Commit | Line | Data |
---|---|---|
d4a7d7b5 | 1 | #! @PYTHON@ |
583b7e4a MW |
2 | ### |
3 | ### Convert a directory tree of audio files | |
4 | ### | |
5 | ### (c) 2010 Mark Wooding | |
6 | ### | |
7 | ||
8 | ###----- Licensing notice --------------------------------------------------- | |
9 | ### | |
9e3a516f MW |
10 | ### This file is part of the `autoys' audio tools collection. |
11 | ### | |
12 | ### `autoys' is free software; you can redistribute it and/or modify | |
583b7e4a MW |
13 | ### it under the terms of the GNU General Public License as published by |
14 | ### the Free Software Foundation; either version 2 of the License, or | |
15 | ### (at your option) any later version. | |
16 | ### | |
9e3a516f | 17 | ### `autoys' is distributed in the hope that it will be useful, |
583b7e4a MW |
18 | ### but WITHOUT ANY WARRANTY; without even the implied warranty of |
19 | ### MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
20 | ### GNU General Public License for more details. | |
21 | ### | |
22 | ### You should have received a copy of the GNU General Public License | |
9e3a516f | 23 | ### along with `autoys'; if not, write to the Free Software Foundation, |
583b7e4a MW |
24 | ### Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. |
25 | ||
26 | ###-------------------------------------------------------------------------- | |
27 | ### External dependencies. | |
28 | ||
29 | ## Language features. | |
30 | from __future__ import with_statement | |
31 | ||
32 | ## Standard Python libraries. | |
33 | import sys as SYS | |
34 | import os as OS | |
35 | import errno as E | |
36 | import time as T | |
37 | import unicodedata as UD | |
38 | import fnmatch as FN | |
39 | import re as RX | |
40 | import shutil as SH | |
41 | import optparse as OP | |
42 | import threading as TH | |
43 | import shlex as L | |
00beb9e5 | 44 | from math import sqrt, ceil |
583b7e4a MW |
45 | from contextlib import contextmanager |
46 | ||
47 | ## eyeD3 tag fettling. | |
608b936e | 48 | import eyed3 as E3 |
583b7e4a | 49 | |
3bf73acf MW |
50 | ## Gstreamer. |
51 | import gi | |
52 | gi.require_version('GLib', '2.0'); from gi.repository import GLib as G | |
53 | gi.require_version('Gio', '2.0'); from gi.repository import Gio as GIO | |
54 | gi.require_version('Gst', '1.0'); from gi.repository import Gst as GS | |
55 | GS.init([]) | |
583b7e4a MW |
56 | |
57 | ## Python Imaging. | |
58 | from PIL import Image as I | |
59 | ||
60 | ## Python parsing. | |
61 | import pyparsing as P | |
62 | ||
63 | ###-------------------------------------------------------------------------- | |
64 | ### Special initialization. | |
65 | ||
d4a7d7b5 | 66 | VERSION = '@VERSION@' |
583b7e4a MW |
67 | |
68 | ## GLib. | |
69 | G.threads_init() | |
70 | ||
71 | ###-------------------------------------------------------------------------- | |
72 | ### Eyecandy progress reports. | |
73 | ||
74 | def charwidth(s): | |
75 | """ | |
76 | Return the width of S, in characters. | |
77 | ||
78 | Specifically, this is the number of backspace characters required to | |
79 | overprint the string S. If the current encoding for `stdout' appears to be | |
80 | Unicode then do a complicated Unicode thing; otherwise assume that | |
81 | characters take up one cell each. | |
82 | ||
83 | None of this handles tab characters in any kind of useful way. Sorry. | |
84 | """ | |
85 | ||
86 | ## If there's no encoding for stdout then we're doing something stupid. | |
87 | if SYS.stdout.encoding is None: return len(s) | |
88 | ||
89 | ## Turn the string into Unicode so we can hack on it properly. Maybe that | |
90 | ## won't work out, in which case fall back to being stupid. | |
91 | try: u = s.decode(SYS.stdout.encoding) | |
92 | except UnicodeError: return len(s) | |
93 | ||
94 | ## Our main problem is combining characters, but we should also try to | |
95 | ## handle wide (mostly Asian) characters, and zero-width ones. This hack | |
96 | ## is taken mostly from http://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c | |
97 | w = 0 | |
98 | for ch in u: | |
99 | cd = ord(ch) | |
100 | if UD.category(ch) in ['Cf', 'Me', 'Mn'] or \ | |
101 | 0x1160 <= cd <= 0x11ff: pass | |
102 | elif UD.east_asian_width(ch) in ['F', 'W']: w += 2 | |
103 | else: w += 1 | |
104 | ||
105 | ## Done. | |
583b7e4a MW |
106 | return w |
107 | ||
108 | class StatusLine (object): | |
109 | """ | |
110 | Maintains a status line containing ephemeral progress information. | |
111 | ||
112 | The status line isn't especially important, but it keeps interactive users | |
113 | amused. | |
114 | ||
115 | There should be only one status line object in your program; otherwise | |
116 | they'll interfere with each other and get confused. | |
117 | ||
118 | The update algorithm (in `set') is fairly careful to do the right thing | |
119 | with long status `lines', and to work properly in an Emacs `shell' buffer. | |
120 | """ | |
121 | ||
122 | def __init__(me): | |
123 | "Initialize the status line." | |
124 | me._last = '' | |
125 | me._lastlen = 0 | |
126 | me.eyecandyp = OS.isatty(SYS.stdout.fileno()) | |
127 | ||
128 | def set(me, line): | |
129 | """ | |
130 | Set the status line contents to LINE, replacing what was there before. | |
131 | ||
132 | This only produces actual output if stdout is interactive. | |
133 | """ | |
134 | n = len(line) | |
135 | ||
136 | ## Eyecandy update. | |
137 | if me.eyecandyp: | |
583b7e4a MW |
138 | |
139 | ## If the old line was longer, we need to clobber its tail, so work out | |
140 | ## what that involves. | |
141 | if n < me._lastlen: | |
142 | b = charwidth(me._last[n:]) | |
143 | pre = '\b'*b + ' '*b | |
144 | else: | |
145 | pre = '' | |
146 | ||
147 | ## Now figure out the length of the common prefix between what we had | |
148 | ## before and what we have now. This reduces the amount of I/O done, | |
149 | ## which keeps network traffic down on SSH links, and keeps down the | |
150 | ## amount of work slow terminal emulators like Emacs have to do. | |
151 | i = 0 | |
152 | m = min(n, me._lastlen) | |
153 | while i < m and line[i] == me._last[i]: | |
154 | i += 1 | |
155 | ||
156 | ## Actually do the output, all in one syscall. | |
157 | b = charwidth(me._last[i:]) | |
158 | SYS.stdout.write(pre + '\b'*b + line[i:]) | |
583b7e4a MW |
159 | SYS.stdout.flush() |
160 | ||
161 | ## Update our idea of what's gone on. | |
162 | me._lastlen = n | |
163 | me._last = line | |
164 | ||
165 | def clear(me): | |
166 | "Clear the status line. Just like set('')." | |
167 | me.set('') | |
168 | ||
169 | def commit(me, line = None): | |
170 | """ | |
171 | Commit the current status line, and maybe the string LINE. | |
172 | ||
173 | If the current status line is nonempty, then commit it to the transcript. | |
174 | If LINE is not None, then commit that to the transcript too. | |
175 | ||
176 | After all of this, we clear the status line to get back to a clean state. | |
177 | """ | |
178 | if me._last: | |
179 | if me.eyecandyp: | |
180 | SYS.stdout.write('\n') | |
181 | else: | |
182 | SYS.stdout.write(me._last + '\n') | |
183 | if line is not None: | |
184 | SYS.stdout.write(line + '\n') | |
185 | me._lastlen = 0 | |
186 | me._last = '' | |
187 | ||
188 | STATUS = StatusLine() | |
189 | ||
190 | def filestatus(file, status): | |
191 | return '%s%s: %s' % (' '*8, OS.path.basename(file), status) | |
192 | ||
193 | class ProgressEyecandy (object): | |
194 | """ | |
195 | Provide amusement while something big and complicated is happening. | |
196 | ||
197 | This is an abstract class. Subclasses must provide a method `progress' | |
198 | returning a pair (CURRENT, MAX) indicating the current progress through the | |
199 | operation. | |
200 | """ | |
201 | ||
202 | def __init__(me, what, silentp = False): | |
203 | """ | |
204 | Initialize a progress meter. | |
205 | ||
206 | WHAT is a prefix string to be written before the progress eyecandy | |
207 | itself. | |
208 | """ | |
209 | me._what = what | |
210 | me._silentp = silentp | |
211 | me._spinner = 0 | |
212 | me._start = T.time() | |
213 | ||
214 | def _fmt_time(me, t): | |
215 | "Format T as a time, in (maybe hours) minutes and seconds." | |
216 | s, t = t % 60, int(t/60) | |
217 | m, h = t % 60, int(t/60) | |
