1 |
=head1 NAME |
2 |
|
3 |
BDB - Asynchronous Berkeley DB access |
4 |
|
5 |
=head1 SYNOPSIS |
6 |
|
7 |
use BDB; |
8 |
|
9 |
my $env = db_env_create; |
10 |
|
11 |
mkdir "bdtest", 0700; |
12 |
db_env_open |
13 |
$env, |
14 |
"bdtest", |
15 |
BDB::INIT_LOCK | BDB::INIT_LOG | BDB::INIT_MPOOL |
16 |
| BDB::INIT_TXN | BDB::RECOVER | BDB::USE_ENVIRON | BDB::CREATE, |
17 |
0600; |
18 |
|
19 |
$env->set_flags (BDB::AUTO_COMMIT | BDB::TXN_NOSYNC, 1); |
20 |
|
21 |
my $db = db_create $env; |
22 |
db_open $db, undef, "table", undef, BDB::BTREE, BDB::AUTO_COMMIT | BDB::CREATE |
23 |
| BDB::READ_UNCOMMITTED, 0600; |
24 |
db_put $db, undef, "key", "data", 0, sub { |
25 |
db_del $db, undef, "key"; |
26 |
}; |
27 |
db_sync $db; |
28 |
|
29 |
# when you also use Coro, management is easy: |
30 |
use Coro::BDB; |
31 |
|
32 |
# automatic event loop intergration with AnyEvent: |
33 |
use AnyEvent::BDB; |
34 |
|
35 |
# automatic result processing with EV: |
36 |
my $WATCHER = EV::io BDB::poll_fileno, EV::READ, \&BDB::poll_cb; |
37 |
|
38 |
# with Glib: |
39 |
add_watch Glib::IO BDB::poll_fileno, |
40 |
in => sub { BDB::poll_cb; 1 }; |
41 |
|
42 |
# or simply flush manually |
43 |
BDB::flush; |
44 |
|
45 |
|
46 |
=head1 DESCRIPTION |
47 |
|
48 |
See the BerkeleyDB documentation (L<http://www.oracle.com/technology/documentation/berkeley-db/db/index.html>). |
49 |
The BDB API is very similar to the C API (the translation has been very faithful). |
50 |
|
51 |
See also the example sections in the document below and possibly the eg/ |
52 |
subdirectory of the BDB distribution. Last not least see the IO::AIO |
53 |
documentation, as that module uses almost the same asynchronous request |
54 |
model as this module. |
55 |
|
56 |
I know this is woefully inadequate documentation. Send a patch! |
57 |
|
58 |
|
59 |
=head1 REQUEST ANATOMY AND LIFETIME |
60 |
|
61 |
Every request method creates a request. which is a C data structure not |
62 |
directly visible to Perl. |
63 |
|
64 |
During their existance, bdb requests travel through the following states, |
65 |
in order: |
66 |
|
67 |
=over 4 |
68 |
|
69 |
=item ready |
70 |
|
71 |
Immediately after a request is created it is put into the ready state, |
72 |
waiting for a thread to execute it. |
73 |
|
74 |
=item execute |
75 |
|
76 |
A thread has accepted the request for processing and is currently |
77 |
executing it (e.g. blocking in read). |
78 |
|
79 |
=item pending |
80 |
|
81 |
The request has been executed and is waiting for result processing. |
82 |
|
83 |
While request submission and execution is fully asynchronous, result |
84 |
processing is not and relies on the perl interpreter calling C<poll_cb> |
85 |
(or another function with the same effect). |
86 |
|
87 |
=item result |
88 |
|
89 |
The request results are processed synchronously by C<poll_cb>. |
90 |
|
91 |
The C<poll_cb> function will process all outstanding aio requests by |
92 |
calling their callbacks, freeing memory associated with them and managing |
93 |
any groups they are contained in. |
94 |
|
95 |
=item done |
96 |
|
97 |
Request has reached the end of its lifetime and holds no resources anymore |
98 |
(except possibly for the Perl object, but its connection to the actual |
99 |
aio request is severed and calling its methods will either do nothing or |
100 |
result in a runtime error). |
101 |
|
102 |
=back |
103 |
|
104 |
=cut |
105 |
|
106 |
package BDB; |
107 |
|
108 |
no warnings; |
109 |
use strict 'vars'; |
110 |
|
111 |
use base 'Exporter'; |
112 |
|
113 |
BEGIN { |
114 |
our $VERSION = '1.5'; |
115 |
|
116 |
our @BDB_REQ = qw( |
117 |
db_env_open db_env_close db_env_txn_checkpoint db_env_lock_detect |
118 |
db_env_memp_sync db_env_memp_trickle |
119 |
db_open db_close db_compact db_sync db_upgrade |
