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 | ### | |
10 | ### This program is free software; you can redistribute it and/or modify | |
11 | ### it under the terms of the GNU General Public License as published by | |
12 | ### the Free Software Foundation; either version 2 of the License, or | |
13 | ### (at your option) any later version. | |
14 | ### | |
15 | ### This program is distributed in the hope that it will be useful, | |
16 | ### but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | ### MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | ### GNU General Public License for more details. | |
19 | ### | |
20 | ### You should have received a copy of the GNU General Public License | |
21 | ### along with this program; if not, write to the Free Software Foundation, | |
22 | ### Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. | |
23 | ||
24 | ###-------------------------------------------------------------------------- | |
25 | ### External dependencies. | |
26 | ||
27 | ## Language features. | |
28 | from __future__ import with_statement | |
29 | ||
30 | ## Standard Python libraries. | |
31 | import sys as SYS | |
32 | import os as OS | |
33 | import errno as E | |
34 | import time as T | |
35 | import unicodedata as UD | |
36 | import fnmatch as FN | |
37 | import re as RX | |
38 | import shutil as SH | |
39 | import optparse as OP | |
40 | import threading as TH | |
41 | import shlex as L | |
42 | from math import sqrt | |
43 | from contextlib import contextmanager | |
44 | ||
45 | ## eyeD3 tag fettling. | |
46 | import eyeD3 as E3 | |
47 | ||
48 | ## Gstreamer. It picks up command-line arguments -- most notably `--help' -- | |
49 | ## and processes them itself. Of course, its help is completely wrong. This | |
50 | ## kludge is due to Jonas Wagner. | |
51 | _argv, SYS.argv = SYS.argv, [] | |
52 | import gobject as G | |
53 | import gio as GIO | |
54 | import gst as GS | |
55 | SYS.argv = _argv | |
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. | |
106 | #print ';; %r -> %d' % (s, w) | |
107 | return w | |
108 | ||
109 | class StatusLine (object): | |
110 | """ | |
111 | Maintains a status line containing ephemeral progress information. | |
112 | ||
113 | The status line isn't especially important, but it keeps interactive users | |
114 | amused. | |
115 | ||
116 | There should be only one status line object in your program; otherwise | |
117 | they'll interfere with each other and get confused. | |
118 | ||
119 | The update algorithm (in `set') is fairly careful to do the right thing | |
120 | with long status `lines', and to work properly in an Emacs `shell' buffer. | |
121 | """ | |
122 | ||
123 | def __init__(me): | |
124 | "Initialize the status line." | |
125 | me._last = '' | |
126 | me._lastlen = 0 | |
127 | me.eyecandyp = OS.isatty(SYS.stdout.fileno()) | |
128 | ||
129 | def set(me, line): | |
130 | """ | |
131 | Set the status line contents to LINE, replacing what was there before. | |
132 | ||
133 | This only produces actual output if stdout is interactive. | |
134 | """ | |
135 | n = len(line) | |
136 | ||
137 | ## Eyecandy update. | |
138 | if me.eyecandyp: | |
139 | ||
140 | #print ';; new status %r' % line | |
141 | ||
142 | ## If the old line was longer, we need to clobber its tail, so work out | |
143 | ## what that involves. | |
144 | if n < me._lastlen: | |
145 | b = charwidth(me._last[n:]) | |
146 | pre = '\b'*b + ' '*b | |
147 | else: | |
148 | pre = '' | |
149 | ||
150 | ## Now figure out the length of the common prefix between what we had | |
151 | ## before and what we have now. This reduces the amount of I/O done, | |
152 | ## which keeps network traffic down on SSH links, and keeps down the | |
153 | ## amount of work slow terminal emulators like Emacs have to do. | |
154 | i = 0 | |
155 | m = min(n, me._lastlen) | |
156 | while i < m and line[i] == me._last[i]: | |
157 | i += 1 | |
158 | ||
159 | ## Actually do the output, all in one syscall. | |
160 | b = charwidth(me._last[i:]) | |
161 | SYS.stdout.write(pre + '\b'*b + line[i:]) | |
162 | #print ';; => %r' % (pre + '\b'*b + line[i:]) | |
163 | SYS.stdout.flush() | |
164 | ||
165 | ## Update our idea of what's gone on. | |
166 | me._lastlen = n | |
167 | me._last = line | |
168 | ||
169 | def clear(me): | |
170 | "Clear the status line. Just like set('')." | |
171 | me.set('') | |
172 | ||
173 | def commit(me, line = None): | |
174 | """ | |
175 | Commit the current status line, and maybe the string LINE. | |
176 | ||
177 | If the current status line is nonempty, then commit it to the transcript. | |
178 | If LINE is not None, then commit that to the transcript too. | |
179 | ||
180 | After all of this, we clear the status line to get back to a clean state. | |
181 | """ | |
182 | if me._last: | |
183 | if me.eyecandyp: | |
184 | SYS.stdout.write('\n') | |
185 | else: | |
186 | SYS.stdout.write(me._last + '\n') | |
187 | if line is not None: | |
188 | SYS.stdout.write(line + '\n') | |
189 | me._lastlen = 0 | |
190 | me._last = '' | |
191 | ||
192 | STATUS = StatusLine() | |
193 | ||
194 | def filestatus(file, status): | |
195 | return '%s%s: %s' % (' '*8, OS.path.basename(file), status) | |
196 | ||
197 | class ProgressEyecandy (object): | |
198 | """ | |
199 | Provide amusement while something big and complicated is happening. | |
200 | ||
201 | This is an abstract class. Subclasses must provide a method `progress' | |
202 | returning a pair (CURRENT, MAX) indicating the current progress through the | |
203 | operation. | |
204 | """ | |
205 | ||
206 | def __init__(me, what, silentp = False): | |
207 | """ | |
208 | Initialize a progress meter. | |
209 | ||
210 | WHAT is a prefix string to be written before the progress eyecandy | |
211 | itself. | |
212 | """ | |
213 | me._what = what | |
214 | me._silentp = silentp | |
215 | me._spinner = 0 | |
216 | me._start = T.time() | |
217 | ||
218 | def _fmt_time(me, t): | |
219 | "Format T as a time, in (maybe hours) minutes and seconds." | |
220 | s, t = t % 60, int(t/60) | |
221 | m, h = t % 60, int(t/60) | |
222 | if h > 0: | |
223 | return '%d:%02d:%02d' % (h, m, s) | |
224 | else: | |
225 | return '%02d:%02d' % (m, s) | |
226 | ||
227 | def show(me): | |
228 | "Show the current level of progress." | |
229 | ||
230 | ## If we're not showing pointless frippery, don't bother at all. | |
231 | if not STATUS.eyecandyp: | |
232 | return | |
233 | ||
234 | ## Update the spinner index. | |
235 | me._spinner = (me._spinner + 1)%4 | |
236 | ||
237 | ## Fetch the current progress information. Note that we always fetch | |
238 | ## both the current and maximum levels, because both might change if an | |
239 | ## operation revises its idea of how much work needs doing. | |
240 | cur, max = me.progress() | |
241 | ||
242 | ## If we couldn't get progress information, display something vaguely | |
243 | ## amusing anyway. | |
244 | if cur is None or max is None: | |
245 | STATUS.set('%s %c [unknown progress]' % | |
246 | (me._what, r'/-\|'[me._spinner])) | |
247 | return | |
248 | ||
249 | ## Work out -- well, guess -- the time remaining. | |
250 | if cur: | |
251 | t = T.time() | |
252 | eta = me._fmt_time((t - me._start)*(max - cur)/cur) | |
253 | else: | |
254 | eta = '???' | |
255 | ||
256 | ## Set the status bar. | |
257 | n = 40*cur/max | |
258 | STATUS.set('%s %c [%s%s] %3d%% (%s)' % \ | |
259 | (me._what, | |
260 | r'/-\|'[me._spinner], | |
261 | '='*n, ' '*(40 - n), | |
262 | 100*cur/max, | |
263 | eta)) | |
264 | ||
265 | def done(me, win = True): | |