218 | if h > 0: | |
219 | return '%d:%02d:%02d' % (h, m, s) | |
220 | else: | |
221 | return '%02d:%02d' % (m, s) | |
222 | ||
223 | def show(me): | |
224 | "Show the current level of progress." | |
225 | ||
226 | ## If we're not showing pointless frippery, don't bother at all. | |
227 | if not STATUS.eyecandyp: | |
228 | return | |
229 | ||
230 | ## Update the spinner index. | |
231 | me._spinner = (me._spinner + 1)%4 | |
232 | ||
233 | ## Fetch the current progress information. Note that we always fetch | |
234 | ## both the current and maximum levels, because both might change if an | |
235 | ## operation revises its idea of how much work needs doing. | |
236 | cur, max = me.progress() | |
237 | ||
238 | ## If we couldn't get progress information, display something vaguely | |
239 | ## amusing anyway. | |
240 | if cur is None or max is None: | |
241 | STATUS.set('%s %c [unknown progress]' % | |
242 | (me._what, r'/-\|'[me._spinner])) | |
243 | return | |
244 | ||
245 | ## Work out -- well, guess -- the time remaining. | |
246 | if cur: | |
247 | t = T.time() | |
00beb9e5 | 248 | eta = me._fmt_time(ceil((t - me._start)*(max - cur)/cur)) |
583b7e4a MW |
249 | else: |
250 | eta = '???' | |
251 | ||
252 | ## Set the status bar. | |
253 | n = 40*cur/max | |
254 | STATUS.set('%s %c [%s%s] %3d%% (%s)' % \ | |
255 | (me._what, | |
256 | r'/-\|'[me._spinner], | |
257 | '='*n, ' '*(40 - n), | |
258 | 100*cur/max, | |
259 | eta)) | |
260 | ||
261 | def done(me, win = True): | |
262 | "Show a completion notice, or a failure if WIN is false." | |
263 | if not win: | |
264 | STATUS.set('%s FAILED!' % me._what) | |
265 | elif not me._silentp: | |
266 | STATUS.set('%s done (%s)' % | |
267 | (me._what, | |
268 | me._fmt_time(T.time() - me._start))) | |
269 | else: | |
270 | return | |
271 | STATUS.commit() | |
272 | ||
273 | ###-------------------------------------------------------------------------- | |
274 | ### Timeout handling. | |
275 | ||
276 | KILLSWITCH = TH.Event() | |
277 | ||
278 | def timeout(t0, t1): | |
279 | T.sleep(t0) | |
280 | KILLSWITCH.set() | |
281 | T.sleep(t1) | |
282 | moan('dying messily due to timeout') | |
283 | OS._exit(3) | |
284 | ||
285 | ###-------------------------------------------------------------------------- | |
286 | ### Parsing utilities. | |
287 | ||
288 | ## Allow hyphens in identifiers. | |
289 | IDCHARS = P.alphanums + '-_' | |
290 | P.Keyword.setDefaultKeywordChars(IDCHARS) | |
291 | ||
292 | ## Some common kinds of tokens. | |
293 | Name = P.Word(IDCHARS) | |
294 | Num = P.Word(P.nums).setParseAction(lambda toks: map(int, toks)) | |
295 | String = P.QuotedString('"', '\\') | |
296 | ||
297 | ## Handy abbreviations for constructed parser elements. | |
298 | def K(k): return P.Keyword(k).suppress() | |
299 | def D(d): return P.Literal(d).suppress() | |
583b7e4a MW |
300 | def R(p): return P.ZeroOrMore(p).setParseAction(lambda s, l, t: [t]) |
301 | O = P.Optional | |
302 | ||
303 | ###-------------------------------------------------------------------------- | |
304 | ### Format identification and conversion. | |
305 | ||
306 | class IdentificationFailure (Exception): | |
307 | pass | |
308 | ||
309 | class FileCategory (object): | |
310 | """ | |
311 | A FileCategory represents a class of files. | |
312 | ||
313 | For example, it's sensible to consider audio, or image files as a | |
314 | category. A file category knows how to recognize member files from | |
315 | MIME content types. | |
316 | """ | |
317 | ||
318 | def __init__(me, name, mime_pats, ident): | |
319 | """ | |
320 | Construct a new category. | |
321 | ||
322 | The PATS are a list of `fnmatch' patterns to be compared with a MIME | |
323 | type. The IDENT is a function which produces an identification object | |
324 | given a file's name and first-guess MIME type. The object is passed to a | |
325 | Format's `check' method to see whether a file needs re-encoding, and to | |
326 | `convert' to assist with the conversion. | |
327 | ||
328 | An identification object must have an attribute `mime' which is a set of | |
329 | possible MIME types accumulated for the object. | |
330 | """ | |
331 | me.name = name | |
332 | me._mime_pats = mime_pats | |
333 | me._ident = ident | |
334 | CATEGORYMAP[name] = me | |
335 | ||
336 | def identify(me, file, mime): | |
337 | """ | |
338 | Attempt to identify FILE, given its apparent MIME type. | |
339 | ||
340 | If identification succeeds, return an identification object which can be | |
341 | used by associated file formats; otherwise return None. | |
342 | """ | |
343 | for p in me._mime_pats: | |
344 | if not FN.fnmatchcase(mime, p): | |
345 | continue | |
346 | try: | |
347 | return me._ident(file, mime) | |
348 | except IdentificationFailure: | |
349 | pass | |
350 | return None | |
351 | ||
352 | class BaseFormat (object): | |
353 | """ | |
354 | A BaseFormat object represents a particular encoding and parameters. | |
355 | ||
356 | The object can verify (the `check' method) whether a particular file | |
357 | matches its requirements, and if necessary (`encode') re-encode a file. | |
358 | ||
359 | Subclasses should define the following methods. | |
360 | ||
361 | check(ID) | |
362 | Answer whether the file identified by ID is acceptable according to | |
363 | the receiver's parameters. | |
364 | ||
365 | convert(MASTER, ID, TARGET) | |
366 | Convert the file MASTER, which has been identified as ID, according | |
367 | to the receiver's parameters, writing the output to TARGET. | |
368 | ||
369 | Subclasses should also provide these attributes. | |
370 | ||
371 | CATEGORY | |
372 | A FileCategory object for the category of files that this format | |
373 | lives within. | |
374 | ||
375 | EXT A file extension to be applied to encoded output files. | |
376 | ||
377 | NAME A user-facing name for the format. | |
378 | ||
379 | PROPS A parser element to parse a property definition. It should produce | |
380 | a pair NAME, VALUE to be stored in a dictionary. | |
381 | ||
382 | Subclasses for different kinds of file may introduce more subclass | |
383 | protocol. | |
384 | """ | |
385 | ||
386 | def fixup(me, path): | |
387 | """Post-encoding fixups.""" | |
388 | pass | |
389 | ||
390 | FORMATMAP = {} | |
391 | CATEGORYMAP = {} | |
392 | ||
393 | def defformat(name, cls): | |
394 | "Define a format NAME using class CLS." | |
395 | if not hasattr(cls, 'NAME'): | |
396 | raise ValueError, 'abstract class' | |
397 | if not hasattr(cls, 'CATEGORY'): | |
398 | raise ValueError, 'no category' | |
399 | FORMATMAP[name] = cls | |
400 | ||
401 | class FormatParser (P.ParserElement): | |
402 | """ | |
403 | Parse a format specifier: | |
404 | ||
405 | format-spec ::= string [format-properties] | |
406 | format-properties ::= `{' format-property (`,' format-property)* `}' | |
407 | ||
408 | The syntax of a format-property is determined by the PROPS attribute on the | |
409 | named format and its superclasses. | |
410 | """ | |
411 | ||
483b52d0 MW |
412 | name = 'format-spec' |
413 | ||
583b7e4a MW |
414 | ## We cache the parser elements we generate to avoid enormous consing. |
415 | CACHE = {} | |
416 | ||
417 | def parseImpl(me, s, loc, actp = True): | |
418 | ||
419 | ## Firstly, determine the format name. | |
420 | loc, r = Name._parse(s, loc, actp) | |
421 | fmt = r[0] | |
422 | ||
423 | ## Look up the format class. | |
424 | try: fcls = FORMATMAP[fmt] | |
425 | except KeyError: | |
426 | raise P.ParseException(s, loc, "Unknown format `%s'" % fmt) | |
427 | ||
428 | ## Fetch the property-list parser from the cache, if possible; else | |
429 | ## construct it. | |
430 | try: | |
431 | pp = me.CACHE[fmt] | |
432 | except KeyError: | |
433 | seen = set() | |
434 | prop = None | |
435 | for c in fcls.mro(): | |
436 | try: p = c.PROPS | |
437 | except AttributeError: continue | |
438 | if p in seen: continue | |
439 | if prop is None: prop = p | |
440 | else: prop |= p | |
441 | seen.add(p) | |
442 | if prop is None: | |
443 | pp = me.CACHE[fmt] = None | |
444 | else: | |
445 | props = P.delimitedList(prop) | |
446 | props.setParseAction(lambda s, l, t: dict(t.asList())) | |
447 | pp = me.CACHE[fmt] = O(D('{') - props - D('}')) | |
448 | ||