120 |
db_put db_get db_pget db_del db_key_range |
121 |
db_txn_commit db_txn_abort db_txn_finish |
122 |
db_c_close db_c_count db_c_put db_c_get db_c_pget db_c_del |
123 |
db_sequence_open db_sequence_close |
124 |
db_sequence_get db_sequence_remove |
125 |
); |
126 |
our @EXPORT = (@BDB_REQ, qw(dbreq_pri dbreq_nice db_env_create db_create)); |
127 |
our @EXPORT_OK = qw( |
128 |
poll_fileno poll_cb poll_wait flush |
129 |
min_parallel max_parallel max_idle |
130 |
nreqs nready npending nthreads |
131 |
max_poll_time max_poll_reqs |
132 |
); |
133 |
|
134 |
require XSLoader; |
135 |
XSLoader::load ("BDB", $VERSION); |
136 |
} |
137 |
|
138 |
=head2 WIN32 FILENAMES/DATABASE NAME MESS |
139 |
|
140 |
Perl on Win32 supports only ASCII filenames (the reason is that it abuses |
141 |
an internal flag to store wether a filename is Unicode or ANSI, but that |
142 |
flag is used for somethign else in the perl core, so there is no way to |
143 |
detect wether a filename is ANSI or Unicode-encoded). The BDB module |
144 |
tries to work around this issue by assuming that the filename is an ANSI |
145 |
filename and BDB was built for unicode support. |
146 |
|
147 |
=head2 BERKELEYDB FUNCTIONS |
148 |
|
149 |
All of these are functions. The create functions simply return a new |
150 |
object and never block. All the remaining functions take an optional |
151 |
callback as last argument. If it is missing, then the function will be |
152 |
executed synchronously. In both cases, C<$!> will reflect the return value |
153 |
of the function. |
154 |
|
155 |
BDB functions that cannot block (mostly functions that manipulate |
156 |
settings) are method calls on the relevant objects, so the rule of thumb |
157 |
is: if it's a method, it's not blocking, if it's a function, it takes a |
158 |
callback as last argument. |
159 |
|
160 |
In the following, C<$int> signifies an integer return value, |
161 |
C<octetstring> is a "binary string" (i.e. a perl string with no character |
162 |
indices >255), C<U32> is an unsigned 32 bit integer, C<int> is some |
163 |
integer, C<NV> is a floating point value. |
164 |
|
165 |
The C<SV *> types are generic perl scalars (for input and output of data |
166 |
values), and the C<SV *callback> is the optional callback function to call |
167 |
when the request is completed. |
168 |
|
169 |
The various C<DB_ENV> etc. arguments are handles return by |
170 |
C<db_env_create>, C<db_create>, C<txn_begin> and so on. If they have an |
171 |
appended C<_ornull> this means they are optional and you can pass C<undef> |
172 |
for them, resulting a NULL pointer on the C level. |
173 |
|
174 |
=head3 BDB functions |
175 |
|
176 |
Functions in the BDB namespace, exported by default: |
177 |
|
178 |
$env = db_env_create (U32 env_flags = 0) |
179 |
flags: RPCCLIENT |
180 |
|
181 |
db_env_open (DB_ENV *env, octetstring db_home, U32 open_flags, int mode, SV *callback = &PL_sv_undef) |
182 |
open_flags: INIT_CDB INIT_LOCK INIT_LOG INIT_MPOOL INIT_REP INIT_TXN RECOVER RECOVER_FATAL USE_ENVIRON USE_ENVIRON_ROOT CREATE LOCKDOWN PRIVATE REGISTER SYSTEM_MEM |
183 |
db_env_close (DB_ENV *env, U32 flags = 0, SV *callback = &PL_sv_undef) |
184 |
db_env_txn_checkpoint (DB_ENV *env, U32 kbyte = 0, U32 min = 0, U32 flags = 0, SV *callback = &PL_sv_undef) |
185 |
flags: FORCE |
186 |
db_env_lock_detect (DB_ENV *env, U32 flags = 0, U32 atype = DB_LOCK_DEFAULT, SV *dummy = 0, SV *callback = &PL_sv_undef) |
187 |
atype: LOCK_DEFAULT LOCK_EXPIRE LOCK_MAXLOCKS LOCK_MAXWRITE LOCK_MINLOCKS LOCK_MINWRITE LOCK_OLDEST LOCK_RANDOM LOCK_YOUNGEST |
188 |
db_env_memp_sync (DB_ENV *env, SV *dummy = 0, SV *callback = &PL_sv_undef) |
189 |
db_env_memp_trickle (DB_ENV *env, int percent, SV *dummy = 0, SV *callback = &PL_sv_undef) |
