This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Add "refactor xsubpp to be a thin wrapper around ExtUtils::ParseXS"
[perl5.git] / pod / perltodo.pod
CommitLineData
7711098a
GS
1=head1 NAME
2
3perltodo - Perl TO-DO List
4
5=head1 DESCRIPTION
e50bb9a1 6
0bdfc961
NC
7This is a list of wishes for Perl. The tasks we think are smaller or easier
8are listed first. Anyone is welcome to work on any of these, but it's a good
9idea to first contact I<perl5-porters@perl.org> to avoid duplication of
10effort. By all means contact a pumpking privately first if you prefer.
e50bb9a1 11
0bdfc961
NC
12Whilst patches to make the list shorter are most welcome, ideas to add to
13the list are also encouraged. Check the perl5-porters archives for past
14ideas, and any discussion about them. One set of archives may be found at:
e50bb9a1 15
0bdfc961 16 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
938c8732 17
938c8732 18
938c8732 19
938c8732 20
e50bb9a1 21
0bdfc961 22=head1 Tasks that only need Perl knowledge
e50bb9a1 23
0bdfc961 24=head2 common test code for timed bail out
e50bb9a1 25
0bdfc961
NC
26Write portable self destruct code for tests to stop them burning CPU in
27infinite loops. This needs to avoid using alarm, as some of the tests are
28testing alarm/sleep or timers.
e50bb9a1 29
0bdfc961 30=head2 POD -> HTML conversion in the core still sucks
e50bb9a1 31
938c8732 32Which is crazy given just how simple POD purports to be, and how simple HTML
adebf063
NC
33can be. It's not actually I<as> simple as it sounds, particularly with the
34flexibility POD allows for C<=item>, but it would be good to improve the
35visual appeal of the HTML generated, and to avoid it having any validation
36errors. See also L</make HTML install work>, as the layout of installation tree
37is needed to improve the cross-linking.
938c8732 38
0bdfc961 39=head2 Make Schwern poorer
e50bb9a1 40
0bdfc961
NC
41We should have for everything. When all the core's modules are tested,
42Schwern has promised to donate to $500 to TPF. We may need volunteers to
43hold him upside down and shake vigorously in order to actually extract the
44cash.
3958b146 45
0bdfc961 46See F<t/lib/1_compile.t> for the 3 remaining modules that need tests.
e50bb9a1 47
0bdfc961 48=head2 Improve the coverage of the core tests
e50bb9a1 49
0bdfc961
NC
50Use Devel::Cover to ascertain the core's test coverage, then add tests that
51are currently missing.
30222c0f 52
0bdfc961 53=head2 test B
e50bb9a1 54
0bdfc961 55A full test suite for the B module would be nice.
e50bb9a1 56
0bdfc961 57=head2 A decent benchmark
e50bb9a1 58
0bdfc961
NC
59perlbench seems impervious to any recent changes made to the perl core. It
60would be useful to have a reasonable general benchmarking suite that roughly
61represented what current perl programs do, and measurably reported whether
62tweaks to the core improve, degrade or don't really affect performance, to
63guide people attempting to optimise the guts of perl. Gisle would welcome
64new tests for perlbench.
6168cf99 65
0bdfc961 66=head2 fix tainting bugs
6168cf99 67
0bdfc961
NC
68Fix the bugs revealed by running the test suite with the C<-t> switch (via
69C<make test.taintwarn>).
e50bb9a1 70
0bdfc961 71=head2 Dual life everything
e50bb9a1 72
0bdfc961
NC
73As part of the "dists" plan, anything that doesn't belong in the smallest perl
74distribution needs to be dual lifed. Anything else can be too. Figure out what
75changes would be needed to package that module and its tests up for CPAN, and
76do so. Test it with older perl releases, and fix the problems you find.
e50bb9a1 77
0bdfc961 78=head2 Improving C<threads::shared>
722d2a37 79
0bdfc961
NC
80Investigate whether C<threads::shared> could share aggregates properly with
81only Perl level changes to shared.pm
722d2a37 82
0bdfc961 83=head2 POSIX memory footprint
e50bb9a1 84
0bdfc961
NC
85Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
86various times worked to cut it down. There is probably still fat to cut out -
87for example POSIX passes Exporter some very memory hungry data structures.