266 | "Show a completion notice, or a failure if WIN is false." | |
267 | if not win: | |
268 | STATUS.set('%s FAILED!' % me._what) | |
269 | elif not me._silentp: | |
270 | STATUS.set('%s done (%s)' % | |
271 | (me._what, | |
272 | me._fmt_time(T.time() - me._start))) | |
273 | else: | |
274 | return | |
275 | STATUS.commit() | |
276 | ||
277 | ###-------------------------------------------------------------------------- | |
278 | ### Timeout handling. | |
279 | ||
280 | KILLSWITCH = TH.Event() | |
281 | ||
282 | def timeout(t0, t1): | |
283 | T.sleep(t0) | |
284 | KILLSWITCH.set() | |
285 | T.sleep(t1) | |
286 | moan('dying messily due to timeout') | |
287 | OS._exit(3) | |
288 | ||
289 | ###-------------------------------------------------------------------------- | |
290 | ### Parsing utilities. | |
291 | ||
292 | ## Allow hyphens in identifiers. | |
293 | IDCHARS = P.alphanums + '-_' | |
294 | P.Keyword.setDefaultKeywordChars(IDCHARS) | |
295 | ||
296 | ## Some common kinds of tokens. | |
297 | Name = P.Word(IDCHARS) | |
298 | Num = P.Word(P.nums).setParseAction(lambda toks: map(int, toks)) | |
299 | String = P.QuotedString('"', '\\') | |
300 | ||
301 | ## Handy abbreviations for constructed parser elements. | |
302 | def K(k): return P.Keyword(k).suppress() | |
303 | def D(d): return P.Literal(d).suppress() | |
304 | ##R = P.ZeroOrMore | |
305 | def R(p): return P.ZeroOrMore(p).setParseAction(lambda s, l, t: [t]) | |
306 | O = P.Optional | |
307 | ||
308 | ###-------------------------------------------------------------------------- | |
309 | ### Format identification and conversion. | |
310 | ||
311 | class IdentificationFailure (Exception): | |
312 | pass | |
313 | ||
314 | class FileCategory (object): | |
315 | """ | |
316 | A FileCategory represents a class of files. | |
317 | ||
318 | For example, it's sensible to consider audio, or image files as a | |
319 | category. A file category knows how to recognize member files from | |
320 | MIME content types. | |
321 | """ | |
322 | ||
323 | def __init__(me, name, mime_pats, ident): | |
324 | """ | |
325 | Construct a new category. | |
326 | ||
327 | The PATS are a list of `fnmatch' patterns to be compared with a MIME | |
328 | type. The IDENT is a function which produces an identification object | |
329 | given a file's name and first-guess MIME type. The object is passed to a | |
330 | Format's `check' method to see whether a file needs re-encoding, and to | |
331 | `convert' to assist with the conversion. | |
332 | ||
333 | An identification object must have an attribute `mime' which is a set of | |
334 | possible MIME types accumulated for the object. | |
335 | """ | |
336 | me.name = name | |
337 | me._mime_pats = mime_pats | |
338 | me._ident = ident | |
339 | CATEGORYMAP[name] = me | |
340 | ||
341 | def identify(me, file, mime): | |
342 | """ | |
343 | Attempt to identify FILE, given its apparent MIME type. | |
344 | ||
345 | If identification succeeds, return an identification object which can be | |
346 | used by associated file formats; otherwise return None. | |
347 | """ | |
348 | for p in me._mime_pats: | |
349 | if not FN.fnmatchcase(mime, p): | |
350 | continue | |
351 | try: | |
352 | return me._ident(file, mime) | |
353 | except IdentificationFailure: | |
354 | pass | |
355 | return None | |
356 | ||
357 | class BaseFormat (object): | |
358 | """ | |
359 | A BaseFormat object represents a particular encoding and parameters. | |
360 | ||
361 | The object can verify (the `check' method) whether a particular file | |
362 | matches its requirements, and if necessary (`encode') re-encode a file. | |
363 | ||
364 | Subclasses should define the following methods. | |
365 | ||
366 | check(ID) | |
367 | Answer whether the file identified by ID is acceptable according to | |
368 | the receiver's parameters. | |
369 | ||
370 | convert(MASTER, ID, TARGET) | |
371 | Convert the file MASTER, which has been identified as ID, according | |
372 | to the receiver's parameters, writing the output to TARGET. | |
373 | ||
374 | Subclasses should also provide these attributes. | |
375 | ||
376 | CATEGORY | |
377 | A FileCategory object for the category of files that this format | |
378 | lives within. | |
379 | ||
380 | EXT A file extension to be applied to encoded output files. | |
381 | ||
382 | NAME A user-facing name for the format. | |
383 | ||
384 | PROPS A parser element to parse a property definition. It should produce | |
385 | a pair NAME, VALUE to be stored in a dictionary. | |
386 | ||
387 | Subclasses for different kinds of file may introduce more subclass | |
388 | protocol. | |
389 | """ | |
390 | ||
391 | def fixup(me, path): | |
392 | """Post-encoding fixups.""" | |
393 | pass | |
394 | ||
395 | FORMATMAP = {} | |
396 | CATEGORYMAP = {} | |
397 | ||
398 | def defformat(name, cls): | |
399 | "Define a format NAME using class CLS." | |
400 | if not hasattr(cls, 'NAME'): | |
401 | raise ValueError, 'abstract class' | |
402 | if not hasattr(cls, 'CATEGORY'): | |
403 | raise ValueError, 'no category' | |
404 | FORMATMAP[name] = cls | |
405 | ||
406 | class FormatParser (P.ParserElement): | |
407 | """ | |
408 | Parse a format specifier: | |
409 | ||
410 | format-spec ::= string [format-properties] | |
411 | format-properties ::= `{' format-property (`,' format-property)* `}' | |
412 | ||
413 | The syntax of a format-property is determined by the PROPS attribute on the | |
414 | named format and its superclasses. | |
415 | """ | |
416 | ||
417 | ## We cache the parser elements we generate to avoid enormous consing. | |
418 | CACHE = {} | |
419 | ||
420 | def parseImpl(me, s, loc, actp = True): | |
421 | ||
422 | ## Firstly, determine the format name. | |
423 | loc, r = Name._parse(s, loc, actp) | |
424 | fmt = r[0] | |
425 | ||
426 | ## Look up the format class. | |
427 | try: fcls = FORMATMAP[fmt] | |
428 | except KeyError: | |
429 | raise P.ParseException(s, loc, "Unknown format `%s'" % fmt) | |
430 | ||
431 | ## Fetch the property-list parser from the cache, if possible; else | |
432 | ## construct it. | |
433 | try: | |
434 | pp = me.CACHE[fmt] | |
435 | except KeyError: | |
436 | seen = set() | |
437 | prop = None | |
438 | for c in fcls.mro(): | |
439 | try: p = c.PROPS | |
440 | except AttributeError: continue | |
441 | if p in seen: continue | |
442 | if prop is None: prop = p | |
443 | else: prop |= p | |
444 | seen.add(p) | |
445 | if prop is None: | |
446 | pp = me.CACHE[fmt] = None | |
447 | else: | |
448 | props = P.delimitedList(prop) | |
449 | props.setParseAction(lambda s, l, t: dict(t.asList())) | |
450 | pp = me.CACHE[fmt] = O(D('{') - props - D('}')) | |
451 | ||
452 | ## Parse the properties. | |
453 | if pp is None: | |
454 | pd = {} | |
455 | else: | |
456 | loc, r = pp._parse(s, loc, actp) | |
457 | if r: pd = r[0] | |
458 | else: pd = {} | |
459 | ||
460 | ## Construct the format object and return it. | |
461 | return loc, fcls(**pd) | |
462 | ||
463 | Format = FormatParser() | |
464 | ||
465 | def prop(kw, pval, tag = None): | |
466 | if tag is None: tag = kw | |
467 | if pval is None: | |
468 | p = K(kw) | |
469 | p.setParseAction(lambda s, l, t: (tag, True)) | |
470 | else: | |
471 | p = K(kw) + D('=') + pval | |
472 | p.setParseAction(lambda s, l, t: (tag, t[0])) | |
473 | return p | |
474 | ||
475 | ###-------------------------------------------------------------------------- | |
476 | ### Policies and actions. | |
477 | ||
478 | class Action (object): | |
479 | """ | |
480 | An Action object represents a conversion action to be performed. | |
