utf8::unicode_to_native( 64 ^ ord($1) ))/e;
98 # When we say unicode_to_native we really mean ascii_to_native,
99 # which matters iff this is a non-ASCII platform (EBCDIC).
104 sub B::IV::int_value {
106 return (($self->FLAGS() & SVf_IVisUV()) ? $self->UVX : $self->IV);
109 sub B::NULL::as_string() {""}
110 *B::IV::as_string = \*B::IV::int_value;
111 *B::PV::as_string = \*B::PV::PV;
113 # The input typemap checking makes no distinction between different SV types,
114 # so the XS body will generate the same C code, despite the different XS
115 # "types". So there is no change in behaviour from doing "newXS" like this,
116 # compared with the old approach of having a (near) duplicate XS body.
117 # We should fix the typemap checking.
118 *B::IV::RV = \*B::PV::RV if $] > 5.012;
125 my ($class, $value) = @_;
127 walkoptree_debug($value);
137 sub parents { \@parents }
142 return sprintf("%s (0x%x) %s", class($op), $$op, $op->name);
145 sub walkoptree_slow {
146 my($op, $method, $level) = @_;
147 $op_count++; # just for statistics
149 warn(sprintf("walkoptree: %d. %s\n", $level, peekop($op))) if $debug;
150 $op->$method($level) if $op->can($method);
151 if ($$op && ($op->flags & OPf_KIDS)) {
153 unshift(@parents, $op);
154 for ($kid = $op->first; $$kid; $kid = $kid->sibling) {
155 walkoptree_slow($kid, $method, $level + 1);
159 if (class($op) eq 'PMOP'
160 && ref($op->pmreplroot)
161 && ${$op->pmreplroot}
162 && $op->pmreplroot->isa( 'B::OP' ))
164 unshift(@parents, $op);
165 walkoptree_slow($op->pmreplroot, $method, $level + 1);
171 return "Total number of OPs processed: $op_count\n";
175 my ($sec, $min, $hr) = localtime;
176 my ($user, $sys) = times;
177 sprintf("%02d:%02d:%02d user=$user sys=$sys",
178 $hr, $min, $sec, $user, $sys);
188 my ($obj, $value) = @_;
189 # warn(sprintf("savesym: sym_%x => %s\n", $$obj, $value)); # debug
190 $symtable{sprintf("sym_%x", $$obj)} = $value;
195 return $symtable{sprintf("sym_%x", $$obj)};
198 sub walkoptree_exec {
199 my ($op, $method, $level) = @_;
202 my $prefix = " " x $level;
203 for (; $$op; $op = $op->next) {
206 print $prefix, "goto $sym\n";
209 savesym($op, sprintf("%s (0x%lx)", class($op), $$op));
210 $op->$method($level);
213 /^(d?or(assign)?|and(assign)?|mapwhile|grepwhile|entertry|range|cond_expr)$/)
215 print $prefix, uc($1), " => {\n";
216 walkoptree_exec($op->other, $method, $level + 1);
217 print $prefix, "}\n";
218 } elsif ($ppname eq "match" || $ppname eq "subst") {
219 my $pmreplstart = $op->pmreplstart;
221 print $prefix, "PMREPLSTART => {\n";
222 walkoptree_exec($pmreplstart, $method, $level + 1);
223 print $prefix, "}\n";
225 } elsif ($ppname eq "substcont") {
226 print $prefix, "SUBSTCONT => {\n";
227 walkoptree_exec($op->other->pmreplstart, $method, $level + 1);
228 print $prefix, "}\n";
230 } elsif ($ppname eq "enterloop") {
231 print $prefix, "REDO => {\n";
232 walkoptree_exec($op->redoop, $method, $level + 1);
233 print $prefix, "}\n", $prefix, "NEXT => {\n";
234 walkoptree_exec($op->nextop, $method, $level + 1);
235 print $prefix, "}\n", $prefix, "LAST => {\n";
236 walkoptree_exec($op->lastop, $method, $level + 1);
237 print $prefix, "}\n";
238 } elsif ($ppname eq "subst") {
239 my $replstart = $op->pmreplstart;
241 print $prefix, "SUBST => {\n";
242 walkoptree_exec($replstart, $method, $level + 1);
243 print $prefix, "}\n";
250 my ($symref, $method, $recurse, $prefix) = @_;
255 $prefix = '' unless defined $prefix;
256 while (($sym, $ref) = each %$symref) {
257 $fullname = "*main::".$prefix.$sym;
259 $sym = $prefix . $sym;
260 if (svref_2object(\*$sym)->NAME ne "main::" && $sym ne "<none>::" && &$recurse($sym)) {
261 walksymtable(\%$fullname, $method, $recurse, $sym);
264 svref_2object(\*$fullname)->$method();
275 my ($class, $section, $symtable, $default) = @_;
276 $output_fh ||= FileHandle->new_tmpfile;
277 my $obj = bless [-1, $section, $symtable, $default], $class;
278 $sections{$section} = $obj;
283 my ($class, $section) = @_;
284 return $sections{$section};
289 while (defined($_ = shift)) {
290 print $output_fh "$section->[1]\t$_\n";
297 return $section->[0];
302 return $section->[1];
307 return $section->[2];
312 return $section->[3];