e50bb9a1 88
e50bb9a1 89
e50bb9a1 90
e50bb9a1 91
e50bb9a1 92
adebf063 93
adebf063 94
0bdfc961 95=head1 Tasks that need a little sysadmin-type knowledge
e50bb9a1 96
0bdfc961
NC
97Or if you prefer, tasks that you would learn from, and broaden your skills
98base...
e50bb9a1 99
cd793d32 100=head2 make HTML install work
e50bb9a1 101
adebf063
NC
102There is an C<installhtml> target in the Makefile. It's marked as
103"experimental". It would be good to get this tested, make it work reliably, and
104remove the "experimental" tag. This would include
105
106=over 4
107
108=item 1
109
110Checking that cross linking between various parts of the documentation works.
111In particular that links work between the modules (files with POD in F<lib/>)
112and the core documentation (files in F<pod/>)
113
114=item 2
115
116Work out how to split perlfunc into chunks, preferably one per function group,
117preferably with general case code that could be used elsewhere. Challenges
118here are correctly identifying the groups of functions that go together, and
119making the right named external cross-links point to the right page. Things to
120be aware of are C<-X>, groups such as C<getpwnam> to C<endservent>, two or
121more C<=items> giving the different parameter lists, such as
122
123 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
124
125 =item substr EXPR,OFFSET,LENGTH
126
127 =item substr EXPR,OFFSET
128
129and different parameter lists having different meanings. (eg C<select>)
130
131=back
3a89a73c 132
0bdfc961
NC
133=head2 compressed man pages
134
135Be able to install them. This would probably need a configure test to see how
136the system does compressed man pages (same directory/different directory?
137same filename/different filename), as well as tweaking the F<installman> script
138to compress as necessary.
139
30222c0f
NC
140=head2 Add a code coverage target to the Makefile
141
142Make it easy for anyone to run Devel::Cover on the core's tests. The steps
143to do this manually are roughly
144
145=over 4
146
147=item *
148
149do a normal C<Configure>, but include Devel::Cover as a module to install
150(see F<INSTALL> for how to do this)
151
152=item *
153
154 make perl
155
156=item *
157
158 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
159
160=item *
161
162Process the resulting Devel::Cover database
163
164=back
165
166This just give you the coverage of the F<.pm>s. To also get the C level
167coverage you need to
168
169=over 4
170
171=item *
172
173Additionally tell C<Configure> to use the appropriate C compiler flags for
174C<gcov>
175
176=item *
177
178 make perl.gcov
179
180(instead of C<make perl>)
181
182=item *
183
184After running the tests run C<gcov> to generate all the F<.gcov> files.
185(Including down in the subdirectories of F<ext/>
186
187=item *
188
189(From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
190to get their stats into the cover_db directory.
191
192=item *
193
194Then process the Devel::Cover database
195
196=back
197
198It would be good to add a single switch to C<Configure> to specify that you
199wanted to perform perl level coverage, and another to specify C level
200coverage, and have C<Configure> and the F<Makefile> do all the right things
201automatically.
202
0bdfc961
NC
203=head2 Make Config.pm cope with differences between build and installed perl
204
205Quite often vendors ship a perl binary compiled with their (pay-for)
206compilers. People install a free compiler, such as gcc. To work out how to
207build extensions, Perl interrogates C<%Config>, so in this situation
208C<%Config> describes compilers that aren't there, and extension building
209fails. This forces people into choosing between re-compiling perl themselves
210using the compiler they have, or only using modules that the vendor ships.
211
212It would be good to find a way teach C<Config.pm> about the installation setup,
213possibly involving probing at install time or later, so that the C<%Config> in
214a binary distribution better describes the installed machine, when the
215installed machine differs from the build machine in some significant way.
216
217=head2 Relocatable perl
218
219The C level patches needed to create a relocatable perl binary are done, as
220is the work on Config.pm. All that's left to do is the C<Configure> tweaking
221to let people specify how they want to do the install.