481 | ||
482 | This class isn't intended to be instantiated directly. It exists to define | |
483 | some protocol common to all Action objects. | |
484 | ||
485 | Action objects have the following attributes. | |
486 | ||
487 | master The name of the master (source) file. | |
488 | ||
489 | target The name of the target (destination) file. | |
490 | ||
491 | PRIORITY The priority of the action, for deciding which of two actions | |
492 | to perform. Higher priorities are more likely to win. | |
493 | ||
494 | Converting an Action to a string describes the action in a simple | |
495 | user-readable manner. The `perform' method actually carries the action | |
496 | out. | |
497 | """ | |
498 | ||
499 | PRIORITY = 0 | |
500 | ||
501 | def __init__(me, master): | |
502 | "Stash the MASTER file name for later." | |
503 | me.master = master | |
504 | ||
505 | def choose(me, him): | |
506 | "Choose either ME or HIM and return one." | |
507 | if him is None or me.PRIORITY > him.PRIORITY: | |
508 | return me | |
509 | else: | |
510 | return him | |
511 | ||
512 | class CopyAction (Action): | |
513 | """ | |
514 | An Action object for simply copying a file. | |
515 | ||
516 | Actually we try to hardlink it first, falling back to a copy later. This | |
517 | is both faster and more efficient with regard to disk space. | |
518 | """ | |
519 | ||
520 | ## Copying is good. Linking is really good, but we can't tell the | |
521 | ## difference at this stage. | |
522 | PRIORITY = 10 | |
523 | ||
524 | def __init__(me, master, targetdir): | |
525 | "Initialize a CopyAction, from MASTER to the TARGETDIR directory." | |
526 | Action.__init__(me, master) | |
527 | me.target = OS.path.join(targetdir, OS.path.basename(master)) | |
528 | ||
529 | def __str__(me): | |
530 | return 'copy/link' | |
531 | ||
532 | def perform(me): | |
533 | "Actually perform a CopyAction." | |
534 | try: | |
535 | STATUS.set(filestatus(me.master, 'link')) | |
536 | OS.link(me.master, me.target) | |
537 | except OSError, err: | |
538 | if err.errno != E.EXDEV: | |
539 | raise | |
540 | STATUS.set(filestatus(me.master, 'copy')) | |
541 | new = me.target + '.new' | |
542 | SH.copyfile(me.master, new) | |
543 | OS.rename(new, me.target) | |
544 | STATUS.commit() | |
545 | ||
546 | class ConvertAction (Action): | |
547 | """ | |
548 | An Action object for converting a file to a given format. | |
549 | ||
550 | Additional attributes: | |
551 | ||
552 | id The identification object for the master file. | |
553 | ||
554 | format The format to which we're meant to conver the master. | |
555 | """ | |
556 | ||
557 | def __init__(me, master, targetdir, id, format): | |
558 | "Initialize a ConvertAction." | |
559 | Action.__init__(me, master) | |
560 | stem, ext = OS.path.splitext(OS.path.basename(master)) | |
561 | me.target = OS.path.join(targetdir, stem + '.' + format.EXT) | |
562 | me.id = id | |
563 | me.format = format | |
564 | ||
565 | def __str__(me): | |
566 | return 'convert to %s' % me.format.NAME | |
567 | ||
568 | def perform(me): | |
569 | "Acually perform a ConvertAction." | |
570 | STATUS.set(filestatus(me.master, me)) | |
571 | me.format.convert(me.master, me.id, me.target) | |
572 | ||
573 | Policy = P.Forward() | |
574 | ||
575 | class FormatPolicy (object): | |
576 | """ | |
577 | A FormatPolicy object represents a set of rules for how to convert files. | |
578 | ||
579 | Given a master file, the FormatPolicy will identify it and return a list of | |
580 | actions to be performed. The methods required of a FormatPolicy are: | |
581 | ||
582 | setcategory(CAT) | |
583 | Store CAT as the policy's category. Check that this is consistent | |
584 | with the policy as stored. | |
585 | ||
586 | actions(MASTER, TARGETDIR, ID, COHORT) | |
587 | Given a MASTER file, identified as ID, a target directory | |
588 | TARGETDIR, and a list COHORT of (FILE, ID) pairs for other files | |
589 | of the same category in the same directory, return a list of | |
590 | actions to be performed to get the target directory into the right | |
591 | form. The list might be empty if the policy object /rejects/ the | |
592 | file. | |
593 | """ | |
594 | ||
595 | class AndPolicy (FormatPolicy): | |
596 | """ | |
597 | A FormatPolicy which does the union of a bunch of other policies. | |
598 | ||
599 | Each subsidiary policy is invoked in turn. The highest-priority action for | |
600 | each target file is returned. | |
601 | """ | |
602 | ||
603 | def __init__(me, policies): | |
604 | me._policies = policies | |
605 | ||
606 | def setcategory(me, cat): | |
607 | me.cat = cat | |
608 | for p in me._policies: | |
609 | p.setcategory(cat) | |
610 | ||
611 | def actions(me, master, targetdir, id, cohort): | |
612 | tmap = {} | |
613 | for p in me._policies: | |
614 | for a in p.actions(master, targetdir, id, cohort): | |
615 | if a.target in tmap: | |
616 | tmap[a.target] = a.choose(tmap.get(a.target)) | |
617 | else: | |
618 | tmap[a.target] = a | |
619 | return tmap.values() | |
620 | ||
621 | And = K('and') - D('{') - R(Policy) - D('}') | |
622 | And.setParseAction(lambda s, l, t: AndPolicy(t[0])) | |
623 | ||
624 | class OrPolicy (FormatPolicy): | |
625 | """ | |
626 | A FormatPolicy which tries other policies and uses the first that accepts. | |
627 | ||
628 | Each subsidiary policy is invoked in turn. If any accepts, the actions it | |
629 | proposes are turned and no further policies are invoked. If none accepts | |
630 | then the file is rejected. | |
631 | """ | |
632 | ||
633 | def __init__(me, policies): | |
634 | me._policies = policies | |
635 | ||
636 | def setcategory(me, cat): | |
637 | me.cat = cat | |
638 | for p in me._policies: | |
639 | p.setcategory(cat) | |
640 | ||
641 | def actions(me, master, targetdir, id, cohort): | |
642 | for p in me._policies: | |
643 | aa = p.actions(master, targetdir, id, cohort) | |
644 | if aa: | |
645 | return aa | |
646 | else: | |
647 | return [] | |
648 | ||
649 | Or = K('or') - D('{') - R(Policy) - D('}') | |
650 | Or.setParseAction(lambda s, l, t: OrPolicy(t[0])) | |
651 | ||
652 | class AcceptPolicy (FormatPolicy): | |
653 | """ | |
654 | A FormatPolicy which copies files in a particular format. | |
655 | ||
656 | If all of the files in a cohort are recognized as being in a particular | |
657 | format (including this one), then accept it with a CopyAction; otherwise | |
658 | reject. | |
659 | """ | |
660 | ||
661 | def __init__(me, format): | |
662 | me._format = format | |
663 | ||
664 | def setcategory(me, cat): | |
665 | if me._format.CATEGORY is not cat: | |
666 | raise ValueError, \ | |
667 | "Accept format `%s' has category `%s', not `%s'" % \ | |
668 | (me._format.__class__.__name__, | |
669 | me._format.CATEGORY.name, cat.name) | |
670 | me.cat = cat | |
671 | ||
672 | def actions(me, master, targetdir, id, cohort): | |
673 | if me._format.check(id) and \ | |
674 | all(me._format.check(cid) for f, cid in cohort): | |
675 | return [CopyAction(master, targetdir)] | |
676 | else: | |
677 | return [] | |
678 | ||
679 | Accept = K('accept') - Format | |
680 | Accept.setParseAction(lambda s, l, t: AcceptPolicy(t[0])) | |
681 | ||
682 | class ConvertPolicy (FormatPolicy): | |
683 | """ | |
684 | A FormatPolicy which copies files in a particular format or converts if | |
685 | necessary. | |
686 | """ | |
687 | def __init__(me, format): | |
688 | me._format = format | |
689 | ||
690 | def setcategory(me, cat): | |
691 | if me._format.CATEGORY is not cat: | |
692 | raise ValueError, \ | |
