ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/IO-AIO/AIO.pm
(Generate patch)

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.38 by root, Sun Aug 28 10:51:33 2005 UTC vs.
Revision 1.67 by root, Tue Oct 24 02:25:16 2006 UTC

15 15
16 aio_read $fh, 30000, 1024, $buffer, 0, sub { 16 aio_read $fh, 30000, 1024, $buffer, 0, sub {
17 $_[0] > 0 or die "read error: $!"; 17 $_[0] > 0 or die "read error: $!";
18 }; 18 };
19 19
20 # Event 20 # version 2+ has request and group objects
21 use IO::AIO 2;
22
23 my $req = aio_unlink "/tmp/file", sub { };
24 $req->cancel; # cancel request if still in queue
25
26 my $grp = aio_group sub { print "all stats done\n" };
27 add $grp aio_stat "..." for ...;
28
29 # AnyEvent integration
30 open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!";
31 my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb });
32
33 # Event integration
21 Event->io (fd => IO::AIO::poll_fileno, 34 Event->io (fd => IO::AIO::poll_fileno,
22 poll => 'r', 35 poll => 'r',
23 cb => \&IO::AIO::poll_cb); 36 cb => \&IO::AIO::poll_cb);
24 37
25 # Glib/Gtk2 38 # Glib/Gtk2 integration
26 add_watch Glib::IO IO::AIO::poll_fileno, 39 add_watch Glib::IO IO::AIO::poll_fileno,
27 in => sub { IO::AIO::poll_cb; 1 }; 40 in => sub { IO::AIO::poll_cb; 1 };
28 41
29 # Tk 42 # Tk integration
30 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", 43 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
31 readable => \&IO::AIO::poll_cb); 44 readable => \&IO::AIO::poll_cb);
32 45
33 # Danga::Socket 46 # Danga::Socket integration
34 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => 47 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
35 \&IO::AIO::poll_cb); 48 \&IO::AIO::poll_cb);
36 49
37
38=head1 DESCRIPTION 50=head1 DESCRIPTION
39 51
40This module implements asynchronous I/O using whatever means your 52This module implements asynchronous I/O using whatever means your
41operating system supports. 53operating system supports.
42 54
43Currently, a number of threads are started that execute your read/writes 55Currently, a number of threads are started that execute your read/writes
44and signal their completion. You don't need thread support in your libc or 56and signal their completion. You don't need thread support in perl, and
45perl, and the threads created by this module will not be visible to the 57the threads created by this module will not be visible to perl. In the
46pthreads library. In the future, this module might make use of the native 58future, this module might make use of the native aio functions available
47aio functions available on many operating systems. However, they are often 59on many operating systems. However, they are often not well-supported
48not well-supported (Linux doesn't allow them on normal files currently, 60(Linux doesn't allow them on normal files currently, for example),
49for example), and they would only support aio_read and aio_write, so the 61and they would only support aio_read and aio_write, so the remaining
50remaining functionality would have to be implemented using threads anyway. 62functionality would have to be implemented using threads anyway.
51 63
52Although the module will work with in the presence of other threads, it is 64Although the module will work with in the presence of other threads, it is
53currently not reentrant, so use appropriate locking yourself, always call 65currently not reentrant, so use appropriate locking yourself, always call
54C<poll_cb> from within the same thread, or never call C<poll_cb> (or other 66C<poll_cb> from within the same thread, or never call C<poll_cb> (or other
55C<aio_> functions) recursively. 67C<aio_> functions) recursively.