449 | ## Parse the properties. | |
450 | if pp is None: | |
451 | pd = {} | |
452 | else: | |
453 | loc, r = pp._parse(s, loc, actp) | |
454 | if r: pd = r[0] | |
455 | else: pd = {} | |
456 | ||
457 | ## Construct the format object and return it. | |
458 | return loc, fcls(**pd) | |
459 | ||
460 | Format = FormatParser() | |
461 | ||
462 | def prop(kw, pval, tag = None): | |
463 | if tag is None: tag = kw | |
464 | if pval is None: | |
465 | p = K(kw) | |
466 | p.setParseAction(lambda s, l, t: (tag, True)) | |
467 | else: | |
468 | p = K(kw) + D('=') + pval | |
469 | p.setParseAction(lambda s, l, t: (tag, t[0])) | |
470 | return p | |
471 | ||
472 | ###-------------------------------------------------------------------------- | |
473 | ### Policies and actions. | |
474 | ||
475 | class Action (object): | |
476 | """ | |
477 | An Action object represents a conversion action to be performed. | |
478 | ||
479 | This class isn't intended to be instantiated directly. It exists to define | |
480 | some protocol common to all Action objects. | |
481 | ||
482 | Action objects have the following attributes. | |
483 | ||
484 | master The name of the master (source) file. | |
485 | ||
486 | target The name of the target (destination) file. | |
487 | ||
488 | PRIORITY The priority of the action, for deciding which of two actions | |
489 | to perform. Higher priorities are more likely to win. | |
490 | ||
491 | Converting an Action to a string describes the action in a simple | |
492 | user-readable manner. The `perform' method actually carries the action | |
493 | out. | |
494 | """ | |
495 | ||
496 | PRIORITY = 0 | |
497 | ||
498 | def __init__(me, master): | |
499 | "Stash the MASTER file name for later." | |
500 | me.master = master | |
501 | ||
502 | def choose(me, him): | |
503 | "Choose either ME or HIM and return one." | |
504 | if him is None or me.PRIORITY > him.PRIORITY: | |
505 | return me | |
506 | else: | |
507 | return him | |
508 | ||
509 | class CopyAction (Action): | |
510 | """ | |
511 | An Action object for simply copying a file. | |
512 | ||
513 | Actually we try to hardlink it first, falling back to a copy later. This | |
514 | is both faster and more efficient with regard to disk space. | |
515 | """ | |
516 | ||
517 | ## Copying is good. Linking is really good, but we can't tell the | |
518 | ## difference at this stage. | |
519 | PRIORITY = 10 | |
520 | ||
521 | def __init__(me, master, targetdir): | |
522 | "Initialize a CopyAction, from MASTER to the TARGETDIR directory." | |
523 | Action.__init__(me, master) | |
524 | me.target = OS.path.join(targetdir, OS.path.basename(master)) | |
525 | ||
526 | def __str__(me): | |
527 | return 'copy/link' | |
528 | ||
529 | def perform(me): | |
530 | "Actually perform a CopyAction." | |
531 | try: | |
532 | STATUS.set(filestatus(me.master, 'link')) | |
533 | OS.link(me.master, me.target) | |
534 | except OSError, err: | |
535 | if err.errno != E.EXDEV: | |
536 | raise | |
537 | STATUS.set(filestatus(me.master, 'copy')) | |
538 | new = me.target + '.new' | |
539 | SH.copyfile(me.master, new) | |
540 | OS.rename(new, me.target) | |
541 | STATUS.commit() | |
542 | ||
543 | class ConvertAction (Action): | |
544 | """ | |
545 | An Action object for converting a file to a given format. | |
546 | ||
547 | Additional attributes: | |
548 | ||
549 | id The identification object for the master file. | |
550 | ||
551 | format The format to which we're meant to conver the master. | |
552 | """ | |
553 | ||
554 | def __init__(me, master, targetdir, id, format): | |
555 | "Initialize a ConvertAction." | |
556 | Action.__init__(me, master) | |
557 | stem, ext = OS.path.splitext(OS.path.basename(master)) | |
558 | me.target = OS.path.join(targetdir, stem + '.' + format.EXT) | |
559 | me.id = id | |
560 | me.format = format | |
561 | ||
562 | def __str__(me): | |
563 | return 'convert to %s' % me.format.NAME | |
564 | ||
565 | def perform(me): | |
566 | "Acually perform a ConvertAction." | |
567 | STATUS.set(filestatus(me.master, me)) | |
568 | me.format.convert(me.master, me.id, me.target) | |
569 | ||
570 | Policy = P.Forward() | |
571 | ||
572 | class FormatPolicy (object): | |
573 | """ | |
574 | A FormatPolicy object represents a set of rules for how to convert files. | |
575 | ||
576 | Given a master file, the FormatPolicy will identify it and return a list of | |
577 | actions to be performed. The methods required of a FormatPolicy are: | |
578 | ||
579 | setcategory(CAT) | |
580 | Store CAT as the policy's category. Check that this is consistent | |
581 | with the policy as stored. | |
582 | ||
583 | actions(MASTER, TARGETDIR, ID, COHORT) | |
584 | Given a MASTER file, identified as ID, a target directory | |
585 | TARGETDIR, and a list COHORT of (FILE, ID) pairs for other files | |
586 | of the same category in the same directory, return a list of | |
587 | actions to be performed to get the target directory into the right | |
588 | form. The list might be empty if the policy object /rejects/ the | |
589 | file. | |
590 | """ | |
591 | ||
592 | class AndPolicy (FormatPolicy): | |
593 | """ | |
594 | A FormatPolicy which does the union of a bunch of other policies. | |
595 | ||
596 | Each subsidiary policy is invoked in turn. The highest-priority action for | |
597 | each target file is returned. | |
598 | """ | |
599 | ||
600 | def __init__(me, policies): | |
601 | me._policies = policies | |
602 | ||
603 | def setcategory(me, cat): | |
604 | me.cat = cat | |
605 | for p in me._policies: | |
606 | p.setcategory(cat) | |
607 | ||
608 | def actions(me, master, targetdir, id, cohort): | |
609 | tmap = {} | |
610 | for p in me._policies: | |
611 | for a in p.actions(master, targetdir, id, cohort): | |
612 | if a.target in tmap: | |
613 | tmap[a.target] = a.choose(tmap.get(a.target)) | |
614 | else: | |
615 | tmap[a.target] = a | |
616 | return tmap.values() | |
617 | ||
618 | And = K('and') - D('{') - R(Policy) - D('}') | |
619 | And.setParseAction(lambda s, l, t: AndPolicy(t[0])) | |
620 | ||
621 | class OrPolicy (FormatPolicy): | |
622 | """ | |
623 | A FormatPolicy which tries other policies and uses the first that accepts. | |
624 | ||
625 | Each subsidiary policy is invoked in turn. If any accepts, the actions it | |
626 | proposes are turned and no further policies are invoked. If none accepts | |
627 | then the file is rejected. | |
628 | """ | |
629 | ||
630 | def __init__(me, policies): | |
631 | me._policies = policies | |
632 | ||
633 | def setcategory(me, cat): | |
634 | me.cat = cat | |
635 | for p in me._policies: | |
636 | p.setcategory(cat) | |
637 | ||
638 | def actions(me, master, targetdir, id, cohort): | |
639 | for p in me._policies: | |
640 | aa = p.actions(master, targetdir, id, cohort) | |
641 | if aa: | |
642 | return aa | |
643 | else: | |
644 | return [] | |
645 | ||
646 | Or = K('or') - D('{') - R(Policy) - D('}') | |
647 | Or.setParseAction(lambda s, l, t: OrPolicy(t[0])) | |
648 | ||
649 | class AcceptPolicy (FormatPolicy): | |
650 | """ | |
651 | A FormatPolicy which copies files in a particular format. | |
652 | ||
653 | If all of the files in a cohort are recognized as being in a particular | |
654 | format (including this one), then accept it with a CopyAction; otherwise | |
655 | reject. | |
656 | """ | |
657 | ||
658 | def __init__(me, format): | |
659 | me._format = format | |
660 | ||
661 | def setcategory(me, cat): | |
662 | if me._format.CATEGORY is not cat: | |
663 | raise ValueError, \ | |
664 | "Accept format `%s' has category `%s', not `%s'" % \ | |
665 | (me._format.__class__.__name__, | |
666 | me._format.CATEGORY.name, cat.name) | |
667 | me.cat = cat | |
668 | ||
669 | def actions(me, master, targetdir, id, cohort): | |
670 | if me._format.check(id) and \ | |
671 | all(me._format.check(cid) for f, cid in cohort): | |
672 | return [CopyAction(master, targetdir)] | |
673 | else: | |
674 | return [] | |
675 | ||
676 | Accept = K('accept') - Format | |
677 | Accept.setParseAction(lambda s, l, t: AcceptPolicy(t[0])) | |
678 | ||
679 | class ConvertPolicy (FormatPolicy): | |
680 | """ | |
681 | A FormatPolicy which copies files in a particular format or converts if | |
682 | necessary. | |
683 | """ | |