693 | "Accept format `%s' has category `%s', not `%s'" % \ | |
694 | (me._format.__class__.__name__, | |
695 | me._format.CATEGORY.name, cat.name) | |
696 | me.cat = cat | |
697 | ||
698 | def actions(me, master, targetdir, id, cohort): | |
699 | if me._format.check(id): | |
700 | return [CopyAction(master, targetdir)] | |
701 | else: | |
702 | return [ConvertAction(master, targetdir, id, me._format)] | |
703 | ||
704 | Convert = K('convert') - Format | |
705 | Convert.setParseAction(lambda s, l, t: ConvertPolicy(t[0])) | |
706 | ||
707 | Policy << (And | Or | Accept | Convert) | |
708 | ||
709 | ###-------------------------------------------------------------------------- | |
710 | ### Audio handling, based on GStreamer. | |
711 | ||
712 | def make_element(factory, name = None, **props): | |
713 | "Return a new element from the FACTORY with the given NAME and PROPS." | |
714 | elt = GS.element_factory_make(factory, name) | |
715 | elt.set_properties(**props) | |
716 | return elt | |
717 | ||
718 | class GStreamerProgressEyecandy (ProgressEyecandy): | |
719 | """ | |
720 | Provide amusement while GStreamer is busy doing something. | |
721 | ||
722 | The GStreamerProgressEyecandy object is a context manager. Wrap it round | |
723 | your GStreamer loop to provide progress information for an operation. | |
724 | """ | |
725 | ||
726 | def __init__(me, what, elt, **kw): | |
727 | """ | |
728 | Initialize a progress meter. | |
729 | ||
730 | WHAT is a prefix string to be written before the progress eyecandy | |
731 | itself. ELT is a GStreamer element to interrogate to find the progress | |
732 | information. | |
733 | """ | |
734 | me._elt = elt | |
735 | ProgressEyecandy.__init__(me, what, **kw) | |
736 | ||
737 | def _update(me): | |
738 | "Called by GLib main event loop to update the eyecandy." | |
739 | me.show() | |
740 | return True | |
741 | ||
742 | def _timer(me): | |
743 | """ | |
744 | Update the progress meter. | |
745 | ||
746 | This is called periodically by the GLib main event-processing loop. | |
747 | """ | |
748 | me.show() | |
749 | return True | |
750 | ||
751 | def progress(me): | |
752 | "Return the current progress as a pair (CURRENT, MAX)." | |
753 | ||
754 | ## Fetch the current progress information. We get the duration each | |
755 | ## time, because (particularly with VBR-encoded MP3 inputs) the estimated | |
756 | ## duration can change as we progress. Hopefully it settles down fairly | |
757 | ## soon. | |
758 | try: | |
759 | t, hunoz = me._elt.query_position(GS.FORMAT_TIME) | |
760 | end, hukairz = me._elt.query_duration(GS.FORMAT_TIME) | |
761 | return t, end | |
762 | except GS.QueryError: | |
763 | return None, None | |
764 | ||
765 | def __enter__(me): | |
766 | "Enter context: attach progress meter display." | |
767 | ||
768 | ## If we're not showing pointless frippery, don't bother at all. | |
769 | if not STATUS.eyecandyp: | |
770 | return | |
771 | ||
772 | ## Update regularly. The pipeline runs asynchronously. | |
773 | me._id = G.timeout_add(200, me._update) | |
774 | ||
775 | def __exit__(me, ty, val, tb): | |
776 | "Leave context: remove display and report completion or failure." | |
777 | ||
778 | ## If we're not showing pointless frippery, there's nothing to remove. | |
779 | if STATUS.eyecandyp: | |
780 | G.source_remove(me._id) | |
781 | ||
782 | ## Report completion anyway. | |
783 | me.done(ty is None) | |
784 | ||
785 | ## As you were. | |
786 | return False | |
787 | ||
788 | class AudioIdentifier (object): | |
789 | """ | |
790 | Analyses and identifies an audio file. | |
791 | ||
792 | Important properties are: | |
793 | ||
794 | cap A capabilities structure describing the audio file data. The most | |
795 | interesting thing in here is probably its name, which is a MIME | |
796 | type describing the data. | |
797 | ||
798 | dcap A capabilities structure describing the decoded audio data. This | |
799 | is of interest during conversion. | |
800 | ||
801 | tags A dictionary containing metadata tags from the file. These are in | |
802 | GStreamer's encoding-independent format. | |
803 | ||
804 | bitrate An approximation to the stream's bitrate, in kilobits per second. | |
805 | This might be slow to work out for some files so it's computed on | |
806 | demand. | |
807 | """ | |
808 | ||
809 | def __init__(me, file, mime): | |
810 | "Initialize the object suitably for identifying FILE." | |
811 | ||
812 | ## Make some initial GStreamer objects. We'll want the pipeline later if | |
813 | ## we need to analyse a poorly tagged MP3 stream, so save it away. | |
814 | me._pipe = GS.Pipeline() | |
815 | me._file = file | |
816 | bus = me._pipe.get_bus() | |
817 | bus.add_signal_watch() | |
818 | loop = G.MainLoop() | |
819 | ||
820 | ## The basic recognition kit is based around `decodebin'. We must keep | |
821 | ## it happy by giving it sinks for the streams it's found, which it | |
822 | ## announces asynchronously. | |
823 | source = make_element('filesrc', 'file', location = file) | |
824 | decoder = make_element('decodebin', 'decode') | |
825 | sink = make_element('fakesink') | |
826 | def decoder_pad_arrived(elt, pad): | |
827 | if pad.get_caps()[0].get_name().startswith('audio/'): | |
828 | elt.link_pads(pad.get_name(), sink, 'sink') | |
829 | dpaid = decoder.connect('pad-added', decoder_pad_arrived) | |
830 | me._pipe.add(source, decoder, sink) | |
831 | GS.element_link_many(source, decoder) | |
832 | ||
833 | ## Arrange to collect tags from the pipeline's bus as they're reported. | |
834 | ## If we reuse the pipeline later, we'll want different bus-message | |
835 | ## handling, so make sure we can take the signal handler away. | |
836 | tags = {} | |
837 | fail = [] | |
838 | def bus_message(bus, msg): | |
839 | if msg.type == GS.MESSAGE_ERROR: | |
840 | fail[:] = (ValueError, msg.structure['debug'], None) | |
841 | loop.quit() | |
842 | elif msg.type == GS.MESSAGE_STATE_CHANGED: | |
843 | if msg.structure['new-state'] == GS.STATE_PAUSED and \ | |
844 | msg.src == me._pipe: | |
845 | loop.quit() | |
846 | elif msg.type == GS.MESSAGE_TAG: | |
847 | tags.update(msg.structure) | |
848 | bmid = bus.connect('message', bus_message) | |
849 | ||
850 | ## We want to identify the kind of stream this is. (Hmm. The MIME type | |
851 | ## recognizer has already done this work, but GStreamer is probably more | |
852 | ## reliable.) The `decodebin' has a `typefind' element inside which will | |
853 | ## announce the identified media type. All we need to do is find it and | |
854 | ## attach a signal handler. (Note that the handler might be run in the | |
855 | ## thread context of the pipeline element, but Python's GIL will keep | |
856 | ## things from being too awful.) | |
857 | me.cap = None | |
858 | me.dcap = None | |
859 | for e in decoder.elements(): | |
860 | if e.get_factory().get_name() == 'typefind': | |
861 | tfelt = e | |
862 | break | |
863 | else: | |
864 | assert False, 'failed to find typefind element' | |
865 | ||
866 | ## Crank up most of the heavy machinery. The message handler will stop | |
867 | ## the loop when things seem to be sufficiently well underway. | |
868 | me._pipe.set_state(GS.STATE_PAUSED) | |
869 | loop.run() | |
870 | bus.disconnect(bmid) | |
871 | decoder.disconnect(dpaid) | |
872 | if fail: | |
873 | me._pipe.set_state(GS.STATE_NULL) | |
874 | raise fail[0], fail[1], fail[2] | |
875 | ||
876 | ## Store the collected tags. | |
877 | me.tags = tags | |
878 | ||
879 | ## Gather the capabilities. The `typefind' element knows the input data | |