57=cut 69=cut
58 70
59package IO::AIO; 71package IO::AIO;
60 72
61no warnings; 73no warnings;
74use strict 'vars';
62 75
63use base 'Exporter'; 76use base 'Exporter';
64 77
65use Fcntl ();
66
67BEGIN { 78BEGIN {
68 $VERSION = 1.6; 79 our $VERSION = '2.0';
69 80
70 @EXPORT = qw(aio_sendfile aio_read aio_write aio_open aio_close stat 81 our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat
71 aio_aio_lstat aio_unlink aio_rmdir aio_readdir aio_symlink 82 aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink
72 aio_fsync aio_fdatasync aio_readahead); 83 aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move
73 @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel 84 aio_group aio_nop);
74 max_outstanding nreqs); 85 our @EXPORT = (@AIO_REQ, qw(aioreq_pri));
86 our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
87 min_parallel max_parallel max_outstanding nreqs);
88
89 @IO::AIO::GRP::ISA = 'IO::AIO::REQ';
75 90
76 require XSLoader; 91 require XSLoader;
77 XSLoader::load IO::AIO, $VERSION; 92 XSLoader::load ("IO::AIO", $VERSION);
78} 93}
79 94
80=head1 FUNCTIONS 95=head1 FUNCTIONS
81 96
82=head2 AIO FUNCTIONS 97=head2 AIO FUNCTIONS
90syscall has been executed asynchronously. 105syscall has been executed asynchronously.
91 106
92All functions expecting a filehandle keep a copy of the filehandle 107All functions expecting a filehandle keep a copy of the filehandle
93internally until the request has finished. 108internally until the request has finished.
94 109
110All requests return objects of type L<IO::AIO::REQ> that allow further
111manipulation of those requests while they are in-flight.
112
95The pathnames you pass to these routines I<must> be absolute and 113The pathnames you pass to these routines I<must> be absolute and
96encoded in byte form. The reason for the former is that at the time the 114encoded in byte form. The reason for the former is that at the time the
97request is being executed, the current working directory could have 115request is being executed, the current working directory could have
98changed. Alternatively, you can make sure that you never change the 116changed. Alternatively, you can make sure that you never change the
99current working directory. 117current working directory.
105environment, d) use Glib::filename_from_unicode on unicode filenames or e) 123environment, d) use Glib::filename_from_unicode on unicode filenames or e)
106use something else. 124use something else.
107 125
108=over 4 126=over 4
109 127
110=item aio_open $pathname, $flags, $mode, $callback 128=item aio_open $pathname, $flags, $mode, $callback->($fh)
111 129
112Asynchronously open or create a file and call the callback with a newly 130Asynchronously open or create a file and call the callback with a newly
113created filehandle for the file. 131created filehandle for the file.
114 132
115The pathname passed to C<aio_open> must be absolute. See API NOTES, above, 133The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
132 } else { 150 } else {
133 die "open failed: $!\n"; 151 die "open failed: $!\n";
134 } 152 }
135 }; 153 };
136 154
137=item aio_close $fh, $callback 155=item aio_close $fh, $callback->($status)
138 156
139Asynchronously close a file and call the callback with the result 157Asynchronously close a file and call the callback with the result
140code. I<WARNING:> although accepted, you should not pass in a perl 158code. I<WARNING:> although accepted, you should not pass in a perl
141filehandle here, as perl will likely close the file descriptor another 159filehandle here, as perl will likely close the file descriptor another
142time when the filehandle is destroyed. Normally, you can safely call perls 160time when the filehandle is destroyed. Normally, you can safely call perls
143C<close> or just let filehandles go out of scope. 161C<close> or just let filehandles go out of scope.
144 162
145This is supposed to be a bug in the API, so that might change. It's 163This is supposed to be a bug in the API, so that might change. It's
146therefore best to avoid this function. 164therefore best to avoid this function.
147 165
148=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback 166=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
149 167
150=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback 168=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
151 169
152Reads or writes C<length> bytes from the specified C<fh> and C<offset> 170Reads or writes C<length> bytes from the specified C<fh> and C<offset>
153into the scalar given by C<data> and offset C<dataoffset> and calls the 171into the scalar given by C<data> and offset C<dataoffset> and calls the
154callback without the actual number of bytes read (or -1 on error, just 172callback without the actual number of bytes read (or -1 on error, just
155like the syscall). 173like the syscall).