222
46925299 223=head2 make parallel builds work
0bdfc961 224
46925299
NC
225Currently parallel builds (such as C<make -j3>) don't work reliably. We believe
226that this is due to incomplete dependency specification in the F<Makefile>.
227It would be good if someone were able to track down the causes of these
228problems, so that parallel builds worked properly.
0bdfc961 229
8523e164
NC
230=head2 Refactor C<xsubpp> to be a thin wrapper around C<ExtUtils::ParseXS>
231
232C<ExtUtils::ParseXS> encapsulates a version of the C<xsubpp> into a module.
233In effect this is a code fork, and it's likely that C<xsubpp> has had some
234bug fixes since the code from C<ExtUtils::ParseXS> was derived. It would be
235good to merge the differences in, reduce down to 1 canonical implementation,
236and convert C<xsubpp> to a very thin command line wrapper to
237C<ExtUtils::ParseXS>.
238
239In theory this needs no real C knowledge, as one way of approaching this task
240is to ensure that C<ExtUtils::ParseXS> generates identical output to C<xsubpp>
241for input XS files, which does not require understanding the contents of the
242output C file. However, some C knowledge is likely to help with testing, and
243locating/producing comprehensive test cases.
244
245
246
0bdfc961
NC
247
248
249=head1 Tasks that need a little C knowledge
250
251These tasks would need a little C knowledge, but don't need any specific
252background or experience with XS, or how the Perl interpreter works
253
254=head2 Make it clear from -v if this is the exact official release
89007cb3
NC
255
256Currently perl from p4/rsync ships with a patchlevel.h file that usually
257defines one local patch, of the form "MAINT12345" or "RC1". The output of
258perl -v doesn't report that a perl isn't an official release, and this
259information can get lost in bugs reports. Because of this, the minor version
fa11829f 260isn't bumped up until RC time, to minimise the possibility of versions of perl
89007cb3
NC
261escaping that believe themselves to be newer than they actually are.
262
263It would be useful to find an elegant way to have the "this is an interim
264maintenance release" or "this is a release candidate" in the terse -v output,
265and have it so that it's easy for the pumpking to remove this just as the
266release tarball is rolled up. This way the version pulled out of rsync would
267always say "I'm a development release" and it would be safe to bump the
268reported minor version as soon as a release ships, which would aid perl
269developers.
270
0bdfc961
NC
271This task is really about thinking of an elegant way to arrange the C source
272such that it's trivial for the Pumpking to flag "this is an official release"
273when making a tarball, yet leave the default source saying "I'm not the
274official release".
275
0f788cd2
NC
276=head2 Ordering of "global" variables.
277
278F<thrdvar.h> and F<intrpvarh> define the "global" variables that need to be
279per-thread under ithreads, where the variables are actually elements in a
280structure. As C dictates, the variables must be laid out in order of
281declaration. There is a comment
282C</* Important ones in the first cache line (if alignment is done right) */>
283which implies that at some point in the past the ordering was carefully chosen
284(at least in part). However, it's clear that the ordering is less than perfect,
285as currently there are things such as 7 C<bool>s in a row, then something
286typically requiring 4 byte alignment, and then an odd C<bool> later on.
287(C<bool>s are typically defined as C<char>s). So it would be good for someone
288to review the ordering of the variables, to see how much alignment padding can
289be removed.
290
0bdfc961
NC
291=head2 bincompat functions
292
293There are lots of functions which are retained for binary compatibility.
294Clean these up. Move them to mathom.c, and don't compile for blead?
295
62403a3c
NC
296=head2 am I hot or not?
297
298The idea of F<pp_hot.c> is that it contains the I<hot> ops, the ops that are
299most commonly used. The idea is that by grouping them, their object code will
300be adjacent in the executable, so they have a greater chance of already being
301in the CPU cache (or swapped in) due to being near another op already in use.
302
303Except that it's not clear if these really are the most commonly used ops. So
304anyone feeling like exercising their skill with coverage and profiling tools
305might want to determine what ops I<really> are the most commonly used. And in
306turn suggest evictions and promotions to achieve a better F<pp_hot.c>.