190 |
|
191 |
$db = db_create (DB_ENV *env = 0, U32 flags = 0) |
192 |
flags: XA_CREATE |
193 |
|
194 |
db_open (DB *db, DB_TXN_ornull *txnid, octetstring file, octetstring database, int type, U32 flags, int mode, SV *callback = &PL_sv_undef) |
195 |
flags: AUTO_COMMIT CREATE EXCL MULTIVERSION NOMMAP RDONLY READ_UNCOMMITTED THREAD TRUNCATE |
196 |
db_close (DB *db, U32 flags = 0, SV *callback = &PL_sv_undef) |
197 |
flags: DB_NOSYNC |
198 |
db_upgrade (DB *db, octetstring file, U32 flags = 0, SV *callback = &PL_sv_undef) |
199 |
db_compact (DB *db, DB_TXN_ornull *txn = 0, SV *start = 0, SV *stop = 0, SV *unused1 = 0, U32 flags = DB_FREE_SPACE, SV *unused2 = 0, SV *callback = &PL_sv_undef) |
200 |
flags: FREELIST_ONLY FREE_SPACE |
201 |
db_sync (DB *db, U32 flags = 0, SV *callback = &PL_sv_undef) |
202 |
db_key_range (DB *db, DB_TXN_ornull *txn, SV *key, SV *key_range, U32 flags = 0, SV *callback = &PL_sv_undef) |
203 |
db_put (DB *db, DB_TXN_ornull *txn, SV *key, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
204 |
flags: APPEND NODUPDATA NOOVERWRITE |
205 |
db_get (DB *db, DB_TXN_ornull *txn, SV *key, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
206 |
flags: CONSUME CONSUME_WAIT GET_BOTH SET_RECNO MULTIPLE READ_COMMITTED READ_UNCOMMITTED RMW |
207 |
db_pget (DB *db, DB_TXN_ornull *txn, SV *key, SV *pkey, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
208 |
flags: CONSUME CONSUME_WAIT GET_BOTH SET_RECNO MULTIPLE READ_COMMITTED READ_UNCOMMITTED RMW |
209 |
db_del (DB *db, DB_TXN_ornull *txn, SV *key, U32 flags = 0, SV *callback = &PL_sv_undef) |
210 |
db_txn_commit (DB_TXN *txn, U32 flags = 0, SV *callback = &PL_sv_undef) |
211 |
flags: TXN_NOSYNC TXN_SYNC |
212 |
db_txn_abort (DB_TXN *txn, SV *callback = &PL_sv_undef) |
213 |
|
214 |
db_c_close (DBC *dbc, SV *callback = &PL_sv_undef) |
215 |
db_c_count (DBC *dbc, SV *count, U32 flags = 0, SV *callback = &PL_sv_undef) |
216 |
db_c_put (DBC *dbc, SV *key, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
217 |
flags: AFTER BEFORE CURRENT KEYFIRST KEYLAST NODUPDATA |
218 |
db_c_get (DBC *dbc, SV *key, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
219 |
flags: CURRENT FIRST GET_BOTH GET_BOTH_RANGE GET_RECNO JOIN_ITEM LAST NEXT NEXT_DUP NEXT_NODUP PREV PREV_DUP PREV_NODUP SET SET_RANGE SET_RECNO READ_UNCOMMITTED MULTIPLE MULTIPLE_KEY RMW |
220 |
db_c_pget (DBC *dbc, SV *key, SV *pkey, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
221 |
db_c_del (DBC *dbc, U32 flags = 0, SV *callback = &PL_sv_undef) |
222 |
|
223 |
db_sequence_open (DB_SEQUENCE *seq, DB_TXN_ornull *txnid, SV *key, U32 flags = 0, SV *callback = &PL_sv_undef) |
224 |
flags: CREATE EXCL |
225 |
db_sequence_close (DB_SEQUENCE *seq, U32 flags = 0, SV *callback = &PL_sv_undef) |
226 |
db_sequence_get (DB_SEQUENCE *seq, DB_TXN_ornull *txnid, int delta, SV *seq_value, U32 flags = DB_TXN_NOSYNC, SV *callback = &PL_sv_undef) |
227 |
flags: TXN_NOSYNC |
228 |
db_sequence_remove (DB_SEQUENCE *seq, DB_TXN_ornull *txnid = 0, U32 flags = 0, SV *callback = &PL_sv_undef) |
229 |
flags: TXN_NOSYNC |
230 |
|
231 |
=head4 db_txn_finish (DB_TXN *txn, U32 flags = 0, SV *callback = &PL_sv_undef) |
232 |
|
233 |
This is not actually a Berkeley DB function but a BDB module |
234 |
extension. The background for this exytension is: It is very annoying to |
235 |
have to check every single BDB function for error returns and provide a |
236 |
codepath out of your transaction. While the BDB module still makes this |
237 |
possible, it contains the following extensions: |
238 |
|
239 |
When a transaction-protected function returns any operating system |
240 |
error (errno > 0), BDB will set the C<TXN_DEADLOCK> flag on the |
241 |