164 aio_read $fh, 7, 15, $buffer, 0, sub { 182 aio_read $fh, 7, 15, $buffer, 0, sub {
165 $_[0] > 0 or die "read error: $!"; 183 $_[0] > 0 or die "read error: $!";
166 print "read $_[0] bytes: <$buffer>\n"; 184 print "read $_[0] bytes: <$buffer>\n";
167 }; 185 };
168 186
187=item aio_move $srcpath, $dstpath, $callback->($status)
188
189[EXPERIMENTAL due to internal aio_group use]
190
191Try to move the I<file> (directories not supported as either source or
192destination) from C<$srcpath> to C<$dstpath> and call the callback with
193the C<0> (error) or C<-1> ok.
194
195This is a composite request that tries to rename(2) the file first. If
196rename files with C<EXDEV>, it creates the destination file with mode 0200
197and copies the contents of the source file into it using C<aio_sendfile>,
198followed by restoring atime, mtime, access mode and uid/gid, in that
199order, and unlinking the C<$srcpath>.
200
201If an error occurs, the partial destination file will be unlinked, if
202possible, except when setting atime, mtime, access mode and uid/gid, where
203errors are being ignored.
204
205=cut
206
207sub aio_move($$$) {
208 my ($src, $dst, $cb) = @_;
209
210 my $grp = aio_group $cb;
211
212 add $grp aio_rename $src, $dst, sub {
213 if ($_[0] && $! == EXDEV) {
214 add $grp aio_open $src, O_RDONLY, 0, sub {
215 if (my $src_fh = $_[0]) {
216 my @stat = stat $src_fh;
217
218 add $grp aio_open $dst, O_WRONLY, 0200, sub {
219 if (my $dst_fh = $_[0]) {
220 add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
221 close $src_fh;
222
223 if ($_[0] == $stat[7]) {
224 utime $stat[8], $stat[9], $dst;
225 chmod $stat[2] & 07777, $dst_fh;
226 chown $stat[4], $stat[5], $dst_fh;
227 close $dst_fh;
228
229 add $grp aio_unlink $src, sub {
230 $grp->result ($_[0]);
231 };
232 } else {
233 my $errno = $!;
234 add $grp aio_unlink $dst, sub {
235 $! = $errno;
236 $grp->result (-1);
237 };
238 }
239 };
240 } else {
241 $grp->result (-1);
242 }
243 },
244
245 } else {
246 $grp->result (-1);
247 }
248 };
249 } else {
250 $grp->result ($_[0]);
251 }
252 };
253
254 $grp
255}
256
169=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback 257=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
170 258
171Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts 259Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
172reading at byte offset C<$in_offset>, and starts writing at the current 260reading at byte offset C<$in_offset>, and starts writing at the current
173file offset of C<$out_fh>. Because of that, it is not safe to issue more 261file offset of C<$out_fh>. Because of that, it is not safe to issue more
174than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each 262than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
187bytes have been read from C<aio_sendfile> alone, as C<aio_sendfile> only 275bytes have been read from C<aio_sendfile> alone, as C<aio_sendfile> only
188provides the number of bytes written to C<$out_fh>. Only if the result 276provides the number of bytes written to C<$out_fh>. Only if the result
189value equals C<$length> one can assume that C<$length> bytes have been 277value equals C<$length> one can assume that C<$length> bytes have been
190read. 278read.
191 279
192=item aio_readahead $fh,$offset,$length, $callback 280=item aio_readahead $fh,$offset,$length, $callback->($retval)
193 281
194C<aio_readahead> populates the page cache with data from a file so that 282C<aio_readahead> populates the page cache with data from a file so that
195subsequent reads from that file will not block on disk I/O. The C<$offset> 283subsequent reads from that file will not block on disk I/O. The C<$offset>
196argument specifies the starting point from which data is to be read and 284argument specifies the starting point from which data is to be read and
197C<$length> specifies the number of bytes to be read. I/O is performed in 285C<$length> specifies the number of bytes to be read. I/O is performed in
201file. The current file offset of the file is left unchanged. 289file. The current file offset of the file is left unchanged.