316 my ($section, $fh, $format) = @_;
317 my $name = $section->name;
318 my $sym = $section->symtable || {};
319 my $default = $section->default;
321 seek($output_fh, 0, 0);
322 while (<$output_fh>) {
327 exists($sym->{$1}) ? $sym->{$1} : $default;
329 printf $fh $format, $_;
341 B - The Perl Compiler Backend
349 The C<B> module supplies classes which allow a Perl program to delve
350 into its own innards. It is the module used to implement the
351 "backends" of the Perl compiler. Usage of the compiler does not
352 require knowledge of this module: see the F<O> module for the
353 user-visible part. The C<B> module is of use to those who want to
354 write new compiler backends. This documentation assumes that the
355 reader knows a fair amount about perl's internals including such
356 things as SVs, OPs and the internal symbol table and syntax tree
361 The C<B> module contains a set of utility functions for querying the
362 current state of the Perl interpreter; typically these functions
363 return objects from the B::SV and B::OP classes, or their derived
364 classes. These classes in turn define methods for querying the
365 resulting objects about their own internal state.
367 =head1 Utility Functions
369 The C<B> module exports a variety of functions: some are simple
370 utility functions, others provide a Perl program with a way to
371 get an initial "handle" on an internal object.
373 =head2 Functions Returning C<B::SV>, C<B::AV>, C<B::HV>, and C<B::CV> objects
375 For descriptions of the class hierarchy of these objects and the
376 methods that can be called on them, see below, L<"OVERVIEW OF
377 CLASSES"> and L<"SV-RELATED CLASSES">.
383 Returns the SV object corresponding to the C variable C<sv_undef>.
387 Returns the SV object corresponding to the C variable C<sv_yes>.
391 Returns the SV object corresponding to the C variable C<sv_no>.
393 =item svref_2object(SVREF)
395 Takes a reference to any Perl value, and turns the referred-to value
396 into an object in the appropriate B::OP-derived or B::SV-derived
397 class. Apart from functions such as C<main_root>, this is the primary
398 way to get an initial "handle" on an internal perl data structure
399 which can then be followed with the other access methods.
401 The returned object will only be valid as long as the underlying OPs
402 and SVs continue to exist. Do not attempt to use the object after the
403 underlying structures are freed.
405 =item amagic_generation
407 Returns the SV object corresponding to the C variable C<amagic_generation>.
411 Returns the AV object (i.e. in class B::AV) representing INIT blocks.
415 Returns the AV object (i.e. in class B::AV) representing CHECK blocks.
419 Returns the AV object (i.e. in class B::AV) representing UNITCHECK blocks.
423 Returns the AV object (i.e. in class B::AV) representing BEGIN blocks.
427 Returns the AV object (i.e. in class B::AV) representing END blocks.
431 Returns the AV object (i.e. in class B::AV) of the global comppadlist.
435 Only when perl was compiled with ithreads.
439 Return the (faked) CV corresponding to the main part of the Perl
444 =head2 Functions for Examining the Symbol Table
448 =item walksymtable(SYMREF, METHOD, RECURSE, PREFIX)
450 Walk the symbol table starting at SYMREF and call METHOD on each
451 symbol (a B::GV object) visited. When the walk reaches package
452 symbols (such as "Foo::") it invokes RECURSE, passing in the symbol
453 name, and only recurses into the package if that sub returns true.
455 PREFIX is the name of the SYMREF you're walking.
459 # Walk CGI's symbol table calling print_subs on each symbol.
460 # Recurse only into CGI::Util::
461 walksymtable(\%CGI::, 'print_subs',
462 sub { $_[0] eq 'CGI::Util::' }, 'CGI::');
464 print_subs() is a B::GV method you have declared. Also see L<"B::GV
469 =head2 Functions Returning C<B::OP> objects or for walking op trees
471 For descriptions of the class hierarchy of these objects and the
472 methods that can be called on them, see below, L<"OVERVIEW OF
473 CLASSES"> and L<"OP-RELATED CLASSES">.
479 Returns the root op (i.e. an object in the appropriate B::OP-derived
480 class) of the main part of the Perl program.