880 | ## type. The 'decodebin' knows the raw data type. | |
881 | me.cap = tfelt.get_pad('src').get_negotiated_caps()[0] | |
882 | me.mime = set([mime, me.cap.get_name()]) | |
883 | me.dcap = sink.get_pad('sink').get_negotiated_caps()[0] | |
884 | ||
885 | ## If we found a plausible bitrate then stash it. Otherwise note that we | |
886 | ## failed. If anybody asks then we'll work it out then. | |
887 | if 'nominal-bitrate' in tags: | |
888 | me._bitrate = tags['nominal-bitrate']/1000 | |
889 | elif 'bitrate' in tags and tags['bitrate'] >= 80000: | |
890 | me._bitrate = tags['bitrate']/1000 | |
891 | else: | |
892 | me._bitrate = None | |
893 | ||
894 | ## The bitrate computation wants the file size. Ideally we'd want the | |
895 | ## total size of the frames' contents, but that seems hard to dredge | |
896 | ## out. If the framing overhead is small, this should be close enough | |
897 | ## for our purposes. | |
898 | me._bytes = OS.stat(file).st_size | |
899 | ||
900 | def __del__(me): | |
901 | "Close the pipeline down so we don't leak file descriptors." | |
902 | me._pipe.set_state(GS.STATE_NULL) | |
903 | ||
904 | @property | |
905 | def bitrate(me): | |
906 | """ | |
907 | Return the approximate bit-rate of the input file. | |
908 | ||
909 | This might take a while if we have to work it out the hard way. | |
910 | """ | |
911 | ||
912 | ## If we already know the answer then just return it. | |
913 | if me._bitrate is not None: | |
914 | return me._bitrate | |
915 | ||
916 | ## Make up a new main loop. | |
917 | loop = G.MainLoop() | |
918 | ||
919 | ## Watch for bus messages. We'll stop when we reach the end of the | |
920 | ## stream: then we'll have a clear idea of how long the track was. | |
921 | fail = [] | |
922 | def bus_message(bus, msg): | |
923 | if msg.type == GS.MESSAGE_ERROR: | |
924 | fail[:] = (ValueError, msg.structure['debug'], None) | |
925 | loop.quit() | |
926 | elif msg.type == GS.MESSAGE_EOS: | |
927 | loop.quit() | |
928 | bus = me._pipe.get_bus() | |
929 | bmid = bus.connect('message', bus_message) | |
930 | ||
931 | ## Get everything moving, and keep the user amused while we work. | |
932 | me._pipe.set_state(GS.STATE_PLAYING) | |
933 | with GStreamerProgressEyecandy(filestatus(file, 'measure bitrate') % | |
934 | me._pipe, | |
935 | silentp = True): | |
936 | loop.run() | |
937 | bus.disconnect(bmid) | |
938 | if fail: | |
939 | me._pipe.set_state(GS.STATE_NULL) | |
940 | raise fail[0], fail[1], fail[2] | |
941 | ||
942 | ## Now we should be able to find out our position accurately and work out | |
943 | ## a bitrate. Cache it in case anybody asks again. | |
944 | t, hukairz = me._pipe.query_position(GS.FORMAT_TIME) | |
945 | me._bitrate = int(8*me._bytes*1e6/t) | |
946 | ||
947 | ## Done. | |
948 | return me._bitrate | |
949 | ||
950 | class AudioFormat (BaseFormat): | |
951 | """ | |
952 | An AudioFormat is a kind of Format specialized for audio files. | |
953 | ||
954 | Format checks are done on an AudioIdentifier object. | |
955 | """ | |
956 | ||
957 | PROPS = prop('bitrate', Num) | |
958 | ||
959 | ## libmagic reports `application/ogg' for Ogg Vorbis files. We've switched | |
960 | ## to GIO now, which reports either `audio/ogg' or `audio/x-vorbis+ogg' | |
961 | ## depending on how thorough it's trying to be. Still, it doesn't do any | |
962 | ## harm here; the main risk is picking up Ogg Theora files by accident, and | |
963 | ## we'll probably be able to extract the audio from them anyway. | |
964 | CATEGORY = FileCategory('audio', ['audio/*', 'application/ogg'], | |
965 | AudioIdentifier) | |
966 | ||
967 | def __init__(me, bitrate = None): | |
968 | "Construct an object, requiring an approximate bitrate." | |
969 | me.bitrate = bitrate | |
970 | ||
971 | def check(me, id): | |
972 | """ | |
973 | Return whether the AudioIdentifier ID is suitable for our purposes. | |
974 | ||
975 | Subclasses can either override this method or provide a property | |
976 | `MIMETYPES', which is a list (other thing that implements `__contains__') | |
977 | of GStreamer MIME types matching this format. | |
978 | """ | |
979 | return id.mime & me.MIMETYPES and \ | |
980 | (me.bitrate is None or id.bitrate <= me.bitrate * sqrt(2)) | |
981 | ||
982 | def encoder(me): | |
983 | """ | |
984 | Constructs a GStreamer element to encode audio input. | |
985 | ||
986 | Subclasses can either override this method (or replace `encode' | |
987 | entirely), or provide a method `encoder_chain' which returns a list of | |
988 | elements to be linked together in sequence. The first element in the | |
989 | chain must have a pad named `sink' and the last must have a pad named | |
990 | `src'. | |
991 | """ | |
992 | elts = me.encoder_chain() | |
993 | bin = GS.Bin() | |
994 | bin.add(*elts) | |
995 | GS.element_link_many(*elts) | |
996 | bin.add_pad(GS.GhostPad('sink', elts[0].get_pad('sink'))) | |
997 | bin.add_pad(GS.GhostPad('src', elts[-1].get_pad('src'))) | |
998 | return bin | |
999 | ||
1000 | def convert(me, master, id, target): | |
1001 | """ | |
1002 | Encode audio from MASTER, already identified as ID, writing it to TARGET. | |
1003 | ||
1004 | See `encoder' for subclasses' responsibilities. | |
1005 | """ | |
1006 | ||
1007 | ## Construct the necessary equipment. | |
1008 | pipe = GS.Pipeline() | |
1009 | bus = pipe.get_bus() | |
1010 | bus.add_signal_watch() | |
1011 | loop = G.MainLoop() | |
1012 | ||
1013 | ## Make sure that there isn't anything in the way of our output. We're | |
1014 | ## going to write to a scratch file so that we don't get confused by | |
1015 | ## half-written rubbish left by a crashed program. | |
1016 | new = target + '.new' | |
1017 | try: | |
1018 | OS.unlink(new) | |
1019 | except OSError, err: | |
1020 | if err.errno != E.ENOENT: | |
1021 | raise | |
1022 | ||
1023 | ## Piece together our pipeline. The annoying part is that the | |
1024 | ## `decodebin' doesn't have any source pads yet, so our chain is in two | |
1025 | ## halves for now. | |
1026 | source = make_element('filesrc', 'source', location = master) | |
1027 | decoder = make_element('decodebin', 'decode') | |
1028 | convert = make_element('audioconvert', 'convert') | |
1029 | encoder = me.encoder() | |
1030 | sink = make_element('filesink', 'sink', location = new) | |
1031 | pipe.add(source, decoder, convert, encoder, sink) | |
1032 | GS.element_link_many(source, decoder) | |
1033 | GS.element_link_many(convert, encoder, sink) | |
1034 | ||
1035 | ## Some decoders (e.g., the AC3 decoder) include channel-position | |
1036 | ## indicators in their output caps. The Vorbis encoder interferes with | |
1037 | ## this, and you end up with a beautifully encoded mono signal from a | |
1038 | ## stereo source. From a quick butchers at the `vorbisenc' source, I | |
1039 | ## /think/ that this is only a problem with stereo signals: mono signals | |
1040 | ## are mono already, and `vorbisenc' accepts channel positions if there | |
1041 | ## are more than two channels. | |
1042 | ## | |
1043 | ## So we have this bodge. We already collected the decoded audio caps | |
1044 | ## during identification. So if we see 2-channel audio with channel | |
1045 | ## positions, we strip the positions off forcibly by adding a filter. | |
1046 | if id.dcap.get_name().startswith('audio/x-raw-') and \ | |
1047 | id.dcap.has_field('channels') and \ | |
1048 | id.dcap['channels'] == 2 and \ | |
1049 | id.dcap.has_field('channel-positions'): | |
1050 | dcap = GS.Caps() | |
1051 | c = id.dcap.copy() | |