202 290
203If that syscall doesn't exist (likely if your OS isn't Linux) it will be 291If that syscall doesn't exist (likely if your OS isn't Linux) it will be
204emulated by simply reading the data, which would have a similar effect. 292emulated by simply reading the data, which would have a similar effect.
205 293
206=item aio_stat $fh_or_path, $callback 294=item aio_stat $fh_or_path, $callback->($status)
207 295
208=item aio_lstat $fh, $callback 296=item aio_lstat $fh, $callback->($status)
209 297
210Works like perl's C<stat> or C<lstat> in void context. The callback will 298Works like perl's C<stat> or C<lstat> in void context. The callback will
211be called after the stat and the results will be available using C<stat _> 299be called after the stat and the results will be available using C<stat _>
212or C<-s _> etc... 300or C<-s _> etc...
213 301
223 aio_stat "/etc/passwd", sub { 311 aio_stat "/etc/passwd", sub {
224 $_[0] and die "stat failed: $!"; 312 $_[0] and die "stat failed: $!";
225 print "size is ", -s _, "\n"; 313 print "size is ", -s _, "\n";
226 }; 314 };
227 315
228=item aio_unlink $pathname, $callback 316=item aio_unlink $pathname, $callback->($status)
229 317
230Asynchronously unlink (delete) a file and call the callback with the 318Asynchronously unlink (delete) a file and call the callback with the
231result code. 319result code.
232 320
321=item aio_link $srcpath, $dstpath, $callback->($status)
322
323Asynchronously create a new link to the existing object at C<$srcpath> at
324the path C<$dstpath> and call the callback with the result code.
325
326=item aio_symlink $srcpath, $dstpath, $callback->($status)
327
328Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
329the path C<$dstpath> and call the callback with the result code.
330
331=item aio_rename $srcpath, $dstpath, $callback->($status)
332
333Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
334rename(2) and call the callback with the result code.
335
233=item aio_rmdir $pathname, $callback 336=item aio_rmdir $pathname, $callback->($status)
234 337
235Asynchronously rmdir (delete) a directory and call the callback with the 338Asynchronously rmdir (delete) a directory and call the callback with the
236result code. 339result code.
237 340
238=item aio_readdir $pathname $callback 341=item aio_readdir $pathname, $callback->($entries)
239 342
240Unlike the POSIX call of the same name, C<aio_readdir> reads an entire 343Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
241directory (i.e. opendir + readdir + closedir). The entries will not be 344directory (i.e. opendir + readdir + closedir). The entries will not be
242sorted, and will B<NOT> include the C<.> and C<..> entries. 345sorted, and will B<NOT> include the C<.> and C<..> entries.
243 346
244The callback a single argument which is either C<undef> or an array-ref 347The callback a single argument which is either C<undef> or an array-ref
245with the filenames. 348with the filenames.
246 349
350=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)
351
352[EXPERIMENTAL due to internal aio_group use]
353
354Scans a directory (similar to C<aio_readdir>) but additionally tries to
355separate the entries of directory C<$path> into two sets of names, ones
356you can recurse into (directories or links to them), and ones you cannot
357recurse into (everything else).
358
359C<aio_scandir> is a composite request that creates of many sub requests_
360C<$maxreq> specifies the maximum number of outstanding aio requests that
361this function generates. If it is C<< <= 0 >>, then a suitable default
362will be chosen (currently 6).
363
364On error, the callback is called without arguments, otherwise it receives
365two array-refs with path-relative entry names.
366
367Example:
368
369 aio_scandir $dir, 0, sub {
370 my ($dirs, $nondirs) = @_;
371 print "real directories: @$dirs\n";
372 print "everything else: @$nondirs\n";
373 };
374
375Implementation notes.