transaction. This flag is also set by Berkeley DB functions themselves |
242 |
when an operation fails with LOCK_DEADLOCK, and it causes all further |
243 |
operations on that transaction (including C<db_txn_commit>) to fail. |
244 |
|
245 |
The C<db_txn_finish> request will look at this flag, and, if it is set, |
246 |
will automatically call C<db_txn_abort> (setting errno to C<LOCK_DEADLOCK> |
247 |
if it isn't set to something else yet). If it isn't set, it will call |
248 |
C<db_txn_commit> and return the error normally. |
249 |
|
250 |
How to use this? Easy: just write your transaction normally: |
251 |
|
252 |
my $txn = $db_env->txn_begin; |
253 |
db_get $db, $txn, "key", my $data; |
254 |
db_put $db, $txn, "key", $data + 1 unless $! == BDB::NOTFOUND; |
255 |
db_txn_finish $txn; |
256 |
die "transaction failed" if $!; |
257 |
|
258 |
That is, handle only the expected errors. If something unexpected happens |
259 |
(EIO, LOCK_NOTGRANTED or a deadlock in either db_get or db_put), then the remaining |
260 |
requests (db_put in this case) will simply be skipped (they will fail with |
261 |
LOCK_DEADLOCK) and the transaction will be aborted. |
262 |
|
263 |
You can use the C<< $txn->failed >> method to check wether a transaction |
264 |
has failed in this way and abort further processing (excluding |
265 |
C<db_txn_finish>). |
266 |
|
267 |
=head3 DB_ENV/database environment methods |
268 |
|
269 |
Methods available on DB_ENV/$env handles: |
270 |
|
271 |
DESTROY (DB_ENV_ornull *env) |
272 |
CODE: |
273 |
if (env) |
274 |
env->close (env, 0); |
275 |
|
276 |
$int = $env->set_data_dir (const char *dir) |
277 |
$int = $env->set_tmp_dir (const char *dir) |
278 |
$int = $env->set_lg_dir (const char *dir) |
279 |
$int = $env->set_shm_key (long shm_key) |
280 |
$int = $env->set_cachesize (U32 gbytes, U32 bytes, int ncache = 0) |
281 |
$int = $env->set_flags (U32 flags, int onoff) |
282 |
$env->set_errfile (FILE *errfile = 0) |
283 |
$env->set_msgfile (FILE *msgfile = 0) |
284 |
$int = $env->set_verbose (U32 which, int onoff = 1) |
285 |
$int = $env->set_encrypt (const char *password, U32 flags = 0) |
286 |
$int = $env->set_timeout (NV timeout_seconds, U32 flags = SET_TXN_TIMEOUT) |
287 |
$int = $env->set_mp_max_openfd (int maxopenfd); |
288 |
$int = $env->set_mp_max_write (int maxwrite, int maxwrite_sleep); |
289 |
$int = $env->set_mp_mmapsize (int mmapsize_mb) |
290 |
$int = $env->set_lk_detect (U32 detect = DB_LOCK_DEFAULT) |
291 |
$int = $env->set_lk_max_lockers (U32 max) |
292 |
$int = $env->set_lk_max_locks (U32 max) |
293 |
$int = $env->set_lk_max_objects (U32 max) |
294 |
$int = $env->set_lg_bsize (U32 max) |
295 |
$int = $env->set_lg_max (U32 max) |
296 |
$int = $env->mutex_set_increment (U32 increment) |
297 |
$int = $env->mutex_set_tas_spins (U32 tas_spins) |
298 |
$int = $env->mutex_set_max (U32 max) |
299 |
$int = $env->mutex_set_align (U32 align) |
300 |
|
301 |
$txn = $env->txn_begin (DB_TXN_ornull *parent = 0, U32 flags = 0) |
302 |
flags: READ_COMMITTED READ_UNCOMMITTED TXN_NOSYNC TXN_NOWAIT TXN_SNAPSHOT TXN_SYNC TXN_WAIT TXN_WRITE_NOSYNC |
303 |
|
304 |
=head4 Example: |
305 |
|
306 |
use AnyEvent; |
307 |
use BDB; |
308 |
|
309 |
our $FH; open $FH, "<&=" . BDB::poll_fileno; |
310 |
our $WATCHER = AnyEvent->io (fh => $FH, poll => 'r', cb => \&BDB::poll_cb); |
311 |
|
312 |
BDB::min_parallel 8; |
313 |
|
314 |
my $env = db_env_create; |
315 |
|
316 |
mkdir "bdtest", 0700; |
317 |
db_env_open |
318 |
$env, |
319 |
"bdtest", |
320 |
BDB::INIT_LOCK | BDB::INIT_LOG | BDB::INIT_MPOOL | BDB::INIT_TXN | BDB::RECOVER | BDB::USE_ENVIRON | BDB::CREATE, |
321 |
0600; |
322 |
|
323 |
$env->set_flags (BDB::AUTO_COMMIT | BDB::TXN_NOSYNC, 1); |
324 |
|
325 |
|
326 |
=head3 DB/database methods |
327 |
|
328 |