484 Returns the starting op of the main part of the Perl program.
486 =item walkoptree(OP, METHOD)
488 Does a tree-walk of the syntax tree based at OP and calls METHOD on
489 each op it visits. Each node is visited before its children. If
490 C<walkoptree_debug> (see below) has been called to turn debugging on then
491 the method C<walkoptree_debug> is called on each op before METHOD is
494 =item walkoptree_debug(DEBUG)
496 Returns the current debugging flag for C<walkoptree>. If the optional
497 DEBUG argument is non-zero, it sets the debugging flag to that. See
498 the description of C<walkoptree> above for what the debugging flag
503 =head2 Miscellaneous Utility Functions
509 Return the PP function name (e.g. "pp_add") of op number OPNUM.
513 Returns a string in the form "0x..." representing the value of the
514 internal hash function used by perl on string STR.
518 Casts I to the internal I32 type used by that perl.
522 Does the equivalent of the C<-c> command-line option. Obviously, this
523 is only useful in a BEGIN block or else the flag is set too late.
527 Returns a double-quote-surrounded escaped version of STR which can
528 be used as a string in C source code.
530 =item perlstring(STR)
532 Returns a double-quote-surrounded escaped version of STR which can
533 be used as a string in Perl source code.
537 Returns the class of an object without the part of the classname
538 preceding the first C<"::">. This is used to turn C<"B::UNOP"> into
539 C<"UNOP"> for example.
543 In a perl compiled for threads, this returns a list of the special
544 per-thread threadsv variables.
548 =head2 Exported utility variables
554 my $op_type = $optype[$op_type_num];
556 A simple mapping of the op type number to its type (like 'COP' or 'BINOP').
558 =item @specialsv_name
560 my $sv_name = $specialsv_name[$sv_index];
562 Certain SV types are considered 'special'. They're represented by
563 B::SPECIAL and are referred to by a number from the specialsv_list.
564 This array maps that number back to the name of the SV (like 'Nullsv'
570 =head1 OVERVIEW OF CLASSES
572 The C structures used by Perl's internals to hold SV and OP
573 information (PVIV, AV, HV, ..., OP, SVOP, UNOP, ...) are modelled on a
574 class hierarchy and the C<B> module gives access to them via a true
575 object hierarchy. Structure fields which point to other objects
576 (whether types of SV or types of OP) are represented by the C<B>
577 module as Perl objects of the appropriate class.
579 The bulk of the C<B> module is the methods for accessing fields of
582 Note that all access is read-only. You cannot modify the internals by
583 using this module. Also, note that the B::OP and B::SV objects created
584 by this module are only valid for as long as the underlying objects
585 exist; their creation doesn't increase the reference counts of the
586 underlying objects. Trying to access the fields of a freed object will
587 give incomprehensible results, or worse.
589 =head2 SV-RELATED CLASSES
591 B::IV, B::NV, B::RV, B::PV, B::PVIV, B::PVNV, B::PVMG, B::BM (5.9.5 and
592 earlier), B::PVLV, B::AV, B::HV, B::CV, B::GV, B::FM, B::IO. These classes
593 correspond in the obvious way to the underlying C structures of similar names.
594 The inheritance hierarchy mimics the underlying C "inheritance". For the
595 5.10.x branch, (I<ie> 5.10.0, 5.10.1 I<etc>) this is:
599 +------------+------------+------------+
601 B::PV B::IV B::NV B::RV
613 +-----+-----+-----+-----+
615 B::AV B::GV B::HV B::CV B::IO
620 For 5.9.0 and earlier, PVLV is a direct subclass of PVMG, and BM is still
621 present as a distinct type, so the base of this diagram is
628 +------+-----+-----+-----+-----+-----+
630 B::PVLV B::BM B::AV B::GV B::HV B::CV B::IO
635 For 5.11.0 and later, B::RV is abolished, and IVs can be used to store
636 references, and a new type B::REGEXP is introduced, giving this structure:
640 +------------+------------+
654 +-------+-------+---+---+-------+-------+
656 B::AV B::GV B::HV B::CV B::IO B::REGEXP
662 Access methods correspond to the underlying C macros for field access,
663 usually with the leading "class indication" prefix removed (Sv, Av,
664 Hv, ...). The leading prefix is only left in cases where its removal
665 would cause a clash in method name. For example, C<GvREFCNT> stays
666 as-is since its abbreviation would clash with the "superclass" method
667 C<REFCNT> (corresponding to the C function C<SvREFCNT>).
679 Returns a reference to the regular scalar corresponding to this
680 B::SV object. In other words, this method is the inverse operation
681 to the svref_2object() subroutine. This scalar and other data it points
682 at should be considered read-only: modifying them is neither safe nor
683 guaranteed to have a sensible effect.