307
c99e3826
NC
308=head2 emulate the per-thread memory pool on Unix
309
310For Windows, ithreads allocates memory for each thread from a separate pool,
311which it discards at thread exit. It also checks that memory is free()d to
312the correct pool. Neither check is done on Unix, so code developed there won't
313be subject to such strictures, so can harbour bugs that only show up when the
314code reaches Windows.
315
316It would be good to be able to optionally emulate the Window pool system on
317Unix, to let developers who only have access to Unix, or want to use
318Unix-specific debugging tools, check for these problems. To do this would
319involve figuring out how the C<PerlMem_*> macros wrap C<malloc()> access, and
320providing a layer that records/checks the identity of the thread making the
321call, and recording all the memory allocated by each thread via this API so
322that it can be summarily free()d at thread exit. One implementation idea
323would be to increase the size of allocation, and store the C<my_perl> pointer
324(to identify the thread) at the start, along with pointers to make a linked
325list of blocks for this thread. To avoid alignment problems it would be
326necessary to do something like
327
328 union memory_header_padded {
329 struct memory_header {
330 void *thread_id; /* For my_perl */
331 void *next; /* Pointer to next block for this thread */
332 } data;
333 long double padding; /* whatever type has maximal alignment constraint */
334 };
335
336
337although C<long double> might not be the only type to add to the padding
338union.
62403a3c 339
077e3186
NC
340=head2 reduce duplication in sv_setsv_flags
341
342C<Perl_sv_setsv_flags> has a comment
343C</* There's a lot of redundancy below but we're going for speed here */>
344
345Whilst this was true 10 years ago, the growing disparity between RAM and CPU
346speeds mean that the trade offs have changed. In addition, the duplicate code
347adds to the maintenance burden. It would be good to see how much of the
348redundancy can be pruned, particular in the less common paths. (Profiling
349tools at the ready...). For example, why does the test for
350"Can't redefine active sort subroutine" need to occur in two places?
351
0bdfc961
NC
352
353
354
0bdfc961 355=head1 Tasks that need a knowledge of XS
e50bb9a1 356
0bdfc961
NC
357These tasks would need C knowledge, and roughly the level of knowledge of
358the perl API that comes from writing modules that use XS to interface to
359C.
360
361=head2 IPv6
362
363Clean this up. Check everything in core works
364
4a750395
NC
365=head2 shrink C<GV>s, C<CV>s
366
367By removing unused elements and careful re-ordering, the structures for C<AV>s
368and C<HV>s have recently been shrunk considerably. It's probable that the same
369approach would find savings in C<GV>s and C<CV>s, if not all the other
370larger-than-C<PVMG> types.
371
e593da1a
NC
372=head2 merge Perl_sv_2[inpu]v
373
374There's a lot of code shared between C<Perl_sv_2iv_flags>,
375C<Perl_sv_2uv_flags>, C<Perl_sv_2nv>, and C<Perl_sv_2pv_flags>. It would be
376interesting to see if some of it can be merged into common shared static
377functions. In particular, C<Perl_sv_2uv_flags> started out as a cut&paste
378from C<Perl_sv_2iv_flags> around 5.005_50 time, and it may be possible to
379replace both with a single function that returns a value or union which is
380split out by the macros in F<sv.h>
381
0bdfc961
NC
382=head2 UTF8 caching code
383
384The string position/offset cache is not optional. It should be.
385
386=head2 Implicit Latin 1 => Unicode translation
387
388Conversions from byte strings to UTF-8 currently map high bit characters
389to Unicode without translation (or, depending on how you look at it, by
390implicitly assuming that the byte strings are in Latin-1). As perl assumes
391the C locale by default, upgrading a string to UTF-8 may change the
392meaning of its contents regarding character classes, case mapping, etc.
393This should probably emit a warning (at least).
