3 perltodo - Perl TO-DO List
7 This is a list of wishes for Perl. The most up to date version of this file
8 is at http://perl5.git.perl.org/perl.git/blob_plain/HEAD:/pod/perltodo.pod
10 The tasks we think are smaller or easier are listed first. Anyone is welcome
11 to work on any of these, but it's a good idea to first contact
12 I<perl5-porters@perl.org> to avoid duplication of effort, and to learn from
13 any previous attempts. By all means contact a pumpking privately first if you
16 Whilst patches to make the list shorter are most welcome, ideas to add to
17 the list are also encouraged. Check the perl5-porters archives for past
18 ideas, and any discussion about them. One set of archives may be found at:
20 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
22 What can we offer you in return? Fame, fortune, and everlasting glory? Maybe
23 not, but if your patch is incorporated, then we'll add your name to the
24 F<AUTHORS> file, which ships in the official distribution. How many other
25 programming languages offer you 1 line of immortality?
27 =head1 Tasks that only need Perl knowledge
29 =head2 Remove duplication of test setup.
31 Schwern notes, that there's duplication of code - lots and lots of tests have
32 some variation on the big block of C<$Is_Foo> checks. We can safely put this
33 into a file, change it to build an C<%Is> hash and require it. Maybe just put
34 it into F<test.pl>. Throw in the handy tainting subroutines.
36 =head2 POD -E<gt> HTML conversion in the core still sucks
38 Which is crazy given just how simple POD purports to be, and how simple HTML
39 can be. It's not actually I<as> simple as it sounds, particularly with the
40 flexibility POD allows for C<=item>, but it would be good to improve the
41 visual appeal of the HTML generated, and to avoid it having any validation
42 errors. See also L</make HTML install work>, as the layout of installation tree
43 is needed to improve the cross-linking.
45 The addition of C<Pod::Simple> and its related modules may make this task
48 =head2 Make ExtUtils::ParseXS use strict;
50 F<lib/ExtUtils/ParseXS.pm> contains this line
52 # use strict; # One of these days...
54 Simply uncomment it, and fix all the resulting issues :-)
56 The more practical approach, to break the task down into manageable chunks, is
57 to work your way though the code from bottom to top, or if necessary adding
58 extra C<{ ... }> blocks, and turning on strict within them.
60 =head2 Parallel testing
62 (This probably impacts much more than the core: also the Test::Harness
63 and TAP::* modules on CPAN.)
65 All of the tests in F<t/> can now be run in parallel, if C<$ENV{TEST_JOBS}>
66 is set. However, tests within each directory in F<ext> and F<lib> are still
67 run in series, with directories run in parallel. This is an adequate
68 heuristic, but it might be possible to relax it further, and get more
69 throughput. Specifically, it would be good to audit all of F<lib/*.t>, and
70 make them use C<File::Temp>.
72 =head2 Make Schwern poorer
74 We should have tests for everything. When all the core's modules are tested,
75 Schwern has promised to donate to $500 to TPF. We may need volunteers to
76 hold him upside down and shake vigorously in order to actually extract the
79 =head2 Improve the coverage of the core tests
81 Use Devel::Cover to ascertain the core modules's test coverage, then add
82 tests that are currently missing.
86 A full test suite for the B module would be nice.
88 =head2 A decent benchmark
90 C<perlbench> seems impervious to any recent changes made to the perl core. It
91 would be useful to have a reasonable general benchmarking suite that roughly
92 represented what current perl programs do, and measurably reported whether
93 tweaks to the core improve, degrade or don't really affect performance, to
94 guide people attempting to optimise the guts of perl. Gisle would welcome
95 new tests for perlbench.
97 =head2 fix tainting bugs
99 Fix the bugs revealed by running the test suite with the C<-t> switch (via
100 C<make test.taintwarn>).
102 =head2 Dual life everything
104 As part of the "dists" plan, anything that doesn't belong in the smallest perl
105 distribution needs to be dual lifed. Anything else can be too. Figure out what
106 changes would be needed to package that module and its tests up for CPAN, and
107 do so. Test it with older perl releases, and fix the problems you find.
109 To make a minimal perl distribution, it's useful to look at
110 F<t/lib/commonsense.t>.
112 =head2 Bundle dual life modules in ext/
114 For maintenance (and branch merging) reasons, it would be useful to move
115 some architecture-independent dual-life modules from lib/ to ext/, if this
116 has no negative impact on the build of perl itself.
118 =head2 POSIX memory footprint
120 Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
121 various times worked to cut it down. There is probably still fat to cut out -
122 for example POSIX passes Exporter some very memory hungry data structures.