376
377The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
378
379After reading the directory, the modification time, size etc. of the
380directory before and after the readdir is checked, and if they match (and
381isn't the current time), the link count will be used to decide how many
382entries are directories (if >= 2). Otherwise, no knowledge of the number
383of subdirectories will be assumed.
384
385Then entries will be sorted into likely directories (everything without
386a non-initial dot currently) and likely non-directories (everything
387else). Then every entry plus an appended C</.> will be C<stat>'ed,
388likely directories first. If that succeeds, it assumes that the entry
389is a directory or a symlink to directory (which will be checked
390seperately). This is often faster than stat'ing the entry itself because
391filesystems might detect the type of the entry without reading the inode
392data (e.g. ext2fs filetype feature).
393
394If the known number of directories (link count - 2) has been reached, the
395rest of the entries is assumed to be non-directories.
396
397This only works with certainty on POSIX (= UNIX) filesystems, which
398fortunately are the vast majority of filesystems around.
399
400It will also likely work on non-POSIX filesystems with reduced efficiency
401as those tend to return 0 or 1 as link counts, which disables the
402directory counting heuristic.
403
404=cut
405
406sub aio_scandir($$$) {
407 my ($path, $maxreq, $cb) = @_;
408
409 my $grp = aio_group $cb;
410
411 $maxreq = 6 if $maxreq <= 0;
412
413 # stat once
414 add $grp aio_stat $path, sub {
415 return $grp->result () if $_[0];
416 my $now = time;
417 my $hash1 = join ":", (stat _)[0,1,3,7,9];
418
419 # read the directory entries
420 add $grp aio_readdir $path, sub {
421 my $entries = shift
422 or return $grp->result ();
423
424 # stat the dir another time
425 add $grp aio_stat $path, sub {
426 my $hash2 = join ":", (stat _)[0,1,3,7,9];
427
428 my $ndirs;
429
430 # take the slow route if anything looks fishy
431 if ($hash1 ne $hash2 or (stat _)[9] == $now) {
432 $ndirs = -1;
433 } else {
434 # if nlink == 2, we are finished
435 # on non-posix-fs's, we rely on nlink < 2
436 $ndirs = (stat _)[3] - 2
437 or return $grp->result ([], $entries);
438 }
439
440 # sort into likely dirs and likely nondirs
441 # dirs == files without ".", short entries first
442 $entries = [map $_->[0],
443 sort { $b->[1] cmp $a->[1] }
444 map [$_, sprintf "%s%04d", (/.\./ ? "1" : "0"), length],
445 @$entries];
446
447 my (@dirs, @nondirs);
448
449 my ($statcb, $schedcb);
450 my $nreq = 0;
451
452 my $statgrp = add $grp aio_group;
453
454 $schedcb = sub {
455 if (@$entries) {
456 if ($nreq < $maxreq) {
457 my $ent = pop @$entries;
458 $nreq++;
459 add $statgrp aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) };
460 }
461 } elsif (!$nreq) {
462 # finished
463 $statgrp->cancel;
464 undef $statcb;
465 undef $schedcb;
466 $grp->result (\@dirs, \@nondirs);
467 }
468 };
469 $statcb = sub {
470 my ($status, $entry) = @_;
471
472 if ($status < 0) {
473 $nreq--;
474 push @nondirs, $entry;
475 &$schedcb;
476 } else {
477 # need to check for real directory
478 add $grp aio_lstat "$path/$entry", sub {
479 $nreq--;
480
481 if (-d _) {
482 push @dirs, $entry;
483
484 if (!--$ndirs) {
485 push @nondirs, @$entries;
486 $entries = [];
487 }
488 } else {
489 push @nondirs, $entry;
490 }
491
492 &$schedcb;
493 }
494 }
495 };
496
497 &$schedcb while @$entries && $nreq < $maxreq;
498 };
499 };
500 };
501
502 $grp
503}
504
247=item aio_fsync $fh, $callback 505=item aio_fsync $fh, $callback->($status)
248 506
249Asynchronously call fsync on the given filehandle and call the callback 507Asynchronously call fsync on the given filehandle and call the callback
250with the fsync result code. 508with the fsync result code.