684 | def __init__(me, format): | |
685 | me._format = format | |
686 | ||
687 | def setcategory(me, cat): | |
688 | if me._format.CATEGORY is not cat: | |
689 | raise ValueError, \ | |
690 | "Accept format `%s' has category `%s', not `%s'" % \ | |
691 | (me._format.__class__.__name__, | |
692 | me._format.CATEGORY.name, cat.name) | |
693 | me.cat = cat | |
694 | ||
695 | def actions(me, master, targetdir, id, cohort): | |
696 | if me._format.check(id): | |
697 | return [CopyAction(master, targetdir)] | |
698 | else: | |
699 | return [ConvertAction(master, targetdir, id, me._format)] | |
700 | ||
701 | Convert = K('convert') - Format | |
702 | Convert.setParseAction(lambda s, l, t: ConvertPolicy(t[0])) | |
703 | ||
704 | Policy << (And | Or | Accept | Convert) | |
705 | ||
706 | ###-------------------------------------------------------------------------- | |
707 | ### Audio handling, based on GStreamer. | |
708 | ||
709 | def make_element(factory, name = None, **props): | |
710 | "Return a new element from the FACTORY with the given NAME and PROPS." | |
3bf73acf MW |
711 | elt = GS.ElementFactory.make(factory, name) |
712 | if elt is None: raise ValueError, 'failed to make `%s\' element' % factory | |
583b7e4a MW |
713 | elt.set_properties(**props) |
714 | return elt | |
715 | ||
0dbdd41c MW |
716 | def link_elements(elts): |
717 | "Link the elements ELTS together, in order." | |
3bf73acf MW |
718 | e0 = None |
719 | for e1 in elts: | |
720 | if e0 is not None: e0.link(e1) | |
721 | e0 = e1 | |
0dbdd41c | 722 | |
2101727d MW |
723 | def bin_children(bin): |
724 | "Iterate over the (direct) children of a BIN." | |
3bf73acf MW |
725 | iter = bin.iterate_elements() |
726 | while True: | |
727 | rc, elt = iter.next() | |
728 | if rc == GS.IteratorResult.DONE: break | |
729 | elif rc != GS.IteratorResult.OK: | |
730 | raise ValueError, 'iteration failed (%s)' % rc | |
731 | else: yield elt | |
2101727d | 732 | |
583b7e4a MW |
733 | class GStreamerProgressEyecandy (ProgressEyecandy): |
734 | """ | |
735 | Provide amusement while GStreamer is busy doing something. | |
736 | ||
737 | The GStreamerProgressEyecandy object is a context manager. Wrap it round | |
738 | your GStreamer loop to provide progress information for an operation. | |
739 | """ | |
740 | ||
741 | def __init__(me, what, elt, **kw): | |
742 | """ | |
743 | Initialize a progress meter. | |
744 | ||
745 | WHAT is a prefix string to be written before the progress eyecandy | |
746 | itself. ELT is a GStreamer element to interrogate to find the progress | |
747 | information. | |
748 | """ | |
749 | me._elt = elt | |
750 | ProgressEyecandy.__init__(me, what, **kw) | |
751 | ||
752 | def _update(me): | |
753 | "Called by GLib main event loop to update the eyecandy." | |
754 | me.show() | |
755 | return True | |
756 | ||
757 | def _timer(me): | |
758 | """ | |
759 | Update the progress meter. | |
760 | ||
761 | This is called periodically by the GLib main event-processing loop. | |
762 | """ | |
763 | me.show() | |
764 | return True | |
765 | ||
766 | def progress(me): | |
767 | "Return the current progress as a pair (CURRENT, MAX)." | |
768 | ||
769 | ## Fetch the current progress information. We get the duration each | |
770 | ## time, because (particularly with VBR-encoded MP3 inputs) the estimated | |
771 | ## duration can change as we progress. Hopefully it settles down fairly | |
772 | ## soon. | |
3bf73acf MW |
773 | ok, t = me._elt.query_position(GS.Format.TIME) |
774 | if ok: ok, end = me._elt.query_duration(GS.Format.TIME) | |
775 | if ok: return t, end | |
776 | else: return None, None | |
583b7e4a MW |
777 | |
778 | def __enter__(me): | |
779 | "Enter context: attach progress meter display." | |
780 | ||
781 | ## If we're not showing pointless frippery, don't bother at all. | |
782 | if not STATUS.eyecandyp: | |
783 | return | |
784 | ||
785 | ## Update regularly. The pipeline runs asynchronously. | |
786 | me._id = G.timeout_add(200, me._update) | |
787 | ||
788 | def __exit__(me, ty, val, tb): | |
789 | "Leave context: remove display and report completion or failure." | |
790 | ||
791 | ## If we're not showing pointless frippery, there's nothing to remove. | |
792 | if STATUS.eyecandyp: | |
793 | G.source_remove(me._id) | |
794 | ||
795 | ## Report completion anyway. | |
796 | me.done(ty is None) | |
797 | ||
798 | ## As you were. | |
799 | return False | |
800 | ||
801 | class AudioIdentifier (object): | |
802 | """ | |
803 | Analyses and identifies an audio file. | |
804 | ||
805 | Important properties are: | |
806 | ||
807 | cap A capabilities structure describing the audio file data. The most | |
808 | interesting thing in here is probably its name, which is a MIME | |
809 | type describing the data. | |
810 | ||
811 | dcap A capabilities structure describing the decoded audio data. This | |
812 | is of interest during conversion. | |
813 | ||
814 | tags A dictionary containing metadata tags from the file. These are in | |
815 | GStreamer's encoding-independent format. | |
816 | ||
817 | bitrate An approximation to the stream's bitrate, in kilobits per second. | |
818 | This might be slow to work out for some files so it's computed on | |
819 | demand. | |
820 | """ | |
821 | ||
822 | def __init__(me, file, mime): | |
823 | "Initialize the object suitably for identifying FILE." | |
824 | ||
825 | ## Make some initial GStreamer objects. We'll want the pipeline later if | |
826 | ## we need to analyse a poorly tagged MP3 stream, so save it away. | |
827 | me._pipe = GS.Pipeline() | |
828 | me._file = file | |
829 | bus = me._pipe.get_bus() | |
583b7e4a MW |
830 | loop = G.MainLoop() |
831 | ||
832 | ## The basic recognition kit is based around `decodebin'. We must keep | |
833 | ## it happy by giving it sinks for the streams it's found, which it | |
834 | ## announces asynchronously. | |
835 | source = make_element('filesrc', 'file', location = file) | |
836 | decoder = make_element('decodebin', 'decode') | |
837 | sink = make_element('fakesink') | |
838 | def decoder_pad_arrived(elt, pad): | |
3bf73acf | 839 | if pad.get_current_caps()[0].get_name().startswith('audio/'): |
583b7e4a MW |
840 | elt.link_pads(pad.get_name(), sink, 'sink') |
841 | dpaid = decoder.connect('pad-added', decoder_pad_arrived) | |
0bf1fa38 | 842 | for i in [source, decoder, sink]: me._pipe.add(i) |
0dbdd41c | 843 | link_elements([source, decoder]) |
583b7e4a MW |
844 | |
845 | ## Arrange to collect tags from the pipeline's bus as they're reported. | |
583b7e4a MW |
846 | tags = {} |
847 | fail = [] | |
848 | def bus_message(bus, msg): | |
3bf73acf MW |
849 | ty, s = msg.type, msg.get_structure() |
850 | if ty == GS.MessageType.ERROR: | |
851 | fail[:] = (ValueError, s['debug'], None) | |
583b7e4a | 852 | loop.quit() |
3bf73acf MW |
853 | elif ty == GS.MessageType.STATE_CHANGED: |
854 | if s['new-state'] == GS.State.PAUSED and \ | |
583b7e4a MW |
855 | msg.src == me._pipe: |
856 | loop.quit() | |
3bf73acf MW |
857 | elif ty == GS.MessageType.TAG: |
858 | tt = s['taglist'] | |
859 | for i in xrange(tt.n_tags()): | |
860 | t = tt.nth_tag_name(i) | |
861 | if tt.get_tag_size(t) != 1: continue | |
862 | v = tt.get_value_index(t, 0) | |
863 | tags[t] = v | |
583b7e4a MW |
864 | bmid = bus.connect('message', bus_message) |
865 | ||
866 | ## We want to identify the kind of stream this is. (Hmm. The MIME type | |
867 | ## recognizer has already done this work, but GStreamer is probably more | |
868 | ## reliable.) The `decodebin' has a `typefind' element inside which will | |
869 | ## announce the identified media type. All we need to do is find it and | |
870 | ## attach a signal handler. (Note that the handler might be run in the | |
871 | ## thread context of the pipeline element, but Python's GIL will keep | |
872 | ## things from being too awful.) | |
873 | me.cap = None | |
874 | me.dcap = None | |
2101727d | 875 | for e in bin_children(decoder): |
583b7e4a MW |
876 | if e.get_factory().get_name() == 'typefind': |
877 | tfelt = e | |
878 | break | |
879 | else: | |
880 | assert False, 'failed to find typefind element' | |
881 | ||
882 | ## Crank up most of the heavy machinery. The message handler will stop | |
883 | ## the loop when things seem to be sufficiently well underway. | |
3bf73acf MW |
884 | bus.add_signal_watch() |
885 | me._pipe.set_state(GS.State.PAUSED) | |
583b7e4a MW |
886 | loop.run() |
887 | bus.disconnect(bmid) | |
888 | decoder.disconnect(dpaid) | |
3bf73acf | 889 | bus.remove_signal_watch() |
583b7e4a | 890 | if fail: |
3bf73acf | 891 | me._pipe.set_state(GS.State.NULL) |
583b7e4a MW |
892 | raise fail[0], fail[1], fail[2] |
893 | ||
894 | ## Store the collected tags. | |
895 | me.tags = tags | |
896 | ||
897 | ## Gather the capabilities. The `typefind' element knows the input data | |
898 | ## type. The 'decodebin' knows the raw data type. | |
3bf73acf | 899 | me.cap = tfelt.get_static_pad('src').get_allowed_caps()[0] |
583b7e4a | 900 | me.mime = set([mime, me.cap.get_name()]) |
3bf73acf | 901 | me.dcap = sink.get_static_pad('sink').get_allowed_caps()[0] |
583b7e4a MW |
902 | |
903 | ## If we found a plausible bitrate then stash it. Otherwise note that we | |
904 | ## failed. If anybody asks then we'll work it out then. | |
905 | if 'nominal-bitrate' in tags: | |
906 | me._bitrate = tags['nominal-bitrate']/1000 | |
907 | elif 'bitrate' in tags and tags['bitrate'] >= 80000: | |
908 | me._bitrate = tags['bitrate']/1000 | |
909 | else: | |
910 | me._bitrate = None | |
911 | ||
912 | ## The bitrate computation wants the file size. Ideally we'd want the | |
913 | ## total size of the frames' contents, but that seems hard to dredge | |
914 | ## out. If the framing overhead is small, this should be close enough | |
915 | ## for our purposes. | |
916 | me._bytes = OS.stat(file).st_size | |
917 | ||
918 | def __del__(me): | |
919 | "Close the pipeline down so we don't leak file descriptors." | |
3bf73acf | 920 | me._pipe.set_state(GS.State.NULL) |
583b7e4a MW |
921 | |
922 | @property | |
923 | def bitrate(me): | |
924 | """ | |
925 | Return the approximate bit-rate of the input file. | |
926 | ||
927 | This might take a while if we have to work it out the hard way. | |
928 | """ | |
929 | ||
930 | ## If we already know the answer then just return it. | |
931 | if me._bitrate is not None: | |
932 | return me._bitrate | |
933 | ||
934 | ## Make up a new main loop. | |
935 | loop = G.MainLoop() | |
936 | ||
937 | ## Watch for bus messages. We'll stop when we reach the end of the | |
938 | ## stream: then we'll have a clear idea of how long the track was. | |
939 | fail = [] | |
940 | def bus_message(bus, msg): | |
3bf73acf MW |
941 | ty, s = msg.type, msg.get_structure() |
942 | if ty == GS.MessageType.ERROR: | |
943 | fail[:] = (ValueError, s['debug'], None) | |
583b7e4a | 944 | loop.quit() |
3bf73acf | 945 | elif ty == GS.MessageType.EOS: |
583b7e4a MW |
946 | loop.quit() |
947 | bus = me._pipe.get_bus() | |
948 | bmid = bus.connect('message', bus_message) | |
949 | ||
950 | ## Get everything moving, and keep the user amused while we work. | |
3bf73acf MW |
951 | bus.add_signal_watch() |
952 | me._pipe.set_state(GS.State.PLAYING) | |
583b7e4a MW |
953 | with GStreamerProgressEyecandy(filestatus(file, 'measure bitrate') % |
954 | me._pipe, | |
955 | silentp = True): | |
956 | loop.run() | |
3bf73acf | 957 | bus.remove_signal_watch() |
583b7e4a MW |
958 | bus.disconnect(bmid) |
959 | if fail: | |
3bf73acf | 960 | me._pipe.set_state(GS.State.NULL) |
583b7e4a MW |
961 | raise fail[0], fail[1], fail[2] |
962 | ||
963 | ## Now we should be able to find out our position accurately and work out | |
964 | ## a bitrate. Cache it in case anybody asks again. | |
3bf73acf MW |
965 | ok, t = pipe.query_position(GS.Format.TIME) |
966 | assert ok, 'failed to discover bitrate' | |
583b7e4a MW |
967 | me._bitrate = int(8*me._bytes*1e6/t) |
968 | ||
969 | ## Done. | |
970 | return me._bitrate | |
971 | ||
972 | class AudioFormat (BaseFormat): | |
973 | """ | |
974 | An AudioFormat is a kind of Format specialized for audio files. | |
975 | ||
976 | Format checks are done on an AudioIdentifier object. | |
977 | """ | |
978 | ||
979 | PROPS = prop('bitrate', Num) | |
980 | ||
981 | ## libmagic reports `application/ogg' for Ogg Vorbis files. We've switched | |
982 | ## to GIO now, which reports either `audio/ogg' or `audio/x-vorbis+ogg' | |
983 | ## depending on how thorough it's trying to be. Still, it doesn't do any | |
984 | ## harm here; the main risk is picking up Ogg Theora files by accident, and | |
985 | ## we'll probably be able to extract the audio from them anyway. | |
986 | CATEGORY = FileCategory('audio', ['audio/*', 'application/ogg'], | |
987 | AudioIdentifier) | |
988 | ||
989 | def __init__(me, bitrate = None): | |
990 | "Construct an object, requiring an approximate bitrate." | |
991 | me.bitrate = bitrate | |
992 | ||
993 | def check(me, id): | |
994 | """ | |
995 | Return whether the AudioIdentifier ID is suitable for our purposes. | |
996 | ||
997 | Subclasses can either override this method or provide a property | |
998 | `MIMETYPES', which is a list (other thing that implements `__contains__') | |
999 | of GStreamer MIME types matching this format. | |
1000 | """ | |
1001 | return id.mime & me.MIMETYPES and \ | |
1002 | (me.bitrate is None or id.bitrate <= me.bitrate * sqrt(2)) | |
1003 | ||
1004 | def encoder(me): | |
1005 | """ | |
1006 | Constructs a GStreamer element to encode audio input. | |
1007 | ||
1008 | Subclasses can either override this method (or replace `encode' | |
1009 | entirely), or provide a method `encoder_chain' which returns a list of | |
1010 | elements to be linked together in sequence. The first element in the | |
1011 | chain must have a pad named `sink' and the last must have a pad named | |
1012 | `src'. | |
1013 | """ | |
1014 | elts = me.encoder_chain() | |
1015 | bin = GS.Bin() | |
0bf1fa38 | 1016 | for i in elts: bin.add(i) |
0dbdd41c | 1017 | link_elements(elts) |
3bf73acf MW |
1018 | bin.add_pad(GS.GhostPad('sink', elts[0].get_static_pad('sink'))) |
1019 | bin.add_pad(GS.GhostPad('src', elts[-1].get_static_pad('src'))) | |
583b7e4a MW |
1020 | return bin |
1021 | ||
1022 | def convert(me, master, id, target): | |
1023 | """ | |
1024 | Encode audio from MASTER, already identified as ID, writing it to TARGET. | |
1025 | ||
1026 | See `encoder' for subclasses' responsibilities. | |
1027 | """ | |
1028 | ||
1029 | ## Construct the necessary equipment. | |
1030 | pipe = GS.Pipeline() | |
1031 | bus = pipe.get_bus() | |
583b7e4a MW |
1032 | loop = G.MainLoop() |
1033 | ||
1034 | ## Make sure that there isn't anything in the way of our output. We're | |
1035 | ## going to write to a scratch file so that we don't get confused by | |
1036 | ## half-written rubbish left by a crashed program. | |
1037 | new = target + '.new' | |
1038 | try: | |
1039 | OS.unlink(new) | |
1040 | except OSError, err: | |
1041 | if err.errno != E.ENOENT: | |
1042 | raise | |
1043 | ||
1044 | ## Piece together our pipeline. The annoying part is that the | |
1045 | ## `decodebin' doesn't have any source pads yet, so our chain is in two | |
1046 | ## halves for now. | |
1047 | source = make_element('filesrc', 'source', location = master) | |
1048 | decoder = make_element('decodebin', 'decode') | |
1049 | convert = make_element('audioconvert', 'convert') | |
1050 | encoder = me.encoder() | |
1051 | sink = make_element('filesink', 'sink', location = new) | |
0bf1fa38 | 1052 | for i in [source, decoder, convert, encoder, sink]: pipe.add(i) |
0dbdd41c MW |
1053 | link_elements([source, decoder]) |
1054 | link_elements([convert, encoder, sink]) | |
583b7e4a MW |
1055 | |
1056 | ## Some decoders (e.g., the AC3 decoder) include channel-position | |
1057 | ## indicators in their output caps. The Vorbis encoder interferes with | |
1058 | ## this, and you end up with a beautifully encoded mono signal from a | |
1059 | ## stereo source. From a quick butchers at the `vorbisenc' source, I | |
1060 | ## /think/ that this is only a problem with stereo signals: mono signals | |
1061 | ## are mono already, and `vorbisenc' accepts channel positions if there | |
1062 | ## are more than two channels. | |
1063 | ## | |
1064 | ## So we have this bodge. We already collected the decoded audio caps | |
1065 | ## during identification. So if we see 2-channel audio with channel | |
1066 | ## positions, we strip the positions off forcibly by adding a filter. | |
1067 | if id.dcap.get_name().startswith('audio/x-raw-') and \ | |
1068 | id.dcap.has_field('channels') and \ | |
1069 | id.dcap['channels'] == 2 and \ | |
1070 | id.dcap.has_field('channel-positions'): | |
1071 | dcap = GS.Caps() | |
1072 | c = id.dcap.copy() | |
1073 | c.remove_field('channel-positions') | |
1074 | dcap.append(c) | |
1075 | else: | |
1076 | dcap = None | |
1077 | ||
1078 | ## Hook onto the `decodebin' so we can link together the two halves of | |
1079 | ## our encoding chain. For now, we'll hope that there's only one audio | |
1080 | ## stream in there, and just throw everything else away. | |
1081 | def decoder_pad_arrived(elt, pad): | |
3bf73acf | 1082 | if pad.get_current_caps()[0].get_name().startswith('audio/'): |
583b7e4a MW |
1083 | if dcap: |
1084 | elt.link_pads_filtered(pad.get_name(), convert, 'sink', dcap) | |
1085 | else: | |
1086 | elt.link_pads(pad.get_name(), convert, 'sink') | |
1087 | decoder.connect('pad-added', decoder_pad_arrived) | |
1088 | ||
1089 | ## Watch the bus for completion messages. | |
1090 | fail = [] | |
1091 | def bus_message(bus, msg): | |
3bf73acf MW |
1092 | if msg.type == GS.MessageType.ERROR: |
1093 | fail[:] = (ValueError, msg.get_structure()['debug'], None) | |
583b7e4a | 1094 | loop.quit() |
3bf73acf | 1095 | elif msg.type == GS.MessageType.EOS: |
583b7e4a MW |
1096 | loop.quit() |
1097 | bmid = bus.connect('message', bus_message) | |
1098 | ||
1099 | ## Get everything ready and let it go. | |
3bf73acf MW |
1100 | bus.add_signal_watch() |
1101 | pipe.set_state(GS.State.PLAYING) | |
583b7e4a MW |
1102 | with GStreamerProgressEyecandy(filestatus(master, |
1103 | 'convert to %s' % me.NAME), | |
1104 | pipe): | |
1105 | loop.run() | |
3bf73acf MW |
1106 | pipe.set_state(GS.State.NULL) |
1107 | bus.remove_signal_watch() | |
1108 | bus.disconnect(bmid) | |
583b7e4a MW |
1109 | if fail: |
1110 | raise fail[0], fail[1], fail[2] | |
1111 | ||
1112 | ## Fix up the output file if we have to. | |
1113 | me.fixup(new) | |
1114 | ||
1115 | ## We're done. | |
1116 | OS.rename(new, target) | |
1117 | ||
1118 | class OggVorbisFormat (AudioFormat): | |
1119 | "AudioFormat object for Ogg Vorbis." | |
1120 | ||
00eb0a5b | 1121 | ## From https://en.wikipedia.org/wiki/Vorbis |
583b7e4a MW |
1122 | QMAP = [(-1, 45), ( 0, 64), ( 1, 80), ( 2, 96), |
1123 | ( 3, 112), ( 4, 128), ( 5, 160), ( 6, 192), | |
1124 | ( 7, 224), ( 8, 256), ( 9, 320), (10, 500)] | |
1125 | ||
1126 | NAME = 'Ogg Vorbis' | |
1127 | MIMETYPES = set(['application/ogg', 'audio/x-vorbis', 'audio/ogg', | |
1128 | 'audio/x-vorbis+ogg']) | |
1129 | EXT = 'ogg' | |
1130 | ||
1131 | def encoder_chain(me): | |
3589c4a4 MW |
1132 | encprops = {} |
1133 | if me.bitrate is not None: | |
1134 | for q, br in me.QMAP: | |
1135 | if br >= me.bitrate: | |
1136 | break | |
cf3c562e MW |
1137 | else: |
1138 | raise ValueError, 'no suitable quality setting found' | |
3589c4a4 MW |
1139 | encprops['quality'] = q/10.0 |
1140 | return [make_element('vorbisenc', **encprops), | |
583b7e4a MW |
1141 | make_element('oggmux')] |
1142 | ||
1143 | defformat('ogg-vorbis', OggVorbisFormat) | |
1144 | ||
1145 | class MP3Format (AudioFormat): | |
1146 | "AudioFormat object for MP3." | |
1147 | ||
1148 | NAME = 'MP3' | |
1149 | MIMETYPES = set(['audio/mpeg']) | |
1150 | EXT = 'mp3' | |
1151 | ||
1152 | def encoder_chain(me): | |
3589c4a4 | 1153 | encprops = {} |
3bf73acf MW |
1154 | if me.bitrate is not None: |
1155 | encprops['bitrate'] = me.bitrate | |
1156 | encprops['target'] = 'bitrate' | |
1157 | else: | |
1158 | encprops['quality'] = 4 | |
1159 | encprops['target'] = 'quality' | |
1160 | return [make_element('lamemp3enc', quality = 4, **encprops), | |
583b7e4a MW |
1161 | make_element('xingmux'), |
1162 | make_element('id3v2mux')] | |
1163 | ||
1164 | def fixup(me, path): | |
1165 | """ | |
1166 | Fix up MP3 files. | |
1167 | ||
1168 | GStreamer produces ID3v2 tags, but not ID3v1. This seems unnecessarily | |
1169 | unkind to stupid players. | |
1170 | """ | |
608b936e MW |
1171 | f = E3.load(path) |
1172 | if f is None: return | |
1173 | t = f.tag | |
1174 | if t is None: return | |
1175 | for v in [E3.id3.ID3_V2_3, E3.id3.ID3_V1]: | |
1176 | try: f.tag.save(version = v) | |
1177 | except (UnicodeEncodeError, | |
1178 | E3.id3.GenreException, | |
1179 | E3.id3.TagException): | |
1180 | pass | |
583b7e4a MW |
1181 | |
1182 | defformat('mp3', MP3Format) | |
1183 | ||
1184 | ###-------------------------------------------------------------------------- | |
1185 | ### Image handling, based on the Python Imaging Library. | |
1186 | ||
1187 | class ImageIdentifier (object): | |
1188 | """ | |
1189 | Analyses and identifies an image file. | |
1190 | ||
1191 | Simply leaves an Image object in the `img' property which can be inspected. | |
1192 | """ | |
1193 | ||
1194 | def __init__(me, file, mime): | |
1195 | ||
1196 | ## Get PIL to open the file. It will magically work out what kind of | |
1197 | ## file it is. | |
1198 | try: | |
1199 | me.img = I.open(file) | |
1200 | except IOError, exc: | |
1201 | ||
1202 | ## Unhelpful thing to raise on identification failure. We can | |
1203 | ## distinguish this from an actual I/O error because it doesn't have an | |
1204 | ## `errno'. | |
1205 | if exc.errno is None: | |
1206 | raise IdentificationFailure | |
1207 | raise | |
1208 | ||
1209 | me.mime = set([mime]) | |
1210 | ||
1211 | class ImageFormat (BaseFormat): | |
1212 | """ | |
1213 | An ImageFormat is a kind of Format specialized for image files. | |
1214 | ||
1215 | Subclasses don't need to provide anything other than the properties | |
1216 | required by all concrete Format subclasses. However, there is a | |
1217 | requirement that the `NAME' property match PIL's `format' name for the | |
1218 | format. | |
1219 | """ | |
1220 | ||
1221 | PROPS = prop('size', Num) | |
1222 | CATEGORY = FileCategory('image', ['image/*'], ImageIdentifier) | |
1223 | ||
1224 | def __init__(me, size = None, **kw): | |
1225 | """ | |
1226 | Initialize an ImageFormat object. | |
1227 | ||
1228 | Additional keywords are used when encoding, and may be recognized by | |
1229 | enhanced `check' methods in subclasses. | |
1230 | """ | |
1231 | me._size = size | |
1232 | me._props = kw | |
1233 | ||
1234 | def check(me, id): | |
1235 | "Check whether the ImageIdentifier ID matches our requirements." | |
1236 | return id.img.format == me.NAME and \ | |
1237 | (me._size is None or | |
1238 | (id.img.size[0] <= me._size and | |
1239 | id.img.size[1] <= me._size)) | |
1240 | ||
1241 | def convert(me, master, id, target): | |
1242 | "Encode the file MASTER, identified as ID, writing the result to TARGET." | |
1243 | ||
1244 | ## Write to a scratch file. | |
1245 | new = target + '.new' | |
1246 | ||
1247 | ## The ImageIdentifier already contains a copy of the open file. It | |
1248 | ## would be wasteful not to use it. | |
1249 | img = id.img | |
1250 | STATUS.set(filestatus(master, 'convert to %s' % me.NAME)) | |
1251 | ||
1252 | ## If there's a stated maximum size then scale the image down to match. | |
1253 | ## But thumbnailing clobbers the original, so take a copy. | |
1254 | if me._size is not None and \ | |
1255 | (img.size[0] > me._size or img.size[1] > me._size): | |
1256 | img = img.copy() | |
1257 | img.thumbnail((me._size, me._size), I.ANTIALIAS) | |
1258 | ||
1259 | ## Write the output image. | |
1260 | img.save(new, me.NAME, **me._props) | |
1261 | ||
1262 | ## Fix it up if necessary. | |
1263 | me.fixup(new) | |
1264 | ||
1265 | ## We're done. | |
1266 | OS.rename(new, target) | |
1267 | STATUS.commit() | |
1268 | ||
1269 | class JPEGFormat (ImageFormat): | |
1270 | """ | |
1271 | Image format for JPEG (actually JFIF) files. | |
1272 | ||
1273 | Interesting properties to set: | |
1274 | ||
1275 | optimize | |
1276 | If present, take a second pass to select optimal encoder settings. | |
1277 | ||
b524aa9d | 1278 | progressive |
583b7e4a MW |
1279 | If present, make a progressive file. |
1280 | ||
1281 | quality Integer from 1--100 (worst to best); default is 75. | |
1282 | """ | |
1283 | EXT = 'jpg' | |
1284 | NAME = 'JPEG' | |
1285 | PROPS = prop('optimize', None) \ | |
1286 | | prop('progressive', None, 'progression') \ | |
1287 | | prop('quality', Num) | |
1288 | ||
1289 | defformat('jpeg', JPEGFormat) | |
1290 | ||
1291 | class PNGFormat (ImageFormat): | |
1292 | """ | |
1293 | Image format for PNG files. | |
1294 | ||
1295 | Interesting properties: | |
1296 | ||
1297 | optimize | |
1298 | If present, make a special effort to minimize the output file. | |
1299 | """ | |
1300 | EXT = 'png' | |
1301 | NAME = 'PNG' | |
1302 | PROPS = prop('optimize', None) | |
1303 | ||
1304 | defformat('png', PNGFormat) | |
1305 | ||
1306 | class BMPFormat (ImageFormat): | |
1307 | """ | |
1308 | Image format for Windows BMP files, as used by RockBox. | |
1309 | ||
1310 | No additional properties. | |
1311 | """ | |
1312 | NAME = 'BMP' | |
1313 | EXT = 'bmp' | |
1314 | ||
1315 | defformat('bmp', BMPFormat) | |
1316 | ||
1317 | ###-------------------------------------------------------------------------- | |
e0361afb MW |
1318 | ### Remaining parsing machinery. |
1319 | ||
1320 | Type = K('type') - Name - D('{') - R(Policy) - D('}') | |
1321 | def build_type(s, l, t): | |
1322 | try: | |
1323 | cat = CATEGORYMAP[t[0]] | |
1324 | except KeyError: | |
1325 | raise P.ParseException(s, loc, "Unknown category `%s'" % t[0]) | |
1326 | pols = t[1] | |
1327 | if len(pols) == 1: pol = pols[0] | |
1328 | else: pol = AndPolicy(pols) | |
1329 | pol.setcategory(cat) | |
1330 | return pol | |
1331 | Type.setParseAction(build_type) | |
1332 | ||
1333 | TARGETS = [] | |
1334 | class TargetJob (object): | |
1335 | def __init__(me, targetdir, policies): | |
1336 | me.targetdir = targetdir | |
1337 | me.policies = policies | |
1338 | def perform(me): | |
1339 | TARGETS.append(me) | |
1340 | ||
1341 | Target = K('target') - String - D('{') - R(Type) - D('}') | |
1342 | def build_target(s, l, t): | |
1343 | return TargetJob(t[0], t[1]) | |
1344 | Target.setParseAction(build_target) | |
1345 | ||
1346 | VARS = { 'master': None } | |
1347 | class VarsJob (object): | |
1348 | def __init__(me, vars): | |
1349 | me.vars = vars | |
1350 | def perform(me): | |
1351 | for k, v in me.vars: | |
1352 | VARS[k] = v | |
1353 | ||
1354 | Var = prop('master', String) | |
1355 | Vars = K('vars') - D('{') - R(Var) - D('}') | |
1356 | def build_vars(s, l, t): | |
1357 | return VarsJob(t[0]) | |
1358 | Vars.setParseAction(build_vars) | |
1359 | ||
1360 | TopLevel = Vars | Target | |
1361 | Config = R(TopLevel) | |
1362 | Config.ignore(P.pythonStyleComment) | |
1363 | ||
1364 | ###-------------------------------------------------------------------------- | |
583b7e4a MW |
1365 | ### The directory grobbler. |
1366 | ||
e0361afb | 1367 | def grobble(master, targets, noact = False): |
583b7e4a | 1368 | """ |
e0361afb MW |
1369 | Work through the MASTER directory, writing converted files to TARGETS. |
1370 | ||
1371 | The TARGETS are a list of `TargetJob' objects, each describing a target | |
1372 | directory and a policy to apply to it. | |
1373 | ||
1374 | If NOACT is true, then don't actually do anything permanent to the | |
1375 | filesystem. | |
583b7e4a MW |
1376 | """ |
1377 | ||
e0361afb MW |
1378 | ## Transform the targets into a more convenient data structure. |
1379 | tpolmap = [] | |
1380 | for t in targets: | |
1381 | pmap = {} | |
1382 | tpolmap.append(pmap) | |
1383 | for p in t.policies: pmap.setdefault(p.cat, []).append(p) | |
583b7e4a | 1384 | |
e0361afb MW |
1385 | ## Keep track of the current position in the master tree. |
1386 | dirs = [] | |
583b7e4a | 1387 | |
e0361afb MW |
1388 | ## And the files which haven't worked. |
1389 | broken = [] | |
583b7e4a | 1390 | |
e0361afb MW |
1391 | def grobble_file(master, pmap, targetdir, cohorts): |
1392 | ## Convert MASTER, writing the result to TARGETDIR. | |
1393 | ## | |
1394 | ## The COHORTS are actually (CAT, ID, COHORT) triples, where a COHORT is | |
1395 | ## a list of (FILENAME, ID) pairs. | |
1396 | ## | |
1397 | ## Since this function might convert the MASTER file, the caller doesn't | |
1398 | ## know the name of the output files, so we return then as a list. | |
583b7e4a MW |
1399 | |
1400 | done = set() | |
1401 | st_m = OS.stat(master) | |
1402 | ||
1403 | ## Work through each category listed and apply its policy. | |
1404 | for cat, id, cohort in cohorts: | |
1405 | ||
1406 | ## Go through the category's policies and see if any match. If we fail | |
1407 | ## here, see if there are more categories to try. | |
e0361afb | 1408 | for pol in pmap[cat]: |
583b7e4a MW |
1409 | acts = pol.actions(master, targetdir, id, cohort) |
1410 | if acts: break | |
1411 | else: | |
1412 | continue | |
1413 | ||
1414 | ## Work through the targets one by one. | |
1415 | for a in acts: | |
1416 | done.add(a.target) | |
1417 | ||
1418 | ## Find out whether the target file already exists and is up-to-date | |
1419 | ## with respect to the master. (Caution here with low-resolution | |
1420 | ## timestamps.) If it's OK, then just move on. | |
1421 | try: | |
1422 | st_t = OS.stat(a.target) | |
1423 | if st_m.st_mtime < st_t.st_mtime or \ | |
1424 | (st_m.st_ino, st_m.st_dev) == (st_t.st_ino, st_t.st_dev): | |
1425 | continue | |
1426 | except OSError, err: | |
1427 | if err.errno not in (E.ENOENT, E.ENOTDIR): | |
1428 | raise | |
1429 | ||
1430 | ## We have real work to do. If there's a current status message, | |
1431 | ## it's the containing directory so flush it so that people know | |
1432 | ## where we are. | |
1433 | STATUS.commit() | |
1434 | ||
1435 | ## Remove the target. (A hardlink will fail if the target already | |
1436 | ## exists.) | |
e0361afb | 1437 | if not noact: |
583b7e4a MW |
1438 | try: |
1439 | OS.unlink(a.target) | |
1440 | except OSError, err: | |
1441 | if err.errno not in (E.ENOENT, E.ENOTDIR): | |
1442 | raise | |
1443 | ||
1444 | ## Do whatever it is we decided to do. | |
e0361afb | 1445 | if noact: |
583b7e4a MW |
1446 | STATUS.commit(filestatus(master, a)) |
1447 | else: | |
1448 | a.perform() | |
1449 | ||
1450 | ## We're done. Return the names of the targets. | |
1451 | return list(done) | |
1452 | ||
1453 | @contextmanager | |
e0361afb MW |
1454 | def wrap(masterfile): |
1455 | ## Handle exceptions found while trying to convert a particular file or | |
1456 | ## directory. | |
583b7e4a MW |
1457 | |
1458 | try: | |
1459 | yield masterfile | |
1460 | ||
1461 | ## Something bad happened. Report the error, but continue. (This list | |
1462 | ## of exceptions needs a lot of work.) | |
1463 | except (IOError, OSError), exc: | |
1464 | STATUS.clear() | |
1465 | STATUS.commit(filestatus(masterfile, 'failed (%s)' % exc)) | |
e0361afb | 1466 | broken.append((masterfile, exc)) |
583b7e4a | 1467 | |
e0361afb MW |
1468 | def grobble_dir(master, targets): |
1469 | ## Recursively convert files in MASTER, writing them to the TARGETS. | |
583b7e4a | 1470 | |
e0361afb MW |
1471 | ## Keep track of the subdirectories we encounter, because we'll need to |
1472 | ## do all of those in one go at the end. | |
1473 | subdirs = set() | |
583b7e4a | 1474 | |
e0361afb MW |
1475 | ## Work through each target directory in turn. |
1476 | for target, pmap in zip(targets, tpolmap): | |
583b7e4a | 1477 | |
e0361afb MW |
1478 | ## Make sure the TARGET exists and is a directory. It's a fundamental |
1479 | ## assumption of this program that the entire TARGET tree is | |
1480 | ## disposable, so if something exists but isn't a directory, we should | |
1481 | ## kill it. | |
1482 | if OS.path.isdir(target): | |
1483 | pass | |
1484 | else: | |
1485 | if OS.path.exists(target): | |
1486 | STATUS.commit(filestatus(target, 'clear nondirectory')) | |
1487 | if not noact: | |
1488 | OS.unlink(target) | |
1489 | STATUS.commit(filestatus(target, 'create directory')) | |
1490 | if not noact: | |
1491 | OS.mkdir(target) | |
1492 | ||
1493 | ## Keep a list of things in the target. As we convert files, we'll | |
1494 | ## check them off. Anything left over is rubbish and needs to be | |
1495 | ## deleted. | |
1496 | checklist = {} | |
1497 | try: | |
1498 | for i in OS.listdir(target): | |
1499 | checklist[i] = False | |
1500 | except OSError, err: | |
1501 | if err.errno not in (E.ENOENT, E.ENOTDIR): | |
1502 | raise | |
1503 | ||
1504 | ## Keep track of the files in each category. | |
1505 | catmap = {} | |
1506 | todo = [] | |
1507 | done = [] | |
1508 | ||
1509 | ## Work through the master files. | |
1510 | for f in sorted(OS.listdir(master)): | |
1511 | ||
1512 | ## If the killswitch has been pulled then stop. The whole idea is | |
1513 | ## that we want to cause a clean shutdown if possible, so we don't | |
1514 | ## want to do it in the middle of encoding because the encoding | |
1515 | ## effort will have been wasted. This is the only place we need to | |
1516 | ## check. If we've exited the loop, then clearing old files will | |
1517 | ## probably be fast, and we'll either end up here when the recursive | |
1518 | ## call returns or we'll be in the same boat as before, clearing old | |
1519 | ## files, only up a level. If worst comes to worst, we'll be killed | |
1520 | ## forcibly somewhere inside `SH.rmtree', and that can continue where | |
1521 | ## it left off. | |
1522 | if KILLSWITCH.is_set(): | |
1523 | return | |
1524 | ||
1525 | ## Do something with the file. | |
1526 | with wrap(OS.path.join(master, f)) as masterfile: | |
1527 | ||
1528 | ## If it's a directory then prepare to grobble it recursively, but | |
1529 | ## don't do that yet. | |
1530 | if OS.path.isdir(masterfile): | |
1531 | subdirs.add(f) | |
1532 | done.append(OS.path.join(target, f)) | |
1533 | ||
1534 | ## Otherwise it's a file. Work out what kind, and stash it under | |
1535 | ## the appropriate categories. Later, we'll apply policy to the | |
1536 | ## files, by category, and work out what to do with them all. | |
1537 | else: | |
3bf73acf MW |
1538 | mime = GIO.file_new_for_path(masterfile) \ |
1539 | .query_info('standard::content-type', 0) \ | |
2a236d3a | 1540 | .get_content_type() |
e0361afb MW |
1541 | cats = [] |
1542 | for cat in pmap.iterkeys(): | |
1543 | id = cat.identify(masterfile, mime) | |
1544 | if id is None: continue | |
1545 | catmap.setdefault(cat, []).append((masterfile, id)) | |
1546 | cats.append((cat, id)) | |
1547 | if not cats: | |
1548 | catmap.setdefault(None, []).append((masterfile, id)) | |
1549 | todo.append((masterfile, cats)) | |
1550 | ||
1551 | ## Work through the categorized files to see what actions to do for | |
1552 | ## them. | |
1553 | for masterfile, cats in todo: | |
1554 | with wrap(masterfile): | |
1555 | done += grobble_file(masterfile, pmap, target, | |
1556 | [(cat, id, catmap[cat]) for cat, id in cats]) | |
1557 | ||
1558 | ## Check the results off the list so that we don't clear it later. | |
1559 | for f in done: | |
1560 | checklist[OS.path.basename(f)] = True | |
1561 | ||
1562 | ## Maybe there's stuff in the target which isn't accounted for. Delete | |
1563 | ## it: either the master has changed, or the policy for this target has | |
1564 | ## changed. Either way, the old files aren't wanted. | |
1565 | for f in checklist: | |
1566 | if not checklist[f]: | |
1567 | STATUS.commit(filestatus(f, 'clear bogus file')) | |
1568 | if not noact: | |
1569 | bogus = OS.path.join(target, f) | |
1570 | try: | |
1571 | if OS.path.isdir(bogus): | |
1572 | SH.rmtree(bogus) | |
1573 | else: | |
1574 | OS.unlink(bogus) | |
1575 | except OSError, err: | |
1576 | if err.errno != E.ENOENT: | |
1577 | raise | |
1578 | ||
1579 | ## If there are subdirectories which want processing then do those. | |
1580 | ## Keep the user amused by telling him where we are in the tree. | |
1581 | for d in sorted(subdirs): | |
1582 | dirs.append(d) | |
1583 | STATUS.set('/'.join(dirs)) | |
1584 | with wrap(OS.path.join(master, d)) as masterdir: | |
1585 | try: | |
1586 | grobble_dir(masterdir, | |
1587 | [OS.path.join(target, d) for target in targets]) | |
1588 | finally: | |
1589 | dirs.pop() | |
1590 | STATUS.set('/'.join(dirs)) | |
583b7e4a | 1591 | |
e0361afb MW |
1592 | ## Right. We're ready to go. |
1593 | grobble_dir(master, [t.targetdir for t in targets]) | |
1594 | return broken | |
583b7e4a MW |
1595 | |
1596 | ###-------------------------------------------------------------------------- | |
1597 | ### Command-line interface. | |
1598 | ||
1599 | QUIS = OS.path.basename(SYS.argv[0]) | |
1600 | ||
1601 | def moan(msg): | |
1602 | "Report a warning message to the user." | |
1603 | SYS.stderr.write('%s: %s\n' % (QUIS, msg)) | |
1604 | ||
1605 | def die(msg): | |
1606 | "Report a fatal error message to the user." | |
1607 | moan(msg) | |
1608 | SYS.exit(1) | |
1609 | ||
1610 | def parse_opts(args): | |
1611 | """ | |
1612 | Parse command-line arguments in ARGS. | |
1613 | ||
1614 | Returns a Grobbler object and the MASTER and TARGET directories to be | |
1615 | grobbled. | |
1616 | """ | |
1617 | ||
1618 | ## Build the option parser object. | |
1619 | op = OP.OptionParser(prog = QUIS, version = VERSION, | |
5379ab85 MW |
1620 | usage = '%prog [-in] [-t TIMEOUT] [-T TIMEOUT] ' |
1621 | 'CONFIG', | |
583b7e4a MW |
1622 | description = """\ |
1623 | Convert a directory tree of files according to the configuration file | |
1624 | CONFIG. | |
1625 | """) | |
1626 | ||
1627 | ## Timeout handling. | |
1628 | def cb_time(opt, ostr, arg, op): | |
1629 | m = RX.match(r'\s*(\d+)\s*([dhms]?)\s*', arg) | |
1630 | if not m: | |
1631 | raise OP.OptionValueerror, 'bad time value `%s\'' % arg | |
1632 | t, u = m.groups() | |
1633 | t = int(t) * { '': 1, 's': 1, 'm': 60, 'h': 3600, 'd': 86400 }[u] | |
1634 | setattr(op.values, opt.dest, t) | |
1635 | op.add_option('-t', '--timeout', type = 'string', metavar = 'SECS', | |
1636 | dest = 'timeout', | |
1637 | help = 'stop processing nicely after SECS', | |
1638 | action = 'callback', callback = cb_time) | |
1639 | op.add_option('-T', '--timeout-nasty', type = 'string', metavar = 'SECS', | |
1640 | dest = 'timeout_nasty', | |
1641 | help = 'stop processing unpleasantly after further SECS', | |
1642 | action = 'callback', callback = cb_time) | |
1643 | ||
1644 | ## Other options. | |
1645 | op.add_option('-i', '--interactive', action = 'store_true', dest = 'tty', | |
1646 | help = 'provide progress information') | |
1647 | op.add_option('-n', '--no-act', action = 'store_true', dest = 'noact', | |
1648 | help = 'don\'t actually modify the filesystem') | |
1649 | ||
1650 | ## Ready to rock. | |
1651 | op.set_defaults(formats = [], noact = False, | |
1652 | timeout = None, timeout_nasty = 300) | |
1653 | opts, args = op.parse_args(args) | |
1654 | ||
1655 | ## Check that we got the non-option arguments that we want. | |
1656 | if len(args) != 1: | |
1657 | op.error('wrong number of arguments') | |
1658 | ||
1659 | ## Act on the options. | |
1660 | if opts.tty: | |
1661 | STATUS.eyecandyp = True | |
1662 | if opts.timeout is not None: | |
1663 | to = TH.Thread(target = timeout, | |
1664 | args = (opts.timeout, opts.timeout_nasty)) | |
1665 | to.daemon = True | |
1666 | to.start() | |
1667 | ||
1668 | ## Parse the configuration file. | |
1669 | with open(args[0]) as conf: | |
1670 | jobs, = Config.parseFile(conf, True) | |
1671 | for j in jobs: | |
1672 | j.perform() | |
1673 | ||
1674 | return opts | |
1675 | ||
1676 | if __name__ == '__main__': | |
1677 | opts = parse_opts(SYS.argv[1:]) | |
1678 | if 'master' not in VARS: | |
1679 | die("no master directory set") | |
e0361afb | 1680 | broken = grobble(VARS['master'], TARGETS, opts.noact) |
583b7e4a MW |
1681 | if broken: |
1682 | moan('failed to convert some files:') | |
1683 | for file, exc in broken: | |
1684 | moan('%s: %s' % (file, exc)) | |
1685 | SYS.exit(1) | |
1686 | ||
1687 | ## This is basically a successful completion: we did what we were asked to | |
1688 | ## do. It seems polite to report a message, though. | |
1689 | ## | |
1690 | ## Why don't we have a nonzero exit status? The idea would be that a | |
1691 | ## calling script would be interested that we used up all of our time, and | |
1692 | ## not attempt to convert some other directory as well. But that doesn't | |
1693 | ## quite work. Such a script would need to account correctly for time we | |
1694 | ## had spent even if we complete successfully. And if the script is having | |
1695 | ## to watch the clock itself, it can do that without our help here. | |
1696 | if KILLSWITCH.is_set(): | |
1697 | moan('killed by timeout') | |
1698 | ||
1699 | ###----- That's all, folks -------------------------------------------------- |