1052 | c.remove_field('channel-positions') | |
1053 | dcap.append(c) | |
1054 | else: | |
1055 | dcap = None | |
1056 | ||
1057 | ## Hook onto the `decodebin' so we can link together the two halves of | |
1058 | ## our encoding chain. For now, we'll hope that there's only one audio | |
1059 | ## stream in there, and just throw everything else away. | |
1060 | def decoder_pad_arrived(elt, pad): | |
1061 | if pad.get_caps()[0].get_name().startswith('audio/'): | |
1062 | if dcap: | |
1063 | elt.link_pads_filtered(pad.get_name(), convert, 'sink', dcap) | |
1064 | else: | |
1065 | elt.link_pads(pad.get_name(), convert, 'sink') | |
1066 | decoder.connect('pad-added', decoder_pad_arrived) | |
1067 | ||
1068 | ## Watch the bus for completion messages. | |
1069 | fail = [] | |
1070 | def bus_message(bus, msg): | |
1071 | if msg.type == GS.MESSAGE_ERROR: | |
1072 | fail[:] = (ValueError, msg.structure['debug'], None) | |
1073 | loop.quit() | |
1074 | elif msg.type == GS.MESSAGE_EOS: | |
1075 | loop.quit() | |
1076 | bmid = bus.connect('message', bus_message) | |
1077 | ||
1078 | ## Get everything ready and let it go. | |
1079 | pipe.set_state(GS.STATE_PLAYING) | |
1080 | with GStreamerProgressEyecandy(filestatus(master, | |
1081 | 'convert to %s' % me.NAME), | |
1082 | pipe): | |
1083 | loop.run() | |
1084 | pipe.set_state(GS.STATE_NULL) | |
1085 | if fail: | |
1086 | raise fail[0], fail[1], fail[2] | |
1087 | ||
1088 | ## Fix up the output file if we have to. | |
1089 | me.fixup(new) | |
1090 | ||
1091 | ## We're done. | |
1092 | OS.rename(new, target) | |
1093 | ||
1094 | class OggVorbisFormat (AudioFormat): | |
1095 | "AudioFormat object for Ogg Vorbis." | |
1096 | ||
1097 | ## From http://en.wikipedia.org/wiki/Vorbis | |
1098 | QMAP = [(-1, 45), ( 0, 64), ( 1, 80), ( 2, 96), | |
1099 | ( 3, 112), ( 4, 128), ( 5, 160), ( 6, 192), | |
1100 | ( 7, 224), ( 8, 256), ( 9, 320), (10, 500)] | |
1101 | ||
1102 | NAME = 'Ogg Vorbis' | |
1103 | MIMETYPES = set(['application/ogg', 'audio/x-vorbis', 'audio/ogg', | |
1104 | 'audio/x-vorbis+ogg']) | |
1105 | EXT = 'ogg' | |
1106 | ||
1107 | def encoder_chain(me): | |
1108 | for q, br in me.QMAP: | |
1109 | if br >= me.bitrate: | |
1110 | break | |
1111 | else: | |
1112 | raise ValueError, 'no suitable quality setting found' | |
1113 | return [make_element('vorbisenc', | |
1114 | quality = q/10.0), | |
1115 | make_element('oggmux')] | |
1116 | ||
1117 | defformat('ogg-vorbis', OggVorbisFormat) | |
1118 | ||
1119 | class MP3Format (AudioFormat): | |
1120 | "AudioFormat object for MP3." | |
1121 | ||
1122 | NAME = 'MP3' | |
1123 | MIMETYPES = set(['audio/mpeg']) | |
1124 | EXT = 'mp3' | |
1125 | ||
1126 | def encoder_chain(me): | |
1127 | return [make_element('lame', | |
1128 | vbr_mean_bitrate = me.bitrate, | |
1129 | vbr = 4), | |
1130 | make_element('xingmux'), | |
1131 | make_element('id3v2mux')] | |
1132 | ||
1133 | def fixup(me, path): | |
1134 | """ | |
1135 | Fix up MP3 files. | |
1136 | ||
1137 | GStreamer produces ID3v2 tags, but not ID3v1. This seems unnecessarily | |
1138 | unkind to stupid players. | |
1139 | """ | |
1140 | tag = E3.Tag() | |
1141 | tag.link(path) | |
1142 | tag.setTextEncoding(E3.UTF_8_ENCODING) | |
1143 | try: | |
1144 | tag.update(E3.ID3_V1_1) | |
1145 | except (UnicodeEncodeError, E3.tag.GenreException): | |
1146 | pass | |
1147 | ||
1148 | defformat('mp3', MP3Format) | |
1149 | ||
1150 | ###-------------------------------------------------------------------------- | |
1151 | ### Image handling, based on the Python Imaging Library. | |
1152 | ||
1153 | class ImageIdentifier (object): | |
1154 | """ | |
1155 | Analyses and identifies an image file. | |
1156 | ||
1157 | Simply leaves an Image object in the `img' property which can be inspected. | |
1158 | """ | |
1159 | ||
1160 | def __init__(me, file, mime): | |
1161 | ||
1162 | ## Get PIL to open the file. It will magically work out what kind of | |
1163 | ## file it is. | |
1164 | try: | |
1165 | me.img = I.open(file) | |
1166 | except IOError, exc: | |
1167 | ||
1168 | ## Unhelpful thing to raise on identification failure. We can | |
1169 | ## distinguish this from an actual I/O error because it doesn't have an | |
1170 | ## `errno'. | |
1171 | if exc.errno is None: | |
1172 | raise IdentificationFailure | |
1173 | raise | |
1174 | ||
1175 | me.mime = set([mime]) | |
1176 | ||
1177 | class ImageFormat (BaseFormat): | |
1178 | """ | |
1179 | An ImageFormat is a kind of Format specialized for image files. | |
1180 | ||
1181 | Subclasses don't need to provide anything other than the properties | |
1182 | required by all concrete Format subclasses. However, there is a | |
1183 | requirement that the `NAME' property match PIL's `format' name for the | |
1184 | format. | |
1185 | """ | |
1186 | ||
1187 | PROPS = prop('size', Num) | |
1188 | CATEGORY = FileCategory('image', ['image/*'], ImageIdentifier) | |
1189 | ||
1190 | def __init__(me, size = None, **kw): | |
1191 | """ | |
1192 | Initialize an ImageFormat object. | |
1193 | ||
1194 | Additional keywords are used when encoding, and may be recognized by | |
1195 | enhanced `check' methods in subclasses. | |
1196 | """ | |
1197 | me._size = size | |
1198 | me._props = kw | |
1199 | ||
1200 | def check(me, id): | |
1201 | "Check whether the ImageIdentifier ID matches our requirements." | |
1202 | return id.img.format == me.NAME and \ | |
1203 | (me._size is None or | |
1204 | (id.img.size[0] <= me._size and | |
1205 | id.img.size[1] <= me._size)) | |
1206 | ||
1207 | def convert(me, master, id, target): | |
1208 | "Encode the file MASTER, identified as ID, writing the result to TARGET." | |
1209 | ||
1210 | ## Write to a scratch file. | |
1211 | new = target + '.new' | |
1212 | ||
1213 | ## The ImageIdentifier already contains a copy of the open file. It | |
1214 | ## would be wasteful not to use it. | |
1215 | img = id.img | |
1216 | STATUS.set(filestatus(master, 'convert to %s' % me.NAME)) | |
1217 | ||
1218 | ## If there's a stated maximum size then scale the image down to match. | |
1219 | ## But thumbnailing clobbers the original, so take a copy. | |
1220 | if me._size is not None and \ | |
1221 | (img.size[0] > me._size or img.size[1] > me._size): | |
1222 | img = img.copy() | |
1223 | img.thumbnail((me._size, me._size), I.ANTIALIAS) | |
1224 | ||
1225 | ## Write the output image. | |
1226 | img.save(new, me.NAME, **me._props) | |
1227 | ||
1228 | ## Fix it up if necessary. | |
1229 | me.fixup(new) | |
1230 | ||
1231 | ## We're done. | |
1232 | OS.rename(new, target) | |
1233 | STATUS.commit() | |
1234 | ||
1235 | class JPEGFormat (ImageFormat): | |
1236 | """ | |
1237 | Image format for JPEG (actually JFIF) files. | |
1238 | ||
1239 | Interesting properties to set: | |
1240 | ||
1241 | optimize | |
1242 | If present, take a second pass to select optimal encoder settings. | |
1243 | ||
1244 | progression | |
1245 | If present, make a progressive file. | |
1246 | ||
1247 | quality Integer from 1--100 (worst to best); default is 75. | |
1248 | """ | |
1249 | EXT = 'jpg' | |
1250 | NAME = 'JPEG' | |
1251 | PROPS = prop('optimize', None) \ | |
1252 | | prop('progressive', None, 'progression') \ | |
1253 | | prop('quality', Num) | |
1254 | ||
1255 | defformat('jpeg', JPEGFormat) | |
1256 | ||
1257 | class PNGFormat (ImageFormat): | |
1258 | """ | |
1259 | Image format for PNG files. | |
1260 | ||
1261 | Interesting properties: | |
1262 | ||
1263 | optimize | |
1264 | If present, make a special effort to minimize the output file. | |
1265 | """ | |
1266 | EXT = 'png' | |