Methods available on DB/$db handles: |
329 |
|
330 |
DESTROY (DB_ornull *db) |
331 |
CODE: |
332 |
if (db) |
333 |
{ |
334 |
SV *env = (SV *)db->app_private; |
335 |
db->close (db, 0); |
336 |
SvREFCNT_dec (env); |
337 |
} |
338 |
|
339 |
$int = $db->set_cachesize (U32 gbytes, U32 bytes, int ncache = 0) |
340 |
$int = $db->set_flags (U32 flags) |
341 |
flags: CHKSUM ENCRYPT TXN_NOT_DURABLE |
342 |
Btree: DUP DUPSORT RECNUM REVSPLITOFF |
343 |
Hash: DUP DUPSORT |
344 |
Queue: INORDER |
345 |
Recno: RENUMBER SNAPSHOT |
346 |
|
347 |
$int = $db->set_encrypt (const char *password, U32 flags) |
348 |
$int = $db->set_lorder (int lorder) |
349 |
$int = $db->set_bt_minkey (U32 minkey) |
350 |
$int = $db->set_re_delim (int delim) |
351 |
$int = $db->set_re_pad (int re_pad) |
352 |
$int = $db->set_re_source (char *source) |
353 |
$int = $db->set_re_len (U32 re_len) |
354 |
$int = $db->set_h_ffactor (U32 h_ffactor) |
355 |
$int = $db->set_h_nelem (U32 h_nelem) |
356 |
$int = $db->set_q_extentsize (U32 extentsize) |
357 |
|
358 |
$dbc = $db->cursor (DB_TXN_ornull *txn = 0, U32 flags = 0) |
359 |
flags: READ_COMMITTED READ_UNCOMMITTED WRITECURSOR TXN_SNAPSHOT |
360 |
$seq = $db->sequence (U32 flags = 0) |
361 |
|
362 |
=head4 Example: |
363 |
|
364 |
my $db = db_create $env; |
365 |
db_open $db, undef, "table", undef, BDB::BTREE, BDB::AUTO_COMMIT | BDB::CREATE | BDB::READ_UNCOMMITTED, 0600; |
366 |
|
367 |
for (1..1000) { |
368 |
db_put $db, undef, "key $_", "data $_"; |
369 |
|
370 |
db_key_range $db, undef, "key $_", my $keyrange; |
371 |
my ($lt, $eq, $gt) = @$keyrange; |
372 |
} |
373 |
|
374 |
db_del $db, undef, "key $_" for 1..1000; |
375 |
|
376 |
db_sync $db; |
377 |
|
378 |
|
379 |
=head3 DB_TXN/transaction methods |
380 |
|
381 |
Methods available on DB_TXN/$txn handles: |
382 |
|
383 |
DESTROY (DB_TXN_ornull *txn) |
384 |
CODE: |
385 |
if (txn) |
386 |
txn->abort (txn); |
387 |
|
388 |
$int = $txn->set_timeout (NV timeout_seconds, U32 flags = SET_TXN_TIMEOUT) |
389 |
flags: SET_LOCK_TIMEOUT SET_TXN_TIMEOUT |
390 |
|
391 |
$bool = $txn->failed |
392 |
# see db_txn_finish documentation, above |
393 |
|
394 |
|
395 |
=head3 DBC/cursor methods |
396 |
|
397 |
Methods available on DBC/$dbc handles: |
398 |
|
399 |
DESTROY (DBC_ornull *dbc) |
400 |
CODE: |
401 |
if (dbc) |
402 |
dbc->c_close (dbc); |
403 |
|
404 |
$int = $cursor->set_priority ($priority = PRIORITY_*) |
405 |
|
406 |
=head4 Example: |
407 |
|
408 |
my $c = $db->cursor; |
409 |
|
410 |
for (;;) { |
411 |
db_c_get $c, my $key, my $data, BDB::NEXT; |
412 |
warn "<$!,$key,$data>"; |
413 |
last if $!; |
414 |
} |
415 |
|
416 |
db_c_close $c; |
417 |
|
418 |
|
419 |
=head3 DB_SEQUENCE/sequence methods |
420 |
|
421 |
Methods available on DB_SEQUENCE/$seq handles: |
422 |
|
423 |
DESTROY (DB_SEQUENCE_ornull *seq) |
424 |
CODE: |
425 |
if (seq) |
426 |
seq->close (seq, 0); |
427 |
|
428 |
$int = $seq->initial_value (db_seq_t value) |
429 |
$int = $seq->set_cachesize (U32 size) |
430 |
$int = $seq->set_flags (U32 flags) |
431 |
flags: SEQ_DEC SEQ_INC SEQ_WRAP |
432 |
$int = $seq->set_range (db_seq_t min, db_seq_t max) |
433 |
|
434 |
=head4 Example: |
435 |
|
436 |
my $seq = $db->sequence; |
437 |
|
438 |
db_sequence_open $seq, undef, "seq", BDB::CREATE; |
439 |
db_sequence_get $seq, undef, 1, my $value; |
440 |
|
441 |
|
442 |
=head2 SUPPORT FUNCTIONS |
443 |
|
444 |
=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION |
445 |
|
446 |
=over 4 |
447 |
|
448 |
=item $msg = BDB::strerror [$errno] |
449 |
|
450 |
Returns the string corresponding to the given errno value. If no argument |
451 |
is given, use C<$!>. |
452 |
|
453 |
Note that the BDB module also patches the C<$!> variable directly, so you |
454 |
should be able to get a bdb error string by simply stringifying C<$!>. |
455 |
|
456 |
=item $fileno = BDB::poll_fileno |
457 |
|
458 |
Return the I<request result pipe file descriptor>. This filehandle must be |
459 |
polled for reading by some mechanism outside this module (e.g. Event or |
460 |
select, see below or the SYNOPSIS). If the pipe becomes readable you have |
461 |
to call C<poll_cb> to check the results. |
462 |
|
463 |
See C<poll_cb> for an example. |
464 |
|
465 |
=item BDB::poll_cb |
466 |
|
467 |
Process some outstanding events on the result pipe. You have to call this |
468 |
regularly. Returns the number of events processed. Returns immediately |
469 |
when no events are outstanding. The amount of events processed depends on |
470 |
the settings of C<BDB::max_poll_req> and C<BDB::max_poll_time>. |
471 |
|
472 |
If not all requests were processed for whatever reason, the filehandle |
473 |
will still be ready when C<poll_cb> returns. |
474 |
|
475 |
Example: Install an Event watcher that automatically calls |
476 |
BDB::poll_cb with high priority: |
477 |
|
478 |
Event->io (fd => BDB::poll_fileno, |
479 |
poll => 'r', async => 1, |
480 |
cb => \&BDB::poll_cb); |
481 |
|
482 |
=item BDB::max_poll_reqs $nreqs |
483 |
|
484 |
=item BDB::max_poll_time $seconds |
485 |
|
486 |
These set the maximum number of requests (default C<0>, meaning infinity) |
487 |
that are being processed by C<BDB::poll_cb> in one call, respectively |
488 |
the maximum amount of time (default C<0>, meaning infinity) spent in |
489 |
C<BDB::poll_cb> to process requests (more correctly the mininum amount |
490 |
of time C<poll_cb> is allowed to use). |
491 |
|
492 |
Setting C<max_poll_time> to a non-zero value creates an overhead of one |
493 |
syscall per request processed, which is not normally a problem unless your |
494 |
callbacks are really really fast or your OS is really really slow (I am |
495 |
not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead. |
496 |
|
497 |
Setting these is useful if you want to ensure some level of |
498 |
interactiveness when perl is not fast enough to process all requests in |
499 |
time. |
500 |
|
501 |
For interactive programs, values such as C<0.01> to C<0.1> should be fine. |
502 |
|
503 |
Example: Install an EV watcher that automatically calls |
504 |
BDB::poll_cb with low priority, to ensure that other parts of the |
505 |
program get the CPU sometimes even under high load. |
506 |
|
507 |
# try not to spend much more than 0.1s in poll_cb |
508 |
BDB::max_poll_time 0.1; |
509 |
|
510 |
my $bdb_poll = EV::io BDB::poll_fileno, EV::READ, \&BDB::poll_cb); |
511 |
|
512 |
=item BDB::poll_wait |
513 |
|
514 |
If there are any outstanding requests and none of them in the result |
515 |
phase, wait till the result filehandle becomes ready for reading (simply |
516 |
does a C<select> on the filehandle. This is useful if you want to |
517 |
synchronously wait for some requests to finish). |
518 |
|
519 |
See C<nreqs> for an example. |
520 |
|
521 |
=item BDB::poll |
522 |
|
523 |
Waits until some requests have been handled. |
524 |
|
525 |
Returns the number of requests processed, but is otherwise strictly |
526 |
equivalent to: |
527 |
|
528 |
BDB::poll_wait, BDB::poll_cb |
529 |
|
530 |
=item BDB::flush |
531 |
|
532 |
Wait till all outstanding BDB requests have been handled. |
533 |
|
534 |
Strictly equivalent to: |
535 |
|
536 |
BDB::poll_wait, BDB::poll_cb |
537 |
while BDB::nreqs; |
538 |
|
539 |
=back |
540 |
|
541 |
=head3 CONTROLLING THE NUMBER OF THREADS |
542 |
|
543 |
=over 4 |
544 |
|
545 |
=item BDB::min_parallel $nthreads |
546 |
|
547 |
Set the minimum number of BDB threads to C<$nthreads>. The current |
548 |
default is C<8>, which means eight asynchronous operations can execute |
549 |
concurrently at any one time (the number of outstanding requests, |
550 |
however, is unlimited). |
551 |
|
552 |
BDB starts threads only on demand, when an BDB request is queued and |
553 |
no free thread exists. Please note that queueing up a hundred requests can |
554 |
create demand for a hundred threads, even if it turns out that everything |
555 |
is in the cache and could have been processed faster by a single thread. |
556 |
|
557 |
It is recommended to keep the number of threads relatively low, as some |
558 |
Linux kernel versions will scale negatively with the number of threads |
559 |
(higher parallelity => MUCH higher latency). With current Linux 2.6 |
560 |
versions, 4-32 threads should be fine. |
561 |
|
562 |
Under most circumstances you don't need to call this function, as the |
563 |
module selects a default that is suitable for low to moderate load. |
564 |
|
565 |
=item BDB::max_parallel $nthreads |
566 |
|
567 |
Sets the maximum number of BDB threads to C<$nthreads>. If more than the |
568 |
specified number of threads are currently running, this function kills |
569 |
them. This function blocks until the limit is reached. |
570 |
|
571 |
While C<$nthreads> are zero, aio requests get queued but not executed |
572 |
until the number of threads has been increased again. |
573 |
|
574 |
This module automatically runs C<max_parallel 0> at program end, to ensure |
575 |
that all threads are killed and that there are no outstanding requests. |
576 |
|
577 |
Under normal circumstances you don't need to call this function. |
578 |
|
579 |
=item BDB::max_idle $nthreads |
580 |
|
581 |
Limit the number of threads (default: 4) that are allowed to idle (i.e., |
582 |
threads that did not get a request to process within 10 seconds). That |
583 |
means if a thread becomes idle while C<$nthreads> other threads are also |
584 |
idle, it will free its resources and exit. |
585 |
|
586 |
This is useful when you allow a large number of threads (e.g. 100 or 1000) |
587 |
to allow for extremely high load situations, but want to free resources |
588 |
under normal circumstances (1000 threads can easily consume 30MB of RAM). |
589 |
|
590 |
The default is probably ok in most situations, especially if thread |
591 |
creation is fast. If thread creation is very slow on your system you might |
592 |
want to use larger values. |
593 |
|
594 |
=item $oldmaxreqs = BDB::max_outstanding $maxreqs |
595 |
|
596 |
This is a very bad function to use in interactive programs because it |
597 |
blocks, and a bad way to reduce concurrency because it is inexact: Better |
598 |
use an C<aio_group> together with a feed callback. |
599 |
|
600 |
Sets the maximum number of outstanding requests to C<$nreqs>. If you |
601 |
to queue up more than this number of requests, the next call to the |
602 |
C<poll_cb> (and C<poll_some> and other functions calling C<poll_cb>) |
603 |
function will block until the limit is no longer exceeded. |
604 |
|
605 |
The default value is very large, so there is no practical limit on the |
606 |
number of outstanding requests. |
607 |
|
608 |
You can still queue as many requests as you want. Therefore, |
609 |
C<max_oustsanding> is mainly useful in simple scripts (with low values) or |
610 |
as a stop gap to shield against fatal memory overflow (with large values). |
611 |
|
612 |
=item BDB::set_sync_prepare $cb |
613 |
|
614 |
Sets a callback that is called whenever a request is created without an |
615 |
explicit callback. It has to return two code references. The first is used |
616 |
as the request callback, and the second is called to wait until the first |
617 |
callback has been called. The default implementation works like this: |
618 |
|
619 |
sub { |
620 |
my $status; |
621 |
( |
622 |
sub { $status = $! }, |
623 |
sub { BDB::poll while !defined $status; $! = $status }, |
624 |
) |
625 |
} |
626 |
|
627 |
=back |
628 |
|
629 |
=head3 STATISTICAL INFORMATION |
630 |
|
631 |
=over 4 |
632 |
|
633 |
=item BDB::nreqs |
634 |
|
635 |
Returns the number of requests currently in the ready, execute or pending |
636 |
states (i.e. for which their callback has not been invoked yet). |
637 |
|
638 |
Example: wait till there are no outstanding requests anymore: |
639 |
|
640 |
BDB::poll_wait, BDB::poll_cb |
641 |
while BDB::nreqs; |
642 |
|
643 |
=item BDB::nready |
644 |
|
645 |
Returns the number of requests currently in the ready state (not yet |
646 |
executed). |
647 |
|
648 |
=item BDB::npending |
649 |
|
650 |
Returns the number of requests currently in the pending state (executed, |
651 |
but not yet processed by poll_cb). |
652 |
|
653 |
=back |
654 |
|
655 |
=cut |
656 |
|
657 |
set_sync_prepare { |
658 |
my $status; |
659 |
( |
660 |
sub { |
661 |
$status = $!; |
662 |
}, |
663 |
sub { |
664 |
BDB::poll while !defined $status; |
665 |
$! = $status; |
666 |
}, |
667 |
) |
668 |
}; |
669 |
|
670 |
min_parallel 8; |
671 |
|
672 |
END { flush } |
673 |
|
674 |
1; |
675 |
|
676 |
=head2 FORK BEHAVIOUR |
677 |
|
678 |
This module should do "the right thing" when the process using it forks: |
679 |
|
680 |
Before the fork, BDB enters a quiescent state where no requests |
681 |
can be added in other threads and no results will be processed. After |
682 |
the fork the parent simply leaves the quiescent state and continues |
683 |
request/result processing, while the child frees the request/result queue |
684 |
(so that the requests started before the fork will only be handled in the |
685 |
parent). Threads will be started on demand until the limit set in the |
686 |
parent process has been reached again. |
687 |
|
688 |
In short: the parent will, after a short pause, continue as if fork had |
689 |
not been called, while the child will act as if BDB has not been used |
690 |
yet. |
691 |
|
692 |
Win32 note: there is no fork on win32, and perls emulation of it is too |
693 |
broken to be supported, so do not use BDB in a windows pseudo-fork, better |
694 |
yet, switch to a more capable platform. |
695 |
|
696 |
=head2 MEMORY USAGE |
697 |
|
698 |
Per-request usage: |
699 |
|
700 |
Each aio request uses - depending on your architecture - around 100-200 |
701 |
bytes of memory. In addition, stat requests need a stat buffer (possibly |
702 |
a few hundred bytes), readdir requires a result buffer and so on. Perl |
703 |
scalars and other data passed into aio requests will also be locked and |
704 |
will consume memory till the request has entered the done state. |
705 |
|
706 |
This is not awfully much, so queuing lots of requests is not usually a |
707 |
problem. |
708 |
|
709 |
Per-thread usage: |
710 |
|
711 |
In the execution phase, some aio requests require more memory for |
712 |
temporary buffers, and each thread requires a stack and other data |
713 |
structures (usually around 16k-128k, depending on the OS). |
714 |
|
715 |
=head1 KNOWN BUGS |
716 |
|
717 |
Known bugs will be fixed in the next release, except: |
718 |
|
719 |
If you use a transaction in any request, and the request returns |
720 |
with an operating system error or DB_LOCK_NOTGRANTED, the internal |
721 |
TXN_DEADLOCK flag will be set on the transaction. See C<db_txn_finish>, |
722 |
above. |
723 |
|
724 |
=head1 SEE ALSO |
725 |
|
726 |
L<AnyEvent::BDB> (event loop integration), L<Coro::BDB> (more natural |
727 |
syntax), L<IO::AIO> (nice to have). |
728 |
|
729 |
=head1 AUTHOR |
730 |
|
731 |
Marc Lehmann <schmorp@schmorp.de> |
732 |
http://home.schmorp.de/ |
733 |
|
734 |
=cut |
735 |
|