124 =head2 embed.pl/makedef.pl
126 There is a script F<embed.pl> that generates several header files to prefix
127 all of Perl's symbols in a consistent way, to provide some semblance of
128 namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
129 in F<interpvar.h>. Quite a few of the functions and variables
130 are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
131 doesn't understand the C macros, so the rules about which symbols are present
132 when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
133 It would be good to teach C<embed.pl> to understand the conditional
134 compilation, and hence remove the duplication, and the mistakes it has caused.
136 =head2 use strict; and AutoLoad
138 Currently if you write
141 use AutoLoader 'AUTOLOAD';
146 print join (' ', No, strict, here), "!\n";
149 then C<use strict;> isn't in force within the autoloaded subroutines. It would
150 be more consistent (and less surprising) to arrange for all lexical pragmas
151 in force at the __END__ block to be in force within each autoloaded subroutine.
153 There's a similar problem with SelfLoader.
155 =head2 profile installman
157 The F<installman> script is slow. All it is doing text processing, which we're
158 told is something Perl is good at. So it would be nice to know what it is doing
159 that is taking so much CPU, and where possible address it.
162 =head1 Tasks that need a little sysadmin-type knowledge
164 Or if you prefer, tasks that you would learn from, and broaden your skills
167 =head2 make HTML install work
169 There is an C<installhtml> target in the Makefile. It's marked as
170 "experimental". It would be good to get this tested, make it work reliably, and
171 remove the "experimental" tag. This would include
177 Checking that cross linking between various parts of the documentation works.
178 In particular that links work between the modules (files with POD in F<lib/>)
179 and the core documentation (files in F<pod/>)
183 Work out how to split C<perlfunc> into chunks, preferably one per function
184 group, preferably with general case code that could be used elsewhere.
185 Challenges here are correctly identifying the groups of functions that go
186 together, and making the right named external cross-links point to the right
187 page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
188 C<endservent>, two or more C<=items> giving the different parameter lists, such
191 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
192 =item substr EXPR,OFFSET,LENGTH
193 =item substr EXPR,OFFSET
195 and different parameter lists having different meanings. (eg C<select>)
199 =head2 compressed man pages
201 Be able to install them. This would probably need a configure test to see how
202 the system does compressed man pages (same directory/different directory?
203 same filename/different filename), as well as tweaking the F<installman> script
204 to compress as necessary.
206 =head2 Add a code coverage target to the Makefile
208 Make it easy for anyone to run Devel::Cover on the core's tests. The steps
209 to do this manually are roughly
215 do a normal C<Configure>, but include Devel::Cover as a module to install
216 (see F<INSTALL> for how to do this)
224 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
228 Process the resulting Devel::Cover database
232 This just give you the coverage of the F<.pm>s. To also get the C level
239 Additionally tell C<Configure> to use the appropriate C compiler flags for
246 (instead of C<make perl>)
250 After running the tests run C<gcov> to generate all the F<.gcov> files.
251 (Including down in the subdirectories of F<ext/>
255 (From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
256 to get their stats into the cover_db directory.
260 Then process the Devel::Cover database
264 It would be good to add a single switch to C<Configure> to specify that you
265 wanted to perform perl level coverage, and another to specify C level
266 coverage, and have C<Configure> and the F<Makefile> do all the right things
269 =head2 Make Config.pm cope with differences between built and installed perl
271 Quite often vendors ship a perl binary compiled with their (pay-for)
272 compilers. People install a free compiler, such as gcc. To work out how to
273 build extensions, Perl interrogates C<%Config>, so in this situation
274 C<%Config> describes compilers that aren't there, and extension building
275 fails. This forces people into choosing between re-compiling perl themselves
276 using the compiler they have, or only using modules that the vendor ships.
278 It would be good to find a way teach C<Config.pm> about the installation setup,
279 possibly involving probing at install time or later, so that the C<%Config> in
280 a binary distribution better describes the installed machine, when the
281 installed machine differs from the build machine in some significant way.
283 =head2 linker specification files
285 Some platforms mandate that you provide a list of a shared library's external
286 symbols to the linker, so the core already has the infrastructure in place to
287 do this for generating shared perl libraries. My understanding is that the
288 GNU toolchain can accept an optional linker specification file, and restrict
289 visibility just to symbols declared in that file. It would be good to extend
290 F<makedef.pl> to support this format, and to provide a means within
291 C<Configure> to enable it. This would allow Unix users to test that the
292 export list is correct, and to build a perl that does not pollute the global
293 namespace with private symbols.
295 =head2 Cross-compile support
297 Currently C<Configure> understands C<-Dusecrosscompile> option. This option
298 arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
299 assumed then to be copied to TARGET machine and used as a replacement of full
302 This could be done little differently. Namely C<miniperl> should be built for
303 HOST and then full C<perl> with extensions should be compiled for TARGET.
304 This, however, might require extra trickery for %Config: we have one config
305 first for HOST and then another for TARGET. Tools like MakeMaker will be
306 mightily confused. Having around two different types of executables and
307 libraries (HOST and TARGET) makes life interesting for Makefiles and
308 shell (and Perl) scripts. There is $Config{run}, normally empty, which
309 can be used as an execution wrapper. Also note that in some
310 cross-compilation/execution environments the HOST and the TARGET do
311 not see the same filesystem(s), the $Config{run} may need to do some
312 file/directory copying back and forth.
316 Make F<pod/roffitall> be updated by F<pod/buildtoc>.
318 =head2 Split "linker" from "compiler"
320 Right now, Configure probes for two commands, and sets two variables:
324 =item * C<cc> (in F<cc.U>)
326 This variable holds the name of a command to execute a C compiler which
327 can resolve multiple global references that happen to have the same
328 name. Usual values are F<cc> and F<gcc>.
329 Fervent ANSI compilers may be called F<c89>. AIX has F<xlc>.
331 =item * C<ld> (in F<dlsrc.U>)
333 This variable indicates the program to be used to link
334 libraries for dynamic loading. On some systems, it is F<ld>.
335 On ELF systems, it should be C<$cc>. Mostly, we'll try to respect
336 the hint file setting.
340 There is an implicit historical assumption from around Perl5.000alpha
341 something, that C<$cc> is also the correct command for linking object files
342 together to make an executable. This may be true on Unix, but it's not true
343 on other platforms, and there are a maze of work arounds in other places (such
344 as F<Makefile.SH>) to cope with this.
346 Ideally, we should create a new variable to hold the name of the executable
347 linker program, probe for it in F<Configure>, and centralise all the special
348 case logic there or in hints files.
350 A small bikeshed issue remains - what to call it, given that C<$ld> is already
351 taken (arguably for the wrong thing now, but on SunOS 4.1 it is the command
352 for creating dynamically-loadable modules) and C<$link> could be confused with
353 the Unix command line executable of the same name, which does something
354 completely different. Andy Dougherty makes the counter argument "In parrot, I
355 tried to call the command used to link object files and libraries into an
356 executable F<link>, since that's what my vaguely-remembered DOS and VMS
357 experience suggested. I don't think any real confusion has ensued, so it's
358 probably a reasonable name for perl5 to use."
360 "Alas, I've always worried that introducing it would make things worse,
361 since now the module building utilities would have to look for
362 C<$Config{link}> and institute a fall-back plan if it weren't found."
363 Although I can see that as confusing, given that C<$Config{d_link}> is true
364 when (hard) links are available.
366 =head2 Configure Windows using PowerShell
368 Currently, Windows uses hard-coded config files based to build the
369 config.h for compiling Perl. Makefiles are also hard-coded and need to be
370 hand edited prior to building Perl. While this makes it easy to create a perl.exe
371 that works across multiple Windows versions, being able to accurately
372 configure a perl.exe for a specific Windows versions and VS C++ would be
373 a nice enhancement. With PowerShell available on Windows XP and up, this
374 may now be possible. Step 1 might be to investigate whether this is possible
375 and use this to clean up our current makefile situation. Step 2 would be to
376 see if there would be a way to use our existing metaconfig units to configure a
377 Windows Perl or whether we go in a separate direction and make it so. Of
378 course, we all know what step 3 is.
380 =head1 Tasks that need a little C knowledge
382 These tasks would need a little C knowledge, but don't need any specific
383 background or experience with XS, or how the Perl interpreter works
385 =head2 Weed out needless PERL_UNUSED_ARG
387 The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
388 unused arguments. Often the arguments can't be removed, as there is an
389 external constraint that determines the prototype of the function, so this
390 approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
391 could be removed. Specifically
397 The prototypes of (nearly all) static functions can be changed
401 Unused arguments generated by short cut macros are wasteful - the short cut
402 macro used can be changed.
406 =head2 Modernize the order of directories in @INC
408 The way @INC is laid out by default, one cannot upgrade core (dual-life)
409 modules without overwriting files. This causes problems for binary
410 package builders. One possible proposal is laid out in this
412 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>.
416 Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
417 On these systems, it might be the default compilation mode, and there
418 is currently no guarantee that passing no use64bitall option to the
419 Configure process will build a 32bit perl. Implementing -Duse32bit*
420 options would be nice for perl 5.12.
422 =head2 Profile Perl - am I hot or not?
424 The Perl source code is stable enough that it makes sense to profile it,
425 identify and optimise the hotspots. It would be good to measure the
426 performance of the Perl interpreter using free tools such as cachegrind,
427 gprof, and dtrace, and work to reduce the bottlenecks they reveal.
429 As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
430 the ops that are most commonly used. The idea is that by grouping them, their
431 object code will be adjacent in the executable, so they have a greater chance
432 of already being in the CPU cache (or swapped in) due to being near another op
435 Except that it's not clear if these really are the most commonly used ops. So
436 as part of exercising your skills with coverage and profiling tools you might
437 want to determine what ops I<really> are the most commonly used. And in turn
438 suggest evictions and promotions to achieve a better F<pp_hot.c>.
440 One piece of Perl code that might make a good testbed is F<installman>.
442 =head2 Allocate OPs from arenas
444 Currently all new OP structures are individually malloc()ed and free()d.
445 All C<malloc> implementations have space overheads, and are now as fast as
446 custom allocates so it would both use less memory and less CPU to allocate
447 the various OP structures from arenas. The SV arena code can probably be
450 Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
451 Perl_Slab_alloc() to pack optrees into a contiguous block, which is
452 probably superior to the use of OP arenas, esp. from a cache locality
453 standpoint. See L<Profile Perl - am I hot or not?>.
455 =head2 Improve win32/wince.c
457 Currently, numerous functions look virtually, if not completely,
458 identical in both C<win32/wince.c> and C<win32/win32.c> files, which can't
461 =head2 Use secure CRT functions when building with VC8 on Win32
463 Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
464 that they were "unsafe" and introduced differently named secure versions of
465 them as replacements, e.g. instead of writing
467 FILE* f = fopen(__FILE__, "r");
472 errno_t err = fopen_s(&f, __FILE__, "r");
474 Currently, the warnings about these deprecations have been disabled by adding
475 -D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
476 warning suppressant and actually make use of the new secure CRT functions.
478 There is also a similar issue with POSIX CRT function names like fileno having
479 been deprecated in favour of ISO C++ conformant names like _fileno. These
480 warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
481 might be nice to do as Microsoft suggest here too, although, unlike the secure
482 functions issue, there is presumably little or no benefit in this case.
484 =head2 Fix POSIX::access() and chdir() on Win32
486 These functions currently take no account of DACLs and therefore do not behave
487 correctly in situations where access is restricted by DACLs (as opposed to the
488 read-only attribute).
490 Furthermore, POSIX::access() behaves differently for directories having the
491 read-only attribute set depending on what CRT library is being used. For
492 example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
493 such directories are not writable, whereas in fact all directories are writable
494 unless access is denied by DACLs. (In the case of directories, the read-only
495 attribute actually only means that the directory cannot be deleted.) This CRT
496 bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
497 not actually be writable if access is indeed denied by DACLs).
499 For the chdir() issue, see ActiveState bug #74552:
500 http://bugs.activestate.com/show_bug.cgi?id=74552
502 Therefore, DACLs should be checked both for consistency across CRTs and for
505 (Note that perl's -w operator should not be modified to check DACLs. It has
506 been written so that it reflects the state of the read-only attribute, even
507 for directories (whatever CRT is being used), for symmetry with chmod().)
509 =head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
511 Maybe create a utility that checks after each libperl.a creation that
512 none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
513 ever creep back to libperl.a.
515 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
517 Note, of course, that this will only tell whether B<your> platform
518 is using those naughty interfaces.
520 =head2 -D_FORTIFY_SOURCE=2, -fstack-protector
522 Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
523 (4.1 onwards?) supports C<-fstack-protector>, both of which give
524 protection against various kinds of buffer overflow problems.
525 These should probably be used for compiling Perl whenever available,
526 Configure and/or hints files should be adjusted to probe for the
527 availability of these features and enable them as appropriate.
529 =head2 Arenas for GPs? For MAGIC?
531 C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
532 It might be a speed or memory saving to change to using arenas. Or it might
533 not. It would need some suitable benchmarking first. In particular, C<GP>s
534 can probably be changed with minimal compatibility impact (probably nothing
535 outside of the core, or even outside of F<gv.c> allocates them), but they
536 probably aren't allocated/deallocated often enough for a speed saving. Whereas
537 C<MAGIC> is allocated/deallocated more often, but in turn, is also something
538 more externally visible, so changing the rules here may bite external code.
542 Several SV body structs are now the same size, notably PVMG and PVGV, PVAV and
543 PVHV, and PVCV and PVFM. It should be possible to allocate and return same
544 sized bodies from the same actual arena, rather than maintaining one arena for
545 each. This could save 4-6K per thread, of memory no longer tied up in the
546 not-yet-allocated part of an arena.
549 =head1 Tasks that need a knowledge of XS
551 These tasks would need C knowledge, and roughly the level of knowledge of
552 the perl API that comes from writing modules that use XS to interface to
555 =head2 Remove the use of SVs as temporaries in dump.c
557 F<dump.c> contains debugging routines to dump out the contains of perl data
558 structures, such as C<SV>s, C<AV>s and C<HV>s. Currently, the dumping code
559 B<uses> C<SV>s for its temporary buffers, which was a logical initial
560 implementation choice, as they provide ready made memory handling.
562 However, they also lead to a lot of confusion when it happens that what you're
563 trying to debug is seen by the code in F<dump.c>, correctly or incorrectly, as
564 a temporary scalar it can use for a temporary buffer. It's also not possible
565 to dump scalars before the interpreter is properly set up, such as during
566 ithreads cloning. It would be good to progressively replace the use of scalars
567 as string accumulation buffers with something much simpler, directly allocated
568 by C<malloc>. The F<dump.c> code is (or should be) only producing 7 bit
569 US-ASCII, so output character sets are not an issue.
571 Producing and proving an internal simple buffer allocation would make it easier
572 to re-write the internals of the PerlIO subsystem to avoid using C<SV>s for
573 B<its> buffers, use of which can cause problems similar to those of F<dump.c>,
576 =head2 safely supporting POSIX SA_SIGINFO
578 Some years ago Jarkko supplied patches to provide support for the POSIX
579 SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
581 Unfortunately, it only works with "unsafe" signals, because under safe
582 signals, by the time Perl gets to run the signal handler, the extra
583 information has been lost. Moreover, it's not easy to store it somewhere,
584 as you can't call mutexs, or do anything else fancy, from inside a signal
587 So it strikes me that we could provide safe SA_SIGINFO support
593 Provide global variables for two file descriptors
597 When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
598 pipe, store the reader in one, the writer in the other
602 In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
603 the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
609 serialise signal number, C<struct siginfo_t> (or at least the parts we care
610 about) into a small auto char buff
614 C<write()> that (non-blocking) to the writer fd
620 if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
621 to the current per-signal-number counts
625 if it writes 0%, assume the pipe is full. Flag the data as lost?
629 if it writes partially, croak a panic, as your OS is broken.
637 in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
638 the pipe", read the data out, deserialise, build the Perl structures on
639 the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
644 I think that this gets us decent C<SA_SIGINFO> support, without the current risk
645 of running Perl code inside the signal handler context. (With all the dangers
646 of things like C<malloc> corruption that that currently offers us)
648 For more information see the thread starting with this message:
649 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html
651 =head2 autovivification
653 Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
655 This task is incremental - even a little bit of work on it will help.
657 =head2 Unicode in Filenames
659 chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
660 opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
661 system, truncate, unlink, utime, -X. All these could potentially accept
662 Unicode filenames either as input or output (and in the case of system
663 and qx Unicode in general, as input or output to/from the shell).
664 Whether a filesystem - an operating system pair understands Unicode in
667 Known combinations that have some level of understanding include
668 Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
669 OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
670 create Unicode filenames, what forms of Unicode are accepted and used
671 (UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
672 and so on, varies. Finding the right level of interfacing to Perl
673 requires some thought. Remember that an OS does not implicate a
676 (The Windows -C command flag "wide API support" has been at least
677 temporarily retired in 5.8.1, and the -C has been repurposed, see
680 Most probably the right way to do this would be this:
681 L</"Virtualize operating system access">.
683 =head2 Unicode in %ENV
685 Currently the %ENV entries are always byte strings.
686 See L</"Virtualize operating system access">.
688 =head2 Unicode and glob()
690 Currently glob patterns and filenames returned from File::Glob::glob()
691 are always byte strings. See L</"Virtualize operating system access">.
693 =head2 Unicode and lc/uc operators
695 Some built-in operators (C<lc>, C<uc>, etc.) behave differently, based on
696 what the internal encoding of their argument is. That should not be the
697 case. Maybe add a pragma to switch behaviour.
699 =head2 use less 'memory'
701 Investigate trade offs to switch out perl's choices on memory usage.
702 Particularly perl should be able to give memory back.
704 This task is incremental - even a little bit of work on it will help.
706 =head2 Re-implement C<:unique> in a way that is actually thread-safe
708 The old implementation made bad assumptions on several levels. A good 90%
709 solution might be just to make C<:unique> work to share the string buffer
710 of SvPVs. That way large constant strings can be shared between ithreads,
711 such as the configuration information in F<Config>.
713 =head2 Make tainting consistent
715 Tainting would be easier to use if it didn't take documented shortcuts and
716 allow taint to "leak" everywhere within an expression.
718 =head2 readpipe(LIST)
720 system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
721 running a shell. readpipe() (the function behind qx//) could be similarly
724 =head2 Audit the code for destruction ordering assumptions
728 /* Need to check SvMAGICAL, as during global destruction it may be that
729 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
730 is now part of the linked list of SV heads, rather than pointing to
731 the original body. */
732 /* FIXME - audit the code for other bugs like this one. */
734 adding the C<SvMAGICAL> check to
736 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
737 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
739 Go through the core and look for similar assumptions that SVs have particular
740 types, as all bets are off during global destruction.
742 =head2 Extend PerlIO and PerlIO::Scalar
744 PerlIO::Scalar doesn't know how to truncate(). Implementing this
745 would require extending the PerlIO vtable.
747 Similarly the PerlIO vtable doesn't know about formats (write()), or
748 about stat(), or chmod()/chown(), utime(), or flock().
750 (For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
753 PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
754 opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
757 See also L</"Virtualize operating system access">.
759 =head2 -C on the #! line
761 It should be possible to make -C work correctly if found on the #! line,
762 given that all perl command line options are strict ASCII, and -C changes
763 only the interpretation of non-ASCII characters, and not for the script file
764 handle. To make it work needs some investigation of the ordering of function
765 calls during startup, and (by implication) a bit of tweaking of that order.
767 =head2 Organize error messages
769 Perl's diagnostics (error messages, see L<perldiag>) could use
770 reorganizing and formalizing so that each error message has its
771 stable-for-all-eternity unique id, categorized by severity, type, and
772 subsystem. (The error messages would be listed in a datafile outside
773 of the Perl source code, and the source code would only refer to the
774 messages by the id.) This clean-up and regularizing should apply
775 for all croak() messages.
777 This would enable all sorts of things: easier translation/localization
778 of the messages (though please do keep in mind the caveats of
779 L<Locale::Maketext> about too straightforward approaches to
780 translation), filtering by severity, and instead of grepping for a
781 particular error message one could look for a stable error id. (Of
782 course, changing the error messages by default would break all the
783 existing software depending on some particular error message...)
785 This kind of functionality is known as I<message catalogs>. Look for
786 inspiration for example in the catgets() system, possibly even use it
787 if available-- but B<only> if available, all platforms will B<not>
790 For the really pure at heart, consider extending this item to cover
791 also the warning messages (see L<perllexwarn>, C<warnings.pl>).
793 =head1 Tasks that need a knowledge of the interpreter
795 These tasks would need C knowledge, and knowledge of how the interpreter works,
796 or a willingness to learn.
798 =head2 error reporting of [$a ; $b]
800 Using C<;> inside brackets is a syntax error, and we don't propose to change
801 that by giving it any meaning. However, it's not reported very helpfully:
803 $ perl -e '$a = [$b; $c];'
804 syntax error at -e line 1, near "$b;"
805 syntax error at -e line 1, near "$c]"
806 Execution of -e aborted due to compilation errors.
808 It should be possible to hook into the tokeniser or the lexer, so that when a
809 C<;> is parsed where it is not legal as a statement terminator (ie inside
810 C<{}> used as a hashref, C<[]> or C<()>) it issues an error something like
811 I<';' isn't legal inside an expression - if you need multiple statements use a
812 do {...} block>. See the thread starting at
813 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-09/msg00573.html
815 =head2 lexicals used only once
819 $ perl -we '$pie = 42'
820 Name "main::pie" used only once: possible typo at -e line 1.
824 $ perl -we 'my $pie = 42'
826 Logically all lexicals used only once should warn, if the user asks for
827 warnings. An unworked RT ticket (#5087) has been open for almost seven
828 years for this discrepancy.
832 The handling of Unicode is unclean in many places. For example, the regexp
833 engine matches in Unicode semantics whenever the string or the pattern is
834 flagged as UTF-8, but that should not be dependent on an internal storage
835 detail of the string. Likewise, case folding behaviour is dependent on the
836 UTF8 internal flag being on or off.
838 =head2 Properly Unicode safe tokeniser and pads.
840 The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
841 variable names are stored in stashes as raw bytes, without the utf-8 flag
842 set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
843 tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
844 source filters. All this could be fixed.
846 =head2 state variable initialization in list context
848 Currently this is illegal:
850 state ($a, $b) = foo();
852 In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
853 semantics, which is tricky to implement in Perl 5 as currently they produce
854 the same opcode trees. The Perl 6 design is firm, so it would be good to
855 implement the necessary code in Perl 5. There are comments in
856 C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
857 constructions involving state variables.
859 =head2 Implement $value ~~ 0 .. $range
861 It would be nice to extend the syntax of the C<~~> operator to also
862 understand numeric (and maybe alphanumeric) ranges.
864 =head2 A does() built-in
866 Like ref(), only useful. It would call the C<DOES> method on objects; it
867 would also tell whether something can be dereferenced as an
868 array/hash/etc., or used as a regexp, etc.
869 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
871 =head2 Tied filehandles and write() don't mix
873 There is no method on tied filehandles to allow them to be called back by
876 =head2 Attach/detach debugger from running program
878 The old perltodo notes "With C<gdb>, you can attach the debugger to a running
879 program if you pass the process ID. It would be good to do this with the Perl
880 debugger on a running Perl program, although I'm not sure how it would be
881 done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
883 =head2 LVALUE functions for lists
885 The old perltodo notes that lvalue functions don't work for list or hash
886 slices. This would be good to fix.
888 =head2 regexp optimiser optional
890 The regexp optimiser is not optional. It should configurable to be, to allow
891 its performance to be measured, and its bugs to be easily demonstrated.
893 =head2 delete &function
895 Allow to delete functions. One can already undef them, but they're still
898 =head2 C</w> regex modifier
900 That flag would enable to match whole words, and also to interpolate
901 arrays as alternations. With it, C</P/w> would be roughly equivalent to:
903 do { local $"='|'; /\b(?:P)\b/ }
905 See L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
908 =head2 optional optimizer
910 Make the peephole optimizer optional. Currently it performs two tasks as
911 it walks the optree - genuine peephole optimisations, and necessary fixups of
912 ops. It would be good to find an efficient way to switch out the
913 optimisations whilst keeping the fixups.
915 =head2 You WANT *how* many
917 Currently contexts are void, scalar and list. split has a special mechanism in
918 place to pass in the number of return values wanted. It would be useful to
919 have a general mechanism for this, backwards compatible and little speed hit.
920 This would allow proposals such as short circuiting sort to be implemented
923 =head2 lexical aliases
925 Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
927 =head2 entersub XS vs Perl
929 At the moment pp_entersub is huge, and has code to deal with entering both
930 perl and XS subroutines. Subroutine implementations rarely change between
931 perl and XS at run time, so investigate using 2 ops to enter subs (one for
932 XS, one for perl) and swap between if a sub is redefined.
936 Self-ties are currently illegal because they caused too many segfaults. Maybe
937 the causes of these could be tracked down and self-ties on all types
940 =head2 Optimize away @_
942 The old perltodo notes "Look at the "reification" code in C<av.c>".
944 =head2 Virtualize operating system access
946 Implement a set of "vtables" that virtualizes operating system access
947 (open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
948 least these interfaces should take SVs as "name" arguments instead of
949 bare char pointers; probably the most flexible and extensible way
950 would be for the Perl-facing interfaces to accept HVs. The system
951 needs to be per-operating-system and per-file-system
952 hookable/filterable, preferably both from XS and Perl level
953 (L<perlport/"Files and Filesystems"> is good reading at this point,
954 in fact, all of L<perlport> is.)
956 This has actually already been implemented (but only for Win32),
957 take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
958 variants go through a set of "vtables" for operating system access,
959 non-Win32 systems currently go straight for the POSIX/UNIX-style
960 system/library call. Similar system as for Win32 should be
961 implemented for all platforms. The existing Win32 implementation
962 probably does not need to survive alongside this proposed new
963 implementation, the approaches could be merged.
965 What would this give us? One often-asked-for feature this would
966 enable is using Unicode for filenames, and other "names" like %ENV,
967 usernames, hostnames, and so forth.
968 (See L<perlunicode/"When Unicode Does Not Happen">.)
970 But this kind of virtualization would also allow for things like
971 virtual filesystems, virtual networks, and "sandboxes" (though as long
972 as dynamic loading of random object code is allowed, not very safe
973 sandboxes since external code of course know not of Perl's vtables).
974 An example of a smaller "sandbox" is that this feature can be used to
975 implement per-thread working directories: Win32 already does this.
977 See also L</"Extend PerlIO and PerlIO::Scalar">.
979 =head2 Investigate PADTMP hash pessimisation
981 The peephole optimiser converts constants used for hash key lookups to shared
982 hash key scalars. Under ithreads, something is undoing this work.
983 See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html
985 =head2 Store the current pad in the OP slab allocator
988 I hope that I got that "current pad" part correct
990 Currently we leak ops in various cases of parse failure. I suggested that we
991 could solve this by always using the op slab allocator, and walking it to
992 free ops. Dave comments that as some ops are already freed during optree
993 creation one would have to mark which ops are freed, and not double free them
994 when walking the slab. He notes that one problem with this is that for some ops
995 you have to know which pad was current at the time of allocation, which does
996 change. I suggested storing a pointer to the current pad in the memory allocated
997 for the slab, and swapping to a new slab each time the pad changes. Dave thinks
998 that this would work.
1000 =head2 repack the optree
1002 Repacking the optree after execution order is determined could allow
1003 removal of NULL ops, and optimal ordering of OPs with respect to cache-line
1004 filling. The slab allocator could be reused for this purpose. I think that
1005 the best way to do this is to make it an optional step just before the
1006 completed optree is attached to anything else, and to use the slab allocator
1007 unchanged, so that freeing ops is identical whether or not this step runs.
1008 Note that the slab allocator allocates ops downwards in memory, so one would
1009 have to actually "allocate" the ops in reverse-execution order to get them
1010 contiguous in memory in execution order.
1012 See http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html
1014 Note that running this copy, and then freeing all the old location ops would
1015 cause their slabs to be freed, which would eliminate possible memory wastage if
1016 the previous suggestion is implemented, and we swap slabs more frequently.
1018 =head2 eliminate incorrect line numbers in warnings
1026 } elsif ($undef == 0) {
1029 used to produce this output:
1031 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1032 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1034 where the line of the second warning was misreported - it should be line 5.
1035 Rafael fixed this - the problem arose because there was no nextstate OP
1036 between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
1037 reports that the currently executing line is line 4. The solution was to inject
1038 a nextstate OPs for each C<elsif>, although it turned out that the nextstate
1039 OP needed to be a nulled OP, rather than a live nextstate OP, else other line
1040 numbers became misreported. (Jenga!)
1042 The problem is more general than C<elsif> (although the C<elsif> case is the
1043 most common and the most confusing). Ideally this code
1053 would produce this output
1055 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1056 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1058 (rather than lines 4 and 5), but this would seem to require every OP to carry
1059 (at least) line number information.
1061 What might work is to have an optional line number in memory just before the
1062 BASEOP structure, with a flag bit in the op to say whether it's present.
1063 Initially during compile every OP would carry its line number. Then add a late
1064 pass to the optimiser (potentially combined with L</repack the optree>) which
1065 looks at the two ops on every edge of the graph of the execution path. If
1066 the line number changes, flags the destination OP with this information.
1067 Once all paths are traced, replace every op with the flag with a
1068 nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1069 control on to the true op. All ops would then be replaced by variants that
1070 do not store the line number. (Which, logically, why it would work best in
1071 conjunction with L</repack the optree>, as that is already copying/reallocating
1074 (Although I should note that we're not certain that doing this for the general
1077 =head2 optimize tail-calls
1079 Tail-calls present an opportunity for broadly applicable optimization;
1080 anywhere that C<< return foo(...) >> is called, the outer return can
1081 be replaced by a goto, and foo will return directly to the outer
1082 caller, saving (conservatively) 25% of perl's call&return cost, which
1083 is relatively higher than in C. The scheme language is known to do
1084 this heavily. B::Concise provides good insight into where this
1085 optimization is possible, ie anywhere entersub,leavesub op-sequence
1088 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1090 Bottom line on this is probably a new pp_tailcall function which
1091 combines the code in pp_entersub, pp_leavesub. This should probably
1092 be done 1st in XS, and using B::Generate to patch the new OP into the
1097 Tasks that will get your name mentioned in the description of the "Highlights
1100 =head2 make ithreads more robust
1102 Generally make ithreads more robust. See also L</iCOW>
1104 This task is incremental - even a little bit of work on it will help, and
1105 will be greatly appreciated.
1107 One bit would be to write the missing code in sv.c:Perl_dirp_dup.
1109 Fix Perl_sv_dup, et al so that threads can return objects.
1113 Sarathy and Arthur have a proposal for an improved Copy On Write which
1114 specifically will be able to COW new ithreads. If this can be implemented
1115 it would be a good thing.
1117 =head2 (?{...}) closures in regexps
1119 Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1121 =head2 A re-entrant regexp engine
1123 This will allow the use of a regex from inside (?{ }), (??{ }) and
1124 (?(?{ })|) constructs.
1126 =head2 Add class set operations to regexp engine
1128 Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1130 demerphq has this on his todo list, but right at the bottom.