1267 | NAME = 'PNG' | |
1268 | PROPS = prop('optimize', None) | |
1269 | ||
1270 | defformat('png', PNGFormat) | |
1271 | ||
1272 | class BMPFormat (ImageFormat): | |
1273 | """ | |
1274 | Image format for Windows BMP files, as used by RockBox. | |
1275 | ||
1276 | No additional properties. | |
1277 | """ | |
1278 | NAME = 'BMP' | |
1279 | EXT = 'bmp' | |
1280 | ||
1281 | defformat('bmp', BMPFormat) | |
1282 | ||
1283 | ###-------------------------------------------------------------------------- | |
1284 | ### The directory grobbler. | |
1285 | ||
1286 | class Grobbler (object): | |
1287 | """ | |
1288 | The directory grobbler copies a directory tree, converting files. | |
1289 | """ | |
1290 | ||
1291 | def __init__(me, policies, noact = False): | |
1292 | """ | |
1293 | Create a new Grobbler, working with the given POLICIES. | |
1294 | """ | |
1295 | me._pmap = {} | |
1296 | me._noact = noact | |
1297 | for p in policies: | |
1298 | me._pmap.setdefault(p.cat, []).append(p) | |
1299 | me._dirs = [] | |
1300 | ||
1301 | def _grobble_file(me, master, targetdir, cohorts): | |
1302 | """ | |
1303 | Convert MASTER, writing the result to TARGETDIR. | |
1304 | ||
1305 | The COHORTS are actually (CAT, ID, COHORT) triples, where a COHORT is a | |
1306 | list of (FILENAME, ID) pairs. | |
1307 | ||
1308 | Since this function might convert the MASTER file, the caller doesn't | |
1309 | know the name of the output files, so we return then as a list. | |
1310 | """ | |
1311 | ||
1312 | done = set() | |
1313 | st_m = OS.stat(master) | |
1314 | ||
1315 | ## Work through each category listed and apply its policy. | |
1316 | for cat, id, cohort in cohorts: | |
1317 | ||
1318 | ## Go through the category's policies and see if any match. If we fail | |
1319 | ## here, see if there are more categories to try. | |
1320 | for pol in me._pmap[cat]: | |
1321 | acts = pol.actions(master, targetdir, id, cohort) | |
1322 | if acts: break | |
1323 | else: | |
1324 | continue | |
1325 | ||
1326 | ## Work through the targets one by one. | |
1327 | for a in acts: | |
1328 | done.add(a.target) | |
1329 | ||
1330 | ## Find out whether the target file already exists and is up-to-date | |
1331 | ## with respect to the master. (Caution here with low-resolution | |
1332 | ## timestamps.) If it's OK, then just move on. | |
1333 | try: | |
1334 | st_t = OS.stat(a.target) | |
1335 | if st_m.st_mtime < st_t.st_mtime or \ | |
1336 | (st_m.st_ino, st_m.st_dev) == (st_t.st_ino, st_t.st_dev): | |
1337 | continue | |
1338 | except OSError, err: | |
1339 | if err.errno not in (E.ENOENT, E.ENOTDIR): | |
1340 | raise | |
1341 | ||
1342 | ## We have real work to do. If there's a current status message, | |
1343 | ## it's the containing directory so flush it so that people know | |
1344 | ## where we are. | |
1345 | STATUS.commit() | |
1346 | ||
1347 | ## Remove the target. (A hardlink will fail if the target already | |
1348 | ## exists.) | |
1349 | if not me._noact: | |
1350 | try: | |
1351 | OS.unlink(a.target) | |
1352 | except OSError, err: | |
1353 | if err.errno not in (E.ENOENT, E.ENOTDIR): | |
1354 | raise | |
1355 | ||
1356 | ## Do whatever it is we decided to do. | |
1357 | if me._noact: | |
1358 | STATUS.commit(filestatus(master, a)) | |
1359 | else: | |
1360 | a.perform() | |
1361 | ||
1362 | ## We're done. Return the names of the targets. | |
1363 | return list(done) | |
1364 | ||
1365 | @contextmanager | |
1366 | def _wrap(me, masterfile): | |
1367 | """ | |
1368 | Handle exceptions found while trying to convert a particular file or | |
1369 | directory. | |
1370 | """ | |
1371 | ||
1372 | try: | |
1373 | yield masterfile | |
1374 | ||
1375 | ## Something bad happened. Report the error, but continue. (This list | |
1376 | ## of exceptions needs a lot of work.) | |
1377 | except (IOError, OSError), exc: | |
1378 | STATUS.clear() | |
1379 | STATUS.commit(filestatus(masterfile, 'failed (%s)' % exc)) | |
1380 | me._broken.append((masterfile, exc)) | |
1381 | ||
1382 | def _grobble_dir(me, master, target): | |
1383 | """ | |
1384 | Recursively convert files in MASTER, writing them to TARGET. | |
1385 | """ | |
1386 | ||
1387 | ## Make sure the TARGET exists and is a directory. It's a fundamental | |
1388 | ## assumption of this program that the entire TARGET tree is disposable, | |
1389 | ## so if something exists but isn't a directory, we should kill it. | |
1390 | if OS.path.isdir(target): | |
1391 | pass | |
1392 | else: | |
1393 | if OS.path.exists(target): | |
1394 | STATUS.commit(filestatus(target, 'clear nondirectory')) | |
1395 | if not me._noact: | |
1396 | OS.unlink(target) | |
1397 | STATUS.commit(filestatus(target, 'create directory')) | |
1398 | if not me._noact: | |
1399 | OS.mkdir(target) | |
1400 | ||
1401 | ## Keep a list of things in the target. As we convert files, we'll check | |
1402 | ## them off. Anything left over is rubbish and needs to be deleted. | |
1403 | checklist = {} | |
1404 | try: | |
1405 | for i in OS.listdir(target): | |
1406 | checklist[i] = False | |
1407 | except OSError, err: | |
1408 | if err.errno not in (E.ENOENT, E.ENOTDIR): | |
1409 | raise | |
1410 | ||
1411 | ## Keep track of the files in each category. | |
1412 | catmap = {} | |
1413 | todo = [] | |
1414 | done = [] | |
1415 | ||
1416 | ## Work through the master files. | |
1417 | for f in sorted(OS.listdir(master)): | |
1418 | ||
1419 | ## If the killswitch has been pulled then stop. The whole idea is that | |
1420 | ## we want to cause a clean shutdown if possible, so we don't want to | |
1421 | ## do it in the middle of encoding because the encoding effort will | |
1422 | ## have been wasted. This is the only place we need to check. If | |
1423 | ## we've exited the loop, then clearing old files will probably be | |
1424 | ## fast, and we'll either end up here when the recursive call returns | |
1425 | ## or we'll be in the same boat as before, clearing old files, only up | |
1426 | ## a level. If worst comes to worst, we'll be killed forcibly | |
1427 | ## somewhere inside `SH.rmtree', and that can continue where it left | |
1428 | ## off. | |
1429 | if KILLSWITCH.is_set(): | |
1430 | return | |
1431 | ||
1432 | ## Do something with the file. | |
1433 | with me._wrap(OS.path.join(master, f)) as masterfile: | |
1434 | ||
1435 | ## If it's a directory then grobble it recursively. Keep the user | |
1436 | ## amused by telling him where we are in the tree. | |
1437 | if OS.path.isdir(masterfile): | |
1438 | me._dirs.append(f) | |
1439 | STATUS.set('/'.join(me._dirs)) | |
1440 | try: | |
1441 | done += me._grobble_dir(masterfile, OS.path.join(target, f)) | |
1442 | finally: | |
1443 | me._dirs.pop() | |
1444 | STATUS.set('/'.join(me._dirs)) | |
1445 | ||
1446 | ## Otherwise it's a file. Work out what kind, and stash it under | |
1447 | ## the appropriate categories. Later, we'll apply policy to the | |
1448 | ## files, by category, and work out what to do with them all. | |
1449 | else: | |
1450 | gf = GIO.File(masterfile) | |
1451 | mime = gf.query_info('standard::content-type').get_content_type() | |
1452 | cats = [] | |
1453 | for cat in me._pmap.iterkeys(): | |
1454 | id = cat.identify(masterfile, mime) | |
1455 | if id is None: continue | |
1456 | catmap.setdefault(cat, []).append((masterfile, id)) | |
1457 | cats.append((cat, id)) | |
1458 | if not cats: | |
1459 | catmap.setdefault(None, []).append((masterfile, id)) | |
1460 | todo.append((masterfile, cats)) | |
1461 | ||
1462 | ## Work through the categorized files to see what actions to do for | |