251 509
252=item aio_fdatasync $fh, $callback 510=item aio_fdatasync $fh, $callback->($status)
253 511
254Asynchronously call fdatasync on the given filehandle and call the 512Asynchronously call fdatasync on the given filehandle and call the
255callback with the fdatasync result code. 513callback with the fdatasync result code.
256 514
257If this call isn't available because your OS lacks it or it couldn't be 515If this call isn't available because your OS lacks it or it couldn't be
258detected, it will be emulated by calling C<fsync> instead. 516detected, it will be emulated by calling C<fsync> instead.
517
518=item aio_group $callback->(...)
519
520[EXPERIMENTAL]
521
522This is a very special aio request: Instead of doing something, it is a
523container for other aio requests, which is useful if you want to bundle
524many requests into a single, composite, request.
525
526Returns an object of class L<IO::AIO::GRP>. See its documentation below
527for more info.
528
529Example:
530
531 my $grp = aio_group sub {
532 print "all stats done\n";
533 };
534
535 add $grp
536 (aio_stat ...),
537 (aio_stat ...),
538 ...;
539
540=item aio_nop $callback->()
541
542This is a special request - it does nothing in itself and is only used for
543side effects, such as when you want to add a dummy request to a group so
544that finishing the requests in the group depends on executing the given
545code.
546
547While this request does nothing, it still goes through the execution
548phase and still requires a worker thread. Thus, the callback will not
549be executed immediately but only after other requests in the queue have
550entered their execution phase. This can be used to measure request
551latency.
552
553=item IO::AIO::aio_sleep $fractional_seconds, $callback->() *NOT EXPORTED*
554
555Mainly used for debugging and benchmarking, this aio request puts one of
556the request workers to sleep for the given time.
557
558While it is theoretically handy to have simple I/O scheduling requests
559like sleep and file handle readable/writable, the overhead this creates
560is immense, so do not use this function except to put your application
561under artificial I/O pressure.
562
563=back
564
565=head2 IO::AIO::REQ CLASS
566
567All non-aggregate C<aio_*> functions return an object of this class when
568called in non-void context.
569
570A request always moves through the following five states in its lifetime,
571in order: B<ready> (request has been created, but has not been executed
572yet), B<execute> (request is currently being executed), B<pending>
573(request has been executed but callback has not been called yet),
574B<result> (results are being processed synchronously, includes calling the
575callback) and B<done> (request has reached the end of its lifetime and
576holds no resources anymore).
577
578=over 4
579
580=item cancel $req
581
582Cancels the request, if possible. Has the effect of skipping execution
583when entering the B<execute> state and skipping calling the callback when
584entering the the B<result> state, but will leave the request otherwise
585untouched. That means that requests that currently execute will not be
586stopped and resources held by the request will not be freed prematurely.
587
588=item cb $req $callback->(...)
589
590Replace (or simply set) the callback registered to the request.
591
592=back
593
594=head2 IO::AIO::GRP CLASS
595
596This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
597objects of this class, too.
598
599A IO::AIO::GRP object is a special request that can contain multiple other
600aio requests.
601
602You create one by calling the C<aio_group> constructing function with a
603callback that will be called when all contained requests have entered the
604C<done> state:
605
606 my $grp = aio_group sub {
607 print "all requests are done\n";
608 };
609
610You add requests by calling the C<add> method with one or more
611C<IO::AIO::REQ> objects:
612
613 $grp->add (aio_unlink "...");
614
615 add $grp aio_stat "...", sub {
616 $_[0] or return $grp->result ("error");
617
618 # add another request dynamically, if first succeeded
619 add $grp aio_open "...", sub {
620 $grp->result ("ok");
621 };
622 };
623
624This makes it very easy to create composite requests (see the source of
625C<aio_move> for an application) that work and feel like simple requests.