394
395This task is incremental - even a little bit of work on it will help.
e50bb9a1 396
cd793d32 397=head2 autovivification
e50bb9a1 398
cd793d32 399Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
e50bb9a1 400
0bdfc961 401This task is incremental - even a little bit of work on it will help.
e50bb9a1 402
0bdfc961 403=head2 Unicode in Filenames
e50bb9a1 404
0bdfc961
NC
405chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
406opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
407system, truncate, unlink, utime, -X. All these could potentially accept
408Unicode filenames either as input or output (and in the case of system
409and qx Unicode in general, as input or output to/from the shell).
410Whether a filesystem - an operating system pair understands Unicode in
411filenames varies.
e50bb9a1 412
0bdfc961
NC
413Known combinations that have some level of understanding include
414Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
415OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
416create Unicode filenames, what forms of Unicode are accepted and used
417(UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
418and so on, varies. Finding the right level of interfacing to Perl
419requires some thought. Remember that an OS does not implicate a
420filesystem.
e50bb9a1 421
0bdfc961
NC
422(The Windows -C command flag "wide API support" has been at least
423temporarily retired in 5.8.1, and the -C has been repurposed, see
424L<perlrun>.)
969e704b 425
0bdfc961 426=head2 Unicode in %ENV
969e704b 427
0bdfc961 428Currently the %ENV entries are always byte strings.
e50bb9a1 429
0bdfc961 430=head2 use less 'memory'
e50bb9a1 431
0bdfc961
NC
432Investigate trade offs to switch out perl's choices on memory usage.
433Particularly perl should be able to give memory back.
e50bb9a1 434
0bdfc961 435This task is incremental - even a little bit of work on it will help.
0abe3f7c 436
0bdfc961 437=head2 Re-implement C<:unique> in a way that is actually thread-safe
0abe3f7c 438
0bdfc961
NC
439The old implementation made bad assumptions on several levels. A good 90%
440solution might be just to make C<:unique> work to share the string buffer
441of SvPVs. That way large constant strings can be shared between ithreads,
442such as the configuration information in F<Config>.
0abe3f7c 443
0bdfc961 444=head2 Make tainting consistent
0abe3f7c 445
0bdfc961
NC
446Tainting would be easier to use if it didn't take documented shortcuts and
447allow taint to "leak" everywhere within an expression.
0abe3f7c 448
0bdfc961 449=head2 readpipe(LIST)
0abe3f7c 450
0bdfc961
NC
451system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
452running a shell. readpipe() (the function behind qx//) could be similarly
453extended.
0abe3f7c 454
e50bb9a1 455
e50bb9a1 456
e50bb9a1 457
f86a8bc5 458
0bdfc961 459=head1 Tasks that need a knowledge of the interpreter
3298bd4d 460
0bdfc961
NC
461These tasks would need C knowledge, and knowledge of how the interpreter works,
462or a willingness to learn.
3298bd4d 463
0bdfc961
NC
464=head2 lexical pragmas
465
466Reimplement the mechanism of lexical pragmas to be more extensible. Fix
467current pragmas that don't work well (or at all) with lexical scopes or in
468run-time eval(STRING) (C<sort>, C<re>, C<encoding> for example). MJD has a
469preliminary patch that implements this.
0562c0e3 470
d10fc472 471=head2 Attach/detach debugger from running program
1626a787 472
cd793d32
NC
473The old perltodo notes "With C<gdb>, you can attach the debugger to a running
474program if you pass the process ID. It would be good to do this with the Perl
0bdfc961
NC
475debugger on a running Perl program, although I'm not sure how it would be
476done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
1626a787 477
0bdfc961 478=head2 inlining autoloaded constants
d10fc472 479
0bdfc961
NC
480Currently the optimiser can inline constants when expressed as subroutines
481with prototype ($) that return a constant. Likewise, many packages wrapping
482C libraries export lots of constants as subroutines which are AUTOLOADed on
483demand. However, these have no prototypes, so can't be seen as constants by
484the optimiser. Some way of cheaply (low syntax, low memory overhead) to the
485perl compiler that a name is a constant would be great, so that it knows to
486call the AUTOLOAD routine at compile time, and then inline the constant.
80b46460 487
0bdfc961 488=head2 Constant folding
80b46460 489
0bdfc961
NC
490The peephole optimiser should trap errors during constant folding, and give
491up on the folding, rather than bailing out at compile time. It is quite
492possible that the unfoldable constant is in unreachable code, eg something
493akin to C<$a = 0/0 if 0;>
494
495=head2 LVALUE functions for lists
496
497The old perltodo notes that lvalue functions don't work for list or hash
498slices. This would be good to fix.
499
500=head2 LVALUE functions in the debugger
501
502The old perltodo notes that lvalue functions don't work in the debugger. This
503would be good to fix.
504
505=head2 _ prototype character
506
507Study the possibility of adding a new prototype character, C<_>, meaning
508"this argument defaults to $_".
509
510=head2 @INC source filter to Filter::Simple
511
512The second return value from a sub in @INC can be a source filter. This isn't
513documented. It should be changed to use Filter::Simple, tested and documented.
514
515=head2 regexp optimiser optional
516
517The regexp optimiser is not optional. It should configurable to be, to allow
518its performance to be measured, and its bugs to be easily demonstrated.
519
520=head2 UNITCHECK
521
522Introduce a new special block, UNITCHECK, which is run at the end of a
523compilation unit (module, file, eval(STRING) block). This will correspond to
524the Perl 6 CHECK. Perl 5's CHECK cannot be changed or removed because the
525O.pm/B.pm backend framework depends on it.
526
527=head2 optional optimizer
528
529Make the peephole optimizer optional. Currently it performs two tasks as
530it walks the optree - genuine peephole optimisations, and necessary fixups of
531ops. It would be good to find an efficient way to switch out the
532optimisations whilst keeping the fixups.
533
534=head2 You WANT *how* many
535
536Currently contexts are void, scalar and list. split has a special mechanism in
537place to pass in the number of return values wanted. It would be useful to
538have a general mechanism for this, backwards compatible and little speed hit.
539This would allow proposals such as short circuiting sort to be implemented
540as a module on CPAN.
541
542=head2 lexical aliases
543
544Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
545
546=head2 entersub XS vs Perl
547
548At the moment pp_entersub is huge, and has code to deal with entering both
549perl and XS subroutines. Subroutine implementations rarely change between
550perl and XS at run time, so investigate using 2 ops to enter subs (one for
551XS, one for perl) and swap between if a sub is redefined.
2810d901
NC
552
553=head2 Self ties
554
555self ties are currently illegal because they caused too many segfaults. Maybe
556the causes of these could be tracked down and self-ties on all types re-
557instated.
0bdfc961
NC
558
559=head2 Optimize away @_
560
561The old perltodo notes "Look at the "reification" code in C<av.c>".
562
563=head2 switch ops
564
565The old perltodo notes "Although we have C<Switch.pm> in core, Larry points to
566the dormant C<nswitch> and C<cswitch> ops in F<pp.c>; using these opcodes would
567be much faster."
568
569=head2 What hooks would assertions need?
570
571Assertions are in the core, and work. However, assertions needed to be added
572as a core patch, rather than an XS module in ext, or a CPAN module, because
573the core has no hooks in the necessary places. It would be useful to
574investigate what hooks would need to be added to make it possible to provide
575the full assertion support from a CPAN module, so that we aren't constraining
576the imagination of future CPAN authors.
577
578
579
580
581
582
583
584=head1 Big projects
585
586Tasks that will get your name mentioned in the description of the "Highlights
587of 5.10"
588
589=head2 make ithreads more robust
590
591Generally make ithreads more robust. See also L<iCOW>
592
593This task is incremental - even a little bit of work on it will help, and
594will be greatly appreciated.
595
596=head2 iCOW
597
598Sarathy and Arthur have a proposal for an improved Copy On Write which
599specifically will be able to COW new ithreads. If this can be implemented
600it would be a good thing.
601
602=head2 (?{...}) closures in regexps
603
604Fix (or rewrite) the implementation of the C</(?{...})/> closures.
605
606=head2 A re-entrant regexp engine
607
608This will allow the use of a regex from inside (?{ }), (??{ }) and
609(?(?{ })|) constructs.