693 Returns the value of the IV, I<interpreted as
694 a signed integer>. This will be misleading
695 if C<FLAGS & SVf_IVisUV>. Perhaps you want the
696 C<int_value> method instead?
704 This method returns the value of the IV as an integer.
705 It differs from C<IV> in that it returns the correct
706 value regardless of whether it's stored signed or
739 This method is the one you usually want. It constructs a
740 string using the length and offset information in the struct:
741 for ordinary scalars it will return the string that you'd see
742 from Perl, even if it contains null characters.
746 Same as B::RV::RV, except that it will die() if the PV isn't
751 This method is less often useful. It assumes that the string
752 stored in the struct is null-terminated, and disregards the
755 It is the appropriate method to use if you need to get the name
756 of a lexical variable from a padname array. Lexical variable names
757 are always stored with a null terminator, and the length field
758 (CUR) is overloaded for other purposes and can't be relied on here.
762 This method returns the internal length field, which consists of the number
763 of internal bytes, not necessarily the number of logical characters.
767 This method returns the number of bytes allocated (via malloc) for storing
768 the string. This is 0 if the scalar does not "own" the string.
772 =head2 B::PVMG Methods
782 =head2 B::MAGIC Methods
790 Only valid on r-magic, returns the string that generated the regexp.
800 Will die() if called on r-magic.
806 Only valid on r-magic, returns the integer value of the REGEX stored
811 =head2 B::PVLV Methods
845 This method returns TRUE if the GP field of the GV is NULL.
851 This method returns the name of the glob, but if the first
852 character of the name is a control character, then it converts
853 it to ^X first, so that *^G would return "^G" rather than "\cG".
855 It's useful if you want to print out the name of a variable.
856 If you restrict yourself to globs which exist at compile-time
857 then the result ought to be unambiguous, because code like
858 C<${"^G"} = 1> is compiled as two ops - a constant string and
859 a dereference (rv2gv) - so that the glob is created at runtime.
861 If you're working with globs at runtime, and need to disambiguate
862 *^G from *{"^G"}, then you should use the raw NAME method.
896 B::IO objects derive from IO objects and you will get more information from
897 the IO object itself.
901 $gvio = B::svref_2object(\*main::stdin)->IO;
902 $IO = $gvio->object_2svref();
931 A character symbolizing the type of IO Handle.
944 \0 closed internal handle
950 Takes one argument ( 'stdin' | 'stdout' | 'stderr' ) and returns true
951 if the IoIFP of the object is equal to the handle whose name was
952 passed as argument; i.e., $io->IsSTD('stderr') is true if
953 IoIFP($io) == PerlIO_stderr().
969 Like C<ARRAY>, but takes an index as an argument to get only one element,
970 rather than a list of all of them.
974 This method is deprecated if running under Perl 5.8, and is no longer present
975 if running under Perl 5.9
979 This method returns the AV specific
980 flags. In Perl 5.9 these are now stored
981 in with the main SV flags, so this method is no longer present.
1011 For constant subroutines, returns the constant SV returned by the subroutine.
1019 =head2 B::HV Methods
1037 This method is not present if running under Perl 5.9, as the PMROOT
1038 information is no longer stored directly in the hash.
1042 =head2 OP-RELATED CLASSES
1044 C<B::OP>, C<B::UNOP>, C<B::BINOP>, C<B::LOGOP>, C<B::LISTOP>, C<B::PMOP>,
1045 C<B::SVOP>, C<B::PADOP>, C<B::PVOP>, C<B::LOOP>, C<B::COP>.
1047 These classes correspond in the obvious way to the underlying C
1048 structures of similar names. The inheritance hierarchy mimics the
1049 underlying C "inheritance":
1053 +---------------+--------+--------+-------+
1055 B::UNOP B::SVOP B::PADOP B::COP B::PVOP
1066 Access methods correspond to the underlying C structre field names,
1067 with the leading "class indication" prefix (C<"op_">) removed.
1069 =head2 B::OP Methods
1071 These methods get the values of similarly named fields within the OP
1072 data structure. See top of C<op.h> for more info.
1082 This returns the op name as a string (e.g. "add", "rv2av").
1086 This returns the function name as a string (e.g. "PL_ppaddr[OP_ADD]",
1087 "PL_ppaddr[OP_RV2AV]").
1091 This returns the op description from the global C PL_op_desc array
1092 (e.g. "addition" "array deref").
1108 =head2 B::UNOP METHOD
1116 =head2 B::BINOP METHOD
1124 =head2 B::LOGOP METHOD
1132 =head2 B::LISTOP METHOD
1140 =head2 B::PMOP Methods
1150 Only up to Perl 5.9.4
1162 Only when perl was compiled with ithreads.
1166 =head2 B::SVOP METHOD
1176 =head2 B::PADOP METHOD
1184 =head2 B::PVOP METHOD
1192 =head2 B::LOOP Methods
1204 =head2 B::COP Methods
1237 Malcolm Beattie, C<mbeattie@sable.ox.ac.uk>