626
627=over 4
628
629=item * The IO::AIO::GRP objects will be cleaned up during calls to
630C<IO::AIO::poll_cb>, just like any other request.
631
632=item * They can be canceled like any other request. Canceling will cancel not
633only the request itself, but also all requests it contains.
634
635=item * They can also can also be added to other IO::AIO::GRP objects.
636
637=item * You must not add requests to a group from within the group callback (or
638any later time).
639
640=item * This does not harmonise well with C<max_outstanding>, so best do
641not combine C<aio_group> with it. Groups and feeders are recommended for
642this kind of concurrency-limiting.
643
644=back
645
646Their lifetime, simplified, looks like this: when they are empty, they
647will finish very quickly. If they contain only requests that are in the
648C<done> state, they will also finish. Otherwise they will continue to
649exist.
650
651That means after creating a group you have some time to add requests. And
652in the callbacks of those requests, you can add further requests to the
653group. And only when all those requests have finished will the the group
654itself finish.
655
656=over 4
657
658=item add $grp ...
659
660=item $grp->add (...)
661
662Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
663be added, including other groups, as long as you do not create circular
664dependencies.
665
666Returns all its arguments.
667
668=item $grp->result (...)
669
670Set the result value(s) that will be passed to the group callback when all
671subrequests have finished. By default, no argument will be passed.
672
673=item feed $grp $callback->($grp)
674
675[VERY EXPERIMENTAL]
676
677Sets a feeder/generator on this group: every group can have an attached
678generator that generates requests if idle. The idea behind this is that,
679although you could just queue as many requests as you want in a group,
680this might starve other requests for a potentially long time. For
681example, C<aio_scandir> might generate hundreds of thousands C<aio_stat>
682requests, delaying any later requests for a long time.
683
684To avoid this, and allow incremental generation of requests, you can
685instead a group and set a feeder on it that generates those requests. The
686feed callback will be called whenever there are few enough (see C<feed_limit>,
687below) requests active in the group itself and is expected to queue more
688requests.
689
690The feed can queue as many requests as it likes (i.e. C<add> does not
691impose any limits).
692
693If the feed does not queue more requests when called, it will be
694automatically removed from the group.
695
696If the feed limit is C<0>, it will be set to C<2> automatically.
697
698Example:
699
700 # stat all files in @files, but only ever use four aio requests concurrently:
701
702 my $grp = aio_group sub { print "finished\n" };
703 feed_limit $grp 4;
704 feed $grp sub {
705 my $file = pop @files
706 or return;
707
708 add $grp aio_stat $file, sub { ... };
709 };
710
711=item feed_limit $grp $num
712
713Sets the feeder limit for the group: The feeder will be called whenever
714the group contains less than this many requests.
715
716Setting the limit to C<0> will pause the feeding process.
259 717
260=back 718=back
261 719
262=head2 SUPPORT FUNCTIONS 720=head2 SUPPORT FUNCTIONS
263 721
321 IO::AIO::poll_wait, IO::AIO::poll_cb 779 IO::AIO::poll_wait, IO::AIO::poll_cb
322 if IO::AIO::nreqs; 780 if IO::AIO::nreqs;
323 781
324=item IO::AIO::min_parallel $nthreads 782=item IO::AIO::min_parallel $nthreads
325 783
326Set the minimum number of AIO threads to C<$nthreads>. The current default 784Set the minimum number of AIO threads to C<$nthreads>. The current
327is C<4>, which means four asynchronous operations can be done at one time 785default is C<8>, which means eight asynchronous operations can execute
328(the number of outstanding operations, however, is unlimited). 786concurrently at any one time (the number of outstanding requests,
787however, is unlimited).
329 788
330IO::AIO starts threads only on demand, when an AIO request is queued and 789IO::AIO starts threads only on demand, when an AIO request is queued and
331no free thread exists. 790no free thread exists.