1463 | ## them. | |
1464 | for masterfile, cats in todo: | |
1465 | with me._wrap(masterfile): | |
1466 | done += me._grobble_file(masterfile, target, | |
1467 | [(cat, id, catmap[cat]) | |
1468 | for cat, id in cats]) | |
1469 | ||
1470 | ## Check the results off the list so that we don't clear it later. | |
1471 | for f in done: | |
1472 | checklist[OS.path.basename(f)] = True | |
1473 | ||
1474 | ## Maybe there's stuff in the target which isn't accounted for. Delete | |
1475 | ## it: either the master has changed, or the policy for this target has | |
1476 | ## changed. Either way, the old files aren't wanted. | |
1477 | for f in checklist: | |
1478 | if not checklist[f]: | |
1479 | STATUS.commit(filestatus(f, 'clear bogus file')) | |
1480 | if not me._noact: | |
1481 | bogus = OS.path.join(target, f) | |
1482 | try: | |
1483 | if OS.path.isdir(bogus): | |
1484 | SH.rmtree(bogus) | |
1485 | else: | |
1486 | OS.unlink(bogus) | |
1487 | except OSError, err: | |
1488 | if err.errno != E.ENOENT: | |
1489 | raise | |
1490 | ||
1491 | ## Return the target name, so that it can be checked off. | |
1492 | return [target] | |
1493 | ||
1494 | def grobble(me, master, target): | |
1495 | """ | |
1496 | Convert MASTER, writing a directory tree TARGET. | |
1497 | ||
1498 | Returns a list of files which couldn't be converted. | |
1499 | """ | |
1500 | try: | |
1501 | me._broken = [] | |
1502 | me._grobble_dir(master, target) | |
1503 | return me._broken | |
1504 | finally: | |
1505 | del me._broken | |
1506 | ||
1507 | ###-------------------------------------------------------------------------- | |
1508 | ### Remaining parsing machinery. | |
1509 | ||
1510 | Type = K('type') - Name - D('{') - R(Policy) - D('}') | |
1511 | def build_type(s, l, t): | |
1512 | try: | |
1513 | cat = CATEGORYMAP[t[0]] | |
1514 | except KeyError: | |
1515 | raise P.ParseException(s, loc, "Unknown category `%s'" % t[0]) | |
1516 | pols = t[1] | |
1517 | if len(pols) == 1: pol = pols[0] | |
1518 | else: pol = AndPolicy(pols) | |
1519 | pol.setcategory(cat) | |
1520 | return pol | |
1521 | Type.setParseAction(build_type) | |
1522 | ||
1523 | TARGETS = [] | |
1524 | class TargetJob (object): | |
1525 | def __init__(me, targetdir, policies): | |
1526 | me.targetdir = targetdir | |
1527 | me.policies = policies | |
1528 | def perform(me): | |
1529 | TARGETS.append(me) | |
1530 | ||
1531 | Target = K('target') - String - D('{') - R(Type) - D('}') | |
1532 | def build_target(s, l, t): | |
1533 | return TargetJob(t[0], t[1]) | |
1534 | Target.setParseAction(build_target) | |
1535 | ||
1536 | VARS = { 'master': None } | |
1537 | class VarsJob (object): | |
1538 | def __init__(me, vars): | |
1539 | me.vars = vars | |
1540 | def perform(me): | |
1541 | for k, v in me.vars: | |
1542 | VARS[k] = v | |
1543 | ||
1544 | Var = prop('master', String) | |
1545 | Vars = K('vars') - D('{') - R(Var) - D('}') | |
1546 | def build_vars(s, l, t): | |
1547 | return VarsJob(t[0]) | |
1548 | Vars.setParseAction(build_vars) | |
1549 | ||
1550 | TopLevel = Vars | Target | |
1551 | Config = R(TopLevel) | |
1552 | Config.ignore(P.pythonStyleComment) | |
1553 | ||
1554 | ###-------------------------------------------------------------------------- | |
1555 | ### Command-line interface. | |
1556 | ||
1557 | QUIS = OS.path.basename(SYS.argv[0]) | |
1558 | ||
1559 | def moan(msg): | |
1560 | "Report a warning message to the user." | |
1561 | SYS.stderr.write('%s: %s\n' % (QUIS, msg)) | |
1562 | ||
1563 | def die(msg): | |
1564 | "Report a fatal error message to the user." | |
1565 | moan(msg) | |
1566 | SYS.exit(1) | |
1567 | ||
1568 | def parse_opts(args): | |
1569 | """ | |
1570 | Parse command-line arguments in ARGS. | |
1571 | ||
1572 | Returns a Grobbler object and the MASTER and TARGET directories to be | |
1573 | grobbled. | |
1574 | """ | |
1575 | ||
1576 | ## Build the option parser object. | |
1577 | op = OP.OptionParser(prog = QUIS, version = VERSION, | |
1578 | usage = '%prog [-t TIMEOUT] CONFIG', | |
1579 | description = """\ | |
1580 | Convert a directory tree of files according to the configuration file | |
1581 | CONFIG. | |
1582 | """) | |
1583 | ||
1584 | ## Timeout handling. | |
1585 | def cb_time(opt, ostr, arg, op): | |
1586 | m = RX.match(r'\s*(\d+)\s*([dhms]?)\s*', arg) | |
1587 | if not m: | |
1588 | raise OP.OptionValueerror, 'bad time value `%s\'' % arg | |
1589 | t, u = m.groups() | |
1590 | t = int(t) * { '': 1, 's': 1, 'm': 60, 'h': 3600, 'd': 86400 }[u] | |
1591 | setattr(op.values, opt.dest, t) | |
1592 | op.add_option('-t', '--timeout', type = 'string', metavar = 'SECS', | |
1593 | dest = 'timeout', | |
1594 | help = 'stop processing nicely after SECS', | |
1595 | action = 'callback', callback = cb_time) | |
1596 | op.add_option('-T', '--timeout-nasty', type = 'string', metavar = 'SECS', | |
1597 | dest = 'timeout_nasty', | |
1598 | help = 'stop processing unpleasantly after further SECS', | |
1599 | action = 'callback', callback = cb_time) | |
1600 | ||
1601 | ## Other options. | |
1602 | op.add_option('-i', '--interactive', action = 'store_true', dest = 'tty', | |
1603 | help = 'provide progress information') | |
1604 | op.add_option('-n', '--no-act', action = 'store_true', dest = 'noact', | |
1605 | help = 'don\'t actually modify the filesystem') | |
1606 | ||
1607 | ## Ready to rock. | |
1608 | op.set_defaults(formats = [], noact = False, | |
1609 | timeout = None, timeout_nasty = 300) | |
1610 | opts, args = op.parse_args(args) | |
1611 | ||
1612 | ## Check that we got the non-option arguments that we want. | |
1613 | if len(args) != 1: | |
1614 | op.error('wrong number of arguments') | |
1615 | ||
1616 | ## Act on the options. | |
1617 | if opts.tty: | |
1618 | STATUS.eyecandyp = True | |
1619 | if opts.timeout is not None: | |
1620 | to = TH.Thread(target = timeout, | |
1621 | args = (opts.timeout, opts.timeout_nasty)) | |
1622 | to.daemon = True | |
1623 | to.start() | |
1624 | ||
1625 | ## Parse the configuration file. | |
1626 | with open(args[0]) as conf: | |
1627 | jobs, = Config.parseFile(conf, True) | |
1628 | for j in jobs: | |
1629 | j.perform() | |
1630 | ||
1631 | return opts | |
1632 | ||
1633 | if __name__ == '__main__': | |
1634 | opts = parse_opts(SYS.argv[1:]) | |
1635 | if 'master' not in VARS: | |
1636 | die("no master directory set") | |
1637 | broken = [] | |
1638 | for t in TARGETS: | |
1639 | g = Grobbler(t.policies, opts.noact) | |
1640 | b = g.grobble(VARS['master'], t.targetdir) | |
1641 | broken += b | |
1642 | if broken: | |
1643 | moan('failed to convert some files:') | |
1644 | for file, exc in broken: | |
1645 | moan('%s: %s' % (file, exc)) | |
1646 | SYS.exit(1) | |
1647 | ||
1648 | ## This is basically a successful completion: we did what we were asked to | |
1649 | ## do. It seems polite to report a message, though. | |
1650 | ## | |
1651 | ## Why don't we have a nonzero exit status? The idea would be that a | |
1652 | ## calling script would be interested that we used up all of our time, and | |
1653 | ## not attempt to convert some other directory as well. But that doesn't | |
1654 | ## quite work. Such a script would need to account correctly for time we | |
1655 | ## had spent even if we complete successfully. And if the script is having | |
1656 | ## to watch the clock itself, it can do that without our help here. | |
1657 | if KILLSWITCH.is_set(): | |
1658 | moan('killed by timeout') | |
1659 | ||
1660 | ###----- That's all, folks -------------------------------------------------- |