332 791
333It is recommended to keep the number of threads low, as some Linux 792It is recommended to keep the number of threads relatively low, as some
334kernel versions will scale negatively with the number of threads (higher 793Linux kernel versions will scale negatively with the number of threads
335parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32 794(higher parallelity => MUCH higher latency). With current Linux 2.6
336threads should be fine. 795versions, 4-32 threads should be fine.
337 796
338Under most circumstances you don't need to call this function, as the 797Under most circumstances you don't need to call this function, as the
339module selects a default that is suitable for low to moderate load. 798module selects a default that is suitable for low to moderate load.
340 799
341=item IO::AIO::max_parallel $nthreads 800=item IO::AIO::max_parallel $nthreads
352 811
353Under normal circumstances you don't need to call this function. 812Under normal circumstances you don't need to call this function.
354 813
355=item $oldnreqs = IO::AIO::max_outstanding $nreqs 814=item $oldnreqs = IO::AIO::max_outstanding $nreqs
356 815
816[DEPRECATED]
817
357Sets the maximum number of outstanding requests to C<$nreqs>. If you 818Sets the maximum number of outstanding requests to C<$nreqs>. If you
358try to queue up more than this number of requests, the caller will block until 819try to queue up more than this number of requests, the caller will block until
359some requests have been handled. 820some requests have been handled.
360 821
361The default is very large, so normally there is no practical limit. If you 822The default is very large, so normally there is no practical limit. If you
362queue up many requests in a loop it often improves speed if you set 823queue up many requests in a loop it often improves speed if you set
363this to a relatively low number, such as C<100>. 824this to a relatively low number, such as C<100>.
825
826This function does not work well together with C<aio_group>'s, and their
827feeder interface is better suited to limiting concurrency, so do not use
828this function.
364 829
365Under normal circumstances you don't need to call this function. 830Under normal circumstances you don't need to call this function.
366 831
367=back 832=back
368 833
382 or return undef; 847 or return undef;
383 848
384 *$sym 849 *$sym
385} 850}
386 851
387min_parallel 4; 852min_parallel 8;
388 853
389END { 854END {
390 max_parallel 0; 855 max_parallel 0;
391} 856}
392 857
3931; 8581;
394 859
395=head2 FORK BEHAVIOUR 860=head2 FORK BEHAVIOUR
861
862This module should do "the right thing" when the process using it forks:
396 863
397Before the fork, IO::AIO enters a quiescent state where no requests 864Before the fork, IO::AIO enters a quiescent state where no requests
398can be added in other threads and no results will be processed. After 865can be added in other threads and no results will be processed. After
399the fork the parent simply leaves the quiescent state and continues 866the fork the parent simply leaves the quiescent state and continues
400request/result processing, while the child clears the request/result 867request/result processing, while the child clears the request/result
401queue (so the requests started before the fork will only be handled in 868queue (so the requests started before the fork will only be handled in
402the parent). Threats will be started on demand until the limit ste in the 869the parent). Threads will be started on demand until the limit ste in the
403parent process has been reached again. 870parent process has been reached again.
404 871
872In short: the parent will, after a short pause, continue as if fork had
873not been called, while the child will act as if IO::AIO has not been used
874yet.
875
876=head2 MEMORY USAGE
877
878Each aio request uses - depending on your architecture - around 128 bytes
879of memory. In addition, stat requests need a stat buffer (possibly a few
880hundred bytes). Perl scalars and other data passed into aio requests will
881also be locked.
882
883This is now awfully much, so queuing lots of requests is not usually a
884problem.
885
886Each thread needs a stack area which is usually around 16k, sometimes much
887larger, depending on the OS.
888
405=head1 SEE ALSO 889=head1 SEE ALSO
406 890
407L<Coro>, L<Linux::AIO>. 891L<Coro>, L<Linux::AIO> (obsolete).
408 892
409=head1 AUTHOR 893=head1 AUTHOR
410 894
411 Marc Lehmann <schmorp@schmorp.de> 895 Marc Lehmann <schmorp@schmorp.de>
412 http://home.schmorp.de/ 896 http://home.schmorp.de/

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines