f4ff73817499a7d9c1fc670b8a8e4e41085b46a9
[autoys] / gremlin / gremlin.in
1 #! @PYTHON@
2 ###
3 ### Convert a directory tree of audio files
4 ###
5 ### (c) 2010 Mark Wooding
6 ###
7
8 ###----- Licensing notice ---------------------------------------------------
9 ###
10 ### This file is part of the `autoys' audio tools collection.
11 ###
12 ### `autoys' is free software; you can redistribute it and/or modify
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 ###
17 ### `autoys' is distributed in the hope that it will be useful,
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
23 ### along with `autoys'; if not, write to the Free Software Foundation,
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 errno as E
34 import fnmatch as FN
35 import optparse as OP
36 import os as OS
37 import re as RX
38 import sys as SYS
39 import time as T
40 import shlex as L
41 import shutil as SH
42 import threading as TH
43 import unicodedata as UD
44 from math import sqrt, ceil
45 from contextlib import contextmanager
46
47 ## eyeD3 tag fettling.
48 import eyed3 as E3
49
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([])
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
66 VERSION = '@VERSION@'
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 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:
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:])
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()
248 eta = me._fmt_time(ceil((t - me._start)*(max - cur)/cur))
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()
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
412 name = 'format-spec'
413
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."
711 elt = GS.ElementFactory.make(factory, name)
712 if elt is None: raise ValueError, 'failed to make `%s\' element' % factory
713 elt.set_properties(**props)
714 return elt
715
716 def link_elements(elts):
717 "Link the elements ELTS together, in order."
718 e0 = None
719 for e1 in elts:
720 if e0 is not None: e0.link(e1)
721 e0 = e1
722
723 def bin_children(bin):
724 "Iterate over the (direct) children of a BIN."
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
732
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.
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
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(100, 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 _prepare_pipeline(me):
823 pipe = GS.Pipeline()
824 bus = pipe.get_bus()
825
826 ## The basic recognition kit is based around `decodebin'. We must keep
827 ## it happy by giving it sinks for the streams it's found, which it
828 ## announces asynchronously.
829 source = make_element('filesrc', 'file', location = me._file)
830 decoder = make_element('decodebin', 'decode')
831 sink = make_element('fakesink')
832 def decoder_pad_arrived(elt, pad):
833 if pad.get_current_caps()[0].get_name().startswith('audio/'):
834 elt.link_pads(pad.get_name(), sink, 'sink')
835 decoder.connect('pad-added', decoder_pad_arrived)
836 for i in [source, decoder, sink]: pipe.add(i)
837 link_elements([source, decoder])
838
839 ## Done.
840 return pipe, bus, decoder, sink
841
842 def __init__(me, file, mime):
843 "Initialize the object suitably for identifying FILE."
844
845 me._file = file
846 pipe, bus, decoder, sink = me._prepare_pipeline()
847
848 ## Make some initial GStreamer objects. We'll want the pipeline later if
849 ## we need to analyse a poorly tagged MP3 stream, so save it away.
850 loop = G.MainLoop()
851
852 ## Arrange to collect tags from the pipeline's bus as they're reported.
853 tags = {}
854 fail = []
855 def bus_message(bus, msg):
856 ty, s = msg.type, msg.get_structure()
857 if ty == GS.MessageType.ERROR:
858 fail[:] = (ValueError, s['debug'], None)
859 loop.quit()
860 elif ty == GS.MessageType.STATE_CHANGED:
861 if s['new-state'] == GS.State.PAUSED and \
862 msg.src == pipe:
863 loop.quit()
864 elif ty == GS.MessageType.TAG:
865 tt = s['taglist']
866 for i in xrange(tt.n_tags()):
867 t = tt.nth_tag_name(i)
868 if tt.get_tag_size(t) != 1: continue
869 v = tt.get_value_index(t, 0)
870 tags[t] = v
871 bmid = bus.connect('message', bus_message)
872
873 ## We want to identify the kind of stream this is. (Hmm. The MIME type
874 ## recognizer has already done this work, but GStreamer is probably more
875 ## reliable.) The `decodebin' has a `typefind' element inside which will
876 ## announce the identified media type. All we need to do is find it and
877 ## attach a signal handler. (Note that the handler might be run in the
878 ## thread context of the pipeline element, but Python's GIL will keep
879 ## things from being too awful.)
880 me.cap = None
881 me.dcap = None
882 for e in bin_children(decoder):
883 if e.get_factory().get_name() == 'typefind':
884 tfelt = e
885 break
886 else:
887 assert False, 'failed to find typefind element'
888
889 ## Crank up most of the heavy machinery. The message handler will stop
890 ## the loop when things seem to be sufficiently well underway.
891 bus.add_signal_watch()
892 pipe.set_state(GS.State.PAUSED)
893 loop.run()
894 bus.disconnect(bmid)
895 bus.remove_signal_watch()
896 if fail:
897 pipe.set_state(GS.State.NULL)
898 raise fail[0], fail[1], fail[2]
899
900 ## Store the collected tags.
901 me.tags = tags
902
903 ## Gather the capabilities. The `typefind' element knows the input data
904 ## type. The 'decodebin' knows the raw data type.
905 me.cap = tfelt.get_static_pad('src').get_allowed_caps()[0]
906 me.mime = set([mime, me.cap.get_name()])
907 me.dcap = sink.get_static_pad('sink').get_allowed_caps()[0]
908
909 ## If we found a plausible bitrate then stash it. Otherwise note that we
910 ## failed. If anybody asks then we'll work it out then.
911 if 'nominal-bitrate' in tags:
912 me._bitrate = tags['nominal-bitrate']/1000
913 elif 'bitrate' in tags and tags['bitrate'] >= 80000:
914 me._bitrate = tags['bitrate']/1000
915 else:
916 ok, n = pipe.query_duration(GS.Format.BYTES)
917 if ok: ok, t = pipe.query_duration(GS.Format.TIME)
918 if ok: me._bitrate = int((8e6*n)/t)
919 else: me._bitrate = None
920 pipe.set_state(GS.State.NULL)
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 pipeline and main loop.
935 pipe, bus, _, _ = me._prepare_pipeline()
936 loop = G.MainLoop()
937
938 ## Watch for bus messages. We'll stop when we reach the end of the
939 ## stream: then we'll have a clear idea of how long the track was.
940 fail = []
941 def bus_message(bus, msg):
942 ty, s = msg.type, msg.get_structure()
943 if ty == GS.MessageType.ERROR:
944 fail[:] = (ValueError, s['debug'], None)
945 loop.quit()
946 elif ty == GS.MessageType.EOS:
947 loop.quit()
948 bus = pipe.get_bus()
949 bmid = bus.connect('message', bus_message)
950
951 ## Get everything moving, and keep the user amused while we work.
952 bus.add_signal_watch()
953 pipe.set_state(GS.State.PLAYING)
954 with GStreamerProgressEyecandy(filestatus(me._file, 'measure bitrate'),
955 pipe, silentp = True):
956 loop.run()
957 bus.remove_signal_watch()
958 bus.disconnect(bmid)
959 if fail:
960 pipe.set_state(GS.State.NULL)
961 raise fail[0], fail[1], fail[2]
962 STATUS.clear()
963
964 ## The bitrate computation wants the file size. Ideally we'd want the
965 ## total size of the frames' contents, but that seems hard to dredge
966 ## out. If the framing overhead is small, this should be close enough
967 ## for our purposes.
968 bytes = OS.stat(me._file).st_size
969
970 ## Now we should be able to find out our position accurately and work out
971 ## a bitrate. Cache it in case anybody asks again.
972 ok, t = pipe.query_position(GS.Format.TIME)
973 assert ok, 'failed to discover bitrate'
974 me._bitrate = int(8*bytes*1e6/t)
975 pipe.set_state(GS.State.NULL)
976
977 ## Done.
978 return me._bitrate
979
980 class AudioFormat (BaseFormat):
981 """
982 An AudioFormat is a kind of Format specialized for audio files.
983
984 Format checks are done on an AudioIdentifier object.
985 """
986
987 PROPS = prop('bitrate', Num)
988
989 ## libmagic reports `application/ogg' for Ogg Vorbis files. We've switched
990 ## to GIO now, which reports either `audio/ogg' or `audio/x-vorbis+ogg'
991 ## depending on how thorough it's trying to be. Still, it doesn't do any
992 ## harm here; the main risk is picking up Ogg Theora files by accident, and
993 ## we'll probably be able to extract the audio from them anyway.
994 CATEGORY = FileCategory('audio', ['audio/*', 'application/ogg'],
995 AudioIdentifier)
996
997 def __init__(me, bitrate = None):
998 "Construct an object, requiring an approximate bitrate."
999 me.bitrate = bitrate
1000
1001 def check(me, id):
1002 """
1003 Return whether the AudioIdentifier ID is suitable for our purposes.
1004
1005 Subclasses can either override this method or provide a property
1006 `MIMETYPES', which is a list (other thing that implements `__contains__')
1007 of GStreamer MIME types matching this format.
1008 """
1009 return id.mime & me.MIMETYPES and \
1010 (me.bitrate is None or id.bitrate <= me.bitrate * sqrt(2))
1011
1012 def encoder(me):
1013 """
1014 Constructs a GStreamer element to encode audio input.
1015
1016 Subclasses can either override this method (or replace `encode'
1017 entirely), or provide a method `encoder_chain' which returns a list of
1018 elements to be linked together in sequence. The first element in the
1019 chain must have a pad named `sink' and the last must have a pad named
1020 `src'.
1021 """
1022 elts = me.encoder_chain()
1023 bin = GS.Bin()
1024 for i in elts: bin.add(i)
1025 link_elements(elts)
1026 bin.add_pad(GS.GhostPad('sink', elts[0].get_static_pad('sink')))
1027 bin.add_pad(GS.GhostPad('src', elts[-1].get_static_pad('src')))
1028 return bin
1029
1030 def convert(me, master, id, target):
1031 """
1032 Encode audio from MASTER, already identified as ID, writing it to TARGET.
1033
1034 See `encoder' for subclasses' responsibilities.
1035 """
1036
1037 ## Construct the necessary equipment.
1038 pipe = GS.Pipeline()
1039 bus = pipe.get_bus()
1040 loop = G.MainLoop()
1041
1042 ## Make sure that there isn't anything in the way of our output. We're
1043 ## going to write to a scratch file so that we don't get confused by
1044 ## half-written rubbish left by a crashed program.
1045 new = target + '.new'
1046 try:
1047 OS.unlink(new)
1048 except OSError, err:
1049 if err.errno != E.ENOENT:
1050 raise
1051
1052 ## Piece together our pipeline. The annoying part is that the
1053 ## `decodebin' doesn't have any source pads yet, so our chain is in two
1054 ## halves for now.
1055 source = make_element('filesrc', 'source', location = master)
1056 decoder = make_element('decodebin', 'decode')
1057 convert = make_element('audioconvert', 'convert')
1058 encoder = me.encoder()
1059 sink = make_element('filesink', 'sink', location = new)
1060 for i in [source, decoder, convert, encoder, sink]: pipe.add(i)
1061 link_elements([source, decoder])
1062 link_elements([convert, encoder, sink])
1063
1064 ## Some decoders (e.g., the AC3 decoder) include channel-position
1065 ## indicators in their output caps. The Vorbis encoder interferes with
1066 ## this, and you end up with a beautifully encoded mono signal from a
1067 ## stereo source. From a quick butchers at the `vorbisenc' source, I
1068 ## /think/ that this is only a problem with stereo signals: mono signals
1069 ## are mono already, and `vorbisenc' accepts channel positions if there
1070 ## are more than two channels.
1071 ##
1072 ## So we have this bodge. We already collected the decoded audio caps
1073 ## during identification. So if we see 2-channel audio with channel
1074 ## positions, we strip the positions off forcibly by adding a filter.
1075 if id.dcap.get_name().startswith('audio/x-raw-') and \
1076 id.dcap.has_field('channels') and \
1077 id.dcap['channels'] == 2 and \
1078 id.dcap.has_field('channel-positions'):
1079 dcap = GS.Caps()
1080 c = id.dcap.copy()
1081 c.remove_field('channel-positions')
1082 dcap.append(c)
1083 else:
1084 dcap = None
1085
1086 ## Hook onto the `decodebin' so we can link together the two halves of
1087 ## our encoding chain. For now, we'll hope that there's only one audio
1088 ## stream in there, and just throw everything else away.
1089 def decoder_pad_arrived(elt, pad):
1090 if pad.get_current_caps()[0].get_name().startswith('audio/'):
1091 if dcap:
1092 elt.link_pads_filtered(pad.get_name(), convert, 'sink', dcap)
1093 else:
1094 elt.link_pads(pad.get_name(), convert, 'sink')
1095 decoder.connect('pad-added', decoder_pad_arrived)
1096
1097 ## Watch the bus for completion messages.
1098 fail = []
1099 def bus_message(bus, msg):
1100 if msg.type == GS.MessageType.ERROR:
1101 fail[:] = (ValueError, msg.get_structure()['debug'], None)
1102 loop.quit()
1103 elif msg.type == GS.MessageType.EOS:
1104 loop.quit()
1105 bmid = bus.connect('message', bus_message)
1106
1107 ## Get everything ready and let it go.
1108 bus.add_signal_watch()
1109 pipe.set_state(GS.State.PLAYING)
1110 with GStreamerProgressEyecandy(filestatus(master,
1111 'convert to %s' % me.NAME),
1112 pipe):
1113 loop.run()
1114 pipe.set_state(GS.State.NULL)
1115 bus.remove_signal_watch()
1116 bus.disconnect(bmid)
1117 if fail:
1118 raise fail[0], fail[1], fail[2]
1119
1120 ## Fix up the output file if we have to.
1121 me.fixup(new)
1122
1123 ## We're done.
1124 OS.rename(new, target)
1125
1126 class OggVorbisFormat (AudioFormat):
1127 "AudioFormat object for Ogg Vorbis."
1128
1129 ## From https://en.wikipedia.org/wiki/Vorbis
1130 QMAP = [(-1, 45), ( 0, 64), ( 1, 80), ( 2, 96),
1131 ( 3, 112), ( 4, 128), ( 5, 160), ( 6, 192),
1132 ( 7, 224), ( 8, 256), ( 9, 320), (10, 500)]
1133
1134 NAME = 'Ogg Vorbis'
1135 MIMETYPES = set(['application/ogg', 'audio/x-vorbis', 'audio/ogg',
1136 'audio/x-vorbis+ogg'])
1137 EXT = 'ogg'
1138
1139 def encoder_chain(me):
1140 encprops = {}
1141 if me.bitrate is not None:
1142 for q, br in me.QMAP:
1143 if br >= me.bitrate:
1144 break
1145 else:
1146 raise ValueError, 'no suitable quality setting found'
1147 encprops['quality'] = q/10.0
1148 return [make_element('vorbisenc', **encprops),
1149 make_element('oggmux')]
1150
1151 defformat('ogg-vorbis', OggVorbisFormat)
1152
1153 class MP3Format (AudioFormat):
1154 "AudioFormat object for MP3."
1155
1156 NAME = 'MP3'
1157 MIMETYPES = set(['audio/mpeg'])
1158 EXT = 'mp3'
1159
1160 def encoder_chain(me):
1161 encprops = {}
1162 if me.bitrate is not None:
1163 encprops['bitrate'] = me.bitrate
1164 encprops['target'] = 'bitrate'
1165 else:
1166 encprops['quality'] = 4
1167 encprops['target'] = 'quality'
1168 return [make_element('lamemp3enc', quality = 4, **encprops),
1169 make_element('xingmux'),
1170 make_element('id3v2mux')]
1171
1172 def fixup(me, path):
1173 """
1174 Fix up MP3 files.
1175
1176 GStreamer produces ID3v2 tags, but not ID3v1. This seems unnecessarily
1177 unkind to stupid players.
1178 """
1179 f = E3.load(path)
1180 if f is None: return
1181 t = f.tag
1182 if t is None: return
1183 for v in [E3.id3.ID3_V2_3, E3.id3.ID3_V1]:
1184 try: f.tag.save(version = v)
1185 except (UnicodeEncodeError,
1186 E3.id3.GenreException,
1187 E3.id3.TagException):
1188 pass
1189
1190 defformat('mp3', MP3Format)
1191
1192 ###--------------------------------------------------------------------------
1193 ### Image handling, based on the Python Imaging Library.
1194
1195 class ImageIdentifier (object):
1196 """
1197 Analyses and identifies an image file.
1198
1199 Simply leaves an Image object in the `img' property which can be inspected.
1200 """
1201
1202 def __init__(me, file, mime):
1203
1204 ## Get PIL to open the file. It will magically work out what kind of
1205 ## file it is.
1206 try:
1207 me.img = I.open(file)
1208 except IOError, exc:
1209
1210 ## Unhelpful thing to raise on identification failure. We can
1211 ## distinguish this from an actual I/O error because it doesn't have an
1212 ## `errno'.
1213 if exc.errno is None:
1214 raise IdentificationFailure
1215 raise
1216
1217 me.mime = set([mime])
1218
1219 class ImageFormat (BaseFormat):
1220 """
1221 An ImageFormat is a kind of Format specialized for image files.
1222
1223 Subclasses don't need to provide anything other than the properties
1224 required by all concrete Format subclasses. However, there is a
1225 requirement that the `NAME' property match PIL's `format' name for the
1226 format.
1227 """
1228
1229 PROPS = prop('size', Num)
1230 CATEGORY = FileCategory('image', ['image/*'], ImageIdentifier)
1231
1232 def __init__(me, size = None, **kw):
1233 """
1234 Initialize an ImageFormat object.
1235
1236 Additional keywords are used when encoding, and may be recognized by
1237 enhanced `check' methods in subclasses.
1238 """
1239 me._size = size
1240 me._props = kw
1241
1242 def check(me, id):
1243 "Check whether the ImageIdentifier ID matches our requirements."
1244 return id.img.format == me.NAME and \
1245 (me._size is None or
1246 (id.img.size[0] <= me._size and
1247 id.img.size[1] <= me._size))
1248
1249 def convert(me, master, id, target):
1250 "Encode the file MASTER, identified as ID, writing the result to TARGET."
1251
1252 ## Write to a scratch file.
1253 new = target + '.new'
1254
1255 ## The ImageIdentifier already contains a copy of the open file. It
1256 ## would be wasteful not to use it.
1257 img = id.img
1258 STATUS.set(filestatus(master, 'convert to %s' % me.NAME))
1259
1260 ## If there's a stated maximum size then scale the image down to match.
1261 ## But thumbnailing clobbers the original, so take a copy.
1262 if me._size is not None and \
1263 (img.size[0] > me._size or img.size[1] > me._size):
1264 img = img.copy()
1265 img.thumbnail((me._size, me._size), I.ANTIALIAS)
1266
1267 ## Write the output image.
1268 img.save(new, me.NAME, **me._props)
1269
1270 ## Fix it up if necessary.
1271 me.fixup(new)
1272
1273 ## We're done.
1274 OS.rename(new, target)
1275 STATUS.commit()
1276
1277 class JPEGFormat (ImageFormat):
1278 """
1279 Image format for JPEG (actually JFIF) files.
1280
1281 Interesting properties to set:
1282
1283 optimize
1284 If present, take a second pass to select optimal encoder settings.
1285
1286 progressive
1287 If present, make a progressive file.
1288
1289 quality Integer from 1--100 (worst to best); default is 75.
1290 """
1291 EXT = 'jpg'
1292 NAME = 'JPEG'
1293 PROPS = prop('optimize', None) \
1294 | prop('progressive', None, 'progression') \
1295 | prop('quality', Num)
1296
1297 defformat('jpeg', JPEGFormat)
1298
1299 class PNGFormat (ImageFormat):
1300 """
1301 Image format for PNG files.
1302
1303 Interesting properties:
1304
1305 optimize
1306 If present, make a special effort to minimize the output file.
1307 """
1308 EXT = 'png'
1309 NAME = 'PNG'
1310 PROPS = prop('optimize', None)
1311
1312 defformat('png', PNGFormat)
1313
1314 class BMPFormat (ImageFormat):
1315 """
1316 Image format for Windows BMP files, as used by RockBox.
1317
1318 No additional properties.
1319 """
1320 NAME = 'BMP'
1321 EXT = 'bmp'
1322
1323 defformat('bmp', BMPFormat)
1324
1325 ###--------------------------------------------------------------------------
1326 ### Remaining parsing machinery.
1327
1328 Type = K('type') - Name - D('{') - R(Policy) - D('}')
1329 def build_type(s, l, t):
1330 try:
1331 cat = CATEGORYMAP[t[0]]
1332 except KeyError:
1333 raise P.ParseException(s, loc, "Unknown category `%s'" % t[0])
1334 pols = t[1]
1335 if len(pols) == 1: pol = pols[0]
1336 else: pol = AndPolicy(pols)
1337 pol.setcategory(cat)
1338 return pol
1339 Type.setParseAction(build_type)
1340
1341 TARGETS = []
1342 class TargetJob (object):
1343 def __init__(me, targetdir, policies):
1344 me.targetdir = targetdir
1345 me.policies = policies
1346 def perform(me):
1347 TARGETS.append(me)
1348
1349 Target = K('target') - String - D('{') - R(Type) - D('}')
1350 def build_target(s, l, t):
1351 return TargetJob(t[0], t[1])
1352 Target.setParseAction(build_target)
1353
1354 VARS = { 'master': None }
1355 class VarsJob (object):
1356 def __init__(me, vars):
1357 me.vars = vars
1358 def perform(me):
1359 for k, v in me.vars:
1360 VARS[k] = v
1361
1362 Var = prop('master', String)
1363 Vars = K('vars') - D('{') - R(Var) - D('}')
1364 def build_vars(s, l, t):
1365 return VarsJob(t[0])
1366 Vars.setParseAction(build_vars)
1367
1368 TopLevel = Vars | Target
1369 Config = R(TopLevel)
1370 Config.ignore(P.pythonStyleComment)
1371
1372 ###--------------------------------------------------------------------------
1373 ### The directory grobbler.
1374
1375 def grobble(master, targets, noact = False):
1376 """
1377 Work through the MASTER directory, writing converted files to TARGETS.
1378
1379 The TARGETS are a list of `TargetJob' objects, each describing a target
1380 directory and a policy to apply to it.
1381
1382 If NOACT is true, then don't actually do anything permanent to the
1383 filesystem.
1384 """
1385
1386 ## Transform the targets into a more convenient data structure.
1387 tpolmap = []
1388 for t in targets:
1389 pmap = {}
1390 tpolmap.append(pmap)
1391 for p in t.policies: pmap.setdefault(p.cat, []).append(p)
1392
1393 ## Keep track of the current position in the master tree.
1394 dirs = []
1395
1396 ## And the files which haven't worked.
1397 broken = []
1398
1399 def grobble_file(master, pmap, targetdir, cohorts):
1400 ## Convert MASTER, writing the result to TARGETDIR.
1401 ##
1402 ## The COHORTS are actually (CAT, ID, COHORT) triples, where a COHORT is
1403 ## a list of (FILENAME, ID) pairs.
1404 ##
1405 ## Since this function might convert the MASTER file, the caller doesn't
1406 ## know the name of the output files, so we return then as a list.
1407
1408 done = set()
1409 st_m = OS.stat(master)
1410
1411 ## Work through each category listed and apply its policy.
1412 for cat, id, cohort in cohorts:
1413
1414 ## Go through the category's policies and see if any match. If we fail
1415 ## here, see if there are more categories to try.
1416 for pol in pmap[cat]:
1417 acts = pol.actions(master, targetdir, id, cohort)
1418 if acts: break
1419 else:
1420 continue
1421
1422 ## Work through the targets one by one.
1423 for a in acts:
1424 done.add(a.target)
1425
1426 ## Find out whether the target file already exists and is up-to-date
1427 ## with respect to the master. (Caution here with low-resolution
1428 ## timestamps.) If it's OK, then just move on.
1429 try:
1430 st_t = OS.stat(a.target)
1431 if st_m.st_mtime < st_t.st_mtime or \
1432 (st_m.st_ino, st_m.st_dev) == (st_t.st_ino, st_t.st_dev):
1433 continue
1434 except OSError, err:
1435 if err.errno not in (E.ENOENT, E.ENOTDIR):
1436 raise
1437
1438 ## We have real work to do. If there's a current status message,
1439 ## it's the containing directory so flush it so that people know
1440 ## where we are.
1441 STATUS.commit()
1442
1443 ## Remove the target. (A hardlink will fail if the target already
1444 ## exists.)
1445 if not noact:
1446 try:
1447 OS.unlink(a.target)
1448 except OSError, err:
1449 if err.errno not in (E.ENOENT, E.ENOTDIR):
1450 raise
1451
1452 ## Do whatever it is we decided to do.
1453 if noact:
1454 STATUS.commit(filestatus(master, a))
1455 else:
1456 a.perform()
1457
1458 ## We're done. Return the names of the targets.
1459 return list(done)
1460
1461 @contextmanager
1462 def wrap(masterfile):
1463 ## Handle exceptions found while trying to convert a particular file or
1464 ## directory.
1465
1466 try:
1467 yield masterfile
1468
1469 ## Something bad happened. Report the error, but continue. (This list
1470 ## of exceptions needs a lot of work.)
1471 except (IOError, OSError), exc:
1472 STATUS.clear()
1473 STATUS.commit(filestatus(masterfile, 'failed (%s)' % exc))
1474 broken.append((masterfile, exc))
1475
1476 def grobble_dir(master, targets):
1477 ## Recursively convert files in MASTER, writing them to the TARGETS.
1478
1479 ## Keep track of the subdirectories we encounter, because we'll need to
1480 ## do all of those in one go at the end.
1481 subdirs = set()
1482
1483 ## Work through each target directory in turn.
1484 for target, pmap in zip(targets, tpolmap):
1485
1486 ## Make sure the TARGET exists and is a directory. It's a fundamental
1487 ## assumption of this program that the entire TARGET tree is
1488 ## disposable, so if something exists but isn't a directory, we should
1489 ## kill it.
1490 if OS.path.isdir(target):
1491 pass
1492 else:
1493 if OS.path.exists(target):
1494 STATUS.commit(filestatus(target, 'clear nondirectory'))
1495 if not noact:
1496 OS.unlink(target)
1497 STATUS.commit(filestatus(target, 'create directory'))
1498 if not noact:
1499 OS.mkdir(target)
1500
1501 ## Keep a list of things in the target. As we convert files, we'll
1502 ## check them off. Anything left over is rubbish and needs to be
1503 ## deleted.
1504 checklist = {}
1505 try:
1506 for i in OS.listdir(target):
1507 checklist[i] = False
1508 except OSError, err:
1509 if err.errno not in (E.ENOENT, E.ENOTDIR):
1510 raise
1511
1512 ## Keep track of the files in each category.
1513 catmap = {}
1514 todo = []
1515 done = []
1516
1517 ## Work through the master files.
1518 for f in sorted(OS.listdir(master)):
1519
1520 ## If the killswitch has been pulled then stop. The whole idea is
1521 ## that we want to cause a clean shutdown if possible, so we don't
1522 ## want to do it in the middle of encoding because the encoding
1523 ## effort will have been wasted. This is the only place we need to
1524 ## check. If we've exited the loop, then clearing old files will
1525 ## probably be fast, and we'll either end up here when the recursive
1526 ## call returns or we'll be in the same boat as before, clearing old
1527 ## files, only up a level. If worst comes to worst, we'll be killed
1528 ## forcibly somewhere inside `SH.rmtree', and that can continue where
1529 ## it left off.
1530 if KILLSWITCH.is_set():
1531 return
1532
1533 ## Do something with the file.
1534 with wrap(OS.path.join(master, f)) as masterfile:
1535
1536 ## If it's a directory then prepare to grobble it recursively, but
1537 ## don't do that yet.
1538 if OS.path.isdir(masterfile):
1539 subdirs.add(f)
1540 done.append(OS.path.join(target, f))
1541
1542 ## Otherwise it's a file. Work out what kind, and stash it under
1543 ## the appropriate categories. Later, we'll apply policy to the
1544 ## files, by category, and work out what to do with them all.
1545 else:
1546 mime = GIO.file_new_for_path(masterfile) \
1547 .query_info('standard::content-type', 0) \
1548 .get_content_type()
1549 cats = []
1550 for cat in pmap.iterkeys():
1551 id = cat.identify(masterfile, mime)
1552 if id is None: continue
1553 catmap.setdefault(cat, []).append((masterfile, id))
1554 cats.append((cat, id))
1555 if not cats:
1556 catmap.setdefault(None, []).append((masterfile, id))
1557 todo.append((masterfile, cats))
1558
1559 ## Work through the categorized files to see what actions to do for
1560 ## them.
1561 for masterfile, cats in todo:
1562 with wrap(masterfile):
1563 done += grobble_file(masterfile, pmap, target,
1564 [(cat, id, catmap[cat]) for cat, id in cats])
1565
1566 ## Check the results off the list so that we don't clear it later.
1567 for f in done:
1568 checklist[OS.path.basename(f)] = True
1569
1570 ## Maybe there's stuff in the target which isn't accounted for. Delete
1571 ## it: either the master has changed, or the policy for this target has
1572 ## changed. Either way, the old files aren't wanted.
1573 for f in checklist:
1574 if not checklist[f]:
1575 STATUS.commit(filestatus(f, 'clear bogus file'))
1576 if not noact:
1577 bogus = OS.path.join(target, f)
1578 try:
1579 if OS.path.isdir(bogus):
1580 SH.rmtree(bogus)
1581 else:
1582 OS.unlink(bogus)
1583 except OSError, err:
1584 if err.errno != E.ENOENT:
1585 raise
1586
1587 ## If there are subdirectories which want processing then do those.
1588 ## Keep the user amused by telling him where we are in the tree.
1589 for d in sorted(subdirs):
1590 dirs.append(d)
1591 STATUS.set('/'.join(dirs))
1592 with wrap(OS.path.join(master, d)) as masterdir:
1593 try:
1594 grobble_dir(masterdir,
1595 [OS.path.join(target, d) for target in targets])
1596 finally:
1597 dirs.pop()
1598 STATUS.set('/'.join(dirs))
1599
1600 ## Right. We're ready to go.
1601 grobble_dir(master, [t.targetdir for t in targets])
1602 return broken
1603
1604 ###--------------------------------------------------------------------------
1605 ### Command-line interface.
1606
1607 QUIS = OS.path.basename(SYS.argv[0])
1608
1609 def moan(msg):
1610 "Report a warning message to the user."
1611 SYS.stderr.write('%s: %s\n' % (QUIS, msg))
1612
1613 def die(msg):
1614 "Report a fatal error message to the user."
1615 moan(msg)
1616 SYS.exit(1)
1617
1618 def parse_opts(args):
1619 """
1620 Parse command-line arguments in ARGS.
1621
1622 Returns a Grobbler object and the MASTER and TARGET directories to be
1623 grobbled.
1624 """
1625
1626 ## Build the option parser object.
1627 op = OP.OptionParser(prog = QUIS, version = VERSION,
1628 usage = '%prog [-in] [-t TIMEOUT] [-T TIMEOUT] '
1629 'CONFIG',
1630 description = """\
1631 Convert a directory tree of files according to the configuration file
1632 CONFIG.
1633 """)
1634
1635 ## Timeout handling.
1636 def cb_time(opt, ostr, arg, op):
1637 m = RX.match(r'\s*(\d+)\s*([dhms]?)\s*', arg)
1638 if not m:
1639 raise OP.OptionValueerror, 'bad time value `%s\'' % arg
1640 t, u = m.groups()
1641 t = int(t) * { '': 1, 's': 1, 'm': 60, 'h': 3600, 'd': 86400 }[u]
1642 setattr(op.values, opt.dest, t)
1643 op.add_option('-t', '--timeout', type = 'string', metavar = 'SECS',
1644 dest = 'timeout',
1645 help = 'stop processing nicely after SECS',
1646 action = 'callback', callback = cb_time)
1647 op.add_option('-T', '--timeout-nasty', type = 'string', metavar = 'SECS',
1648 dest = 'timeout_nasty',
1649 help = 'stop processing unpleasantly after further SECS',
1650 action = 'callback', callback = cb_time)
1651
1652 ## Other options.
1653 op.add_option('-i', '--interactive', action = 'store_true', dest = 'tty',
1654 help = 'provide progress information')
1655 op.add_option('-n', '--no-act', action = 'store_true', dest = 'noact',
1656 help = 'don\'t actually modify the filesystem')
1657
1658 ## Ready to rock.
1659 op.set_defaults(formats = [], noact = False,
1660 timeout = None, timeout_nasty = 300)
1661 opts, args = op.parse_args(args)
1662
1663 ## Check that we got the non-option arguments that we want.
1664 if len(args) != 1:
1665 op.error('wrong number of arguments')
1666
1667 ## Act on the options.
1668 if opts.tty:
1669 STATUS.eyecandyp = True
1670 if opts.timeout is not None:
1671 to = TH.Thread(target = timeout,
1672 args = (opts.timeout, opts.timeout_nasty))
1673 to.daemon = True
1674 to.start()
1675
1676 ## Parse the configuration file.
1677 with open(args[0]) as conf:
1678 jobs, = Config.parseFile(conf, True)
1679 for j in jobs:
1680 j.perform()
1681
1682 return opts
1683
1684 if __name__ == '__main__':
1685 opts = parse_opts(SYS.argv[1:])
1686 if 'master' not in VARS:
1687 die("no master directory set")
1688 broken = grobble(VARS['master'], TARGETS, opts.noact)
1689 if broken:
1690 moan('failed to convert some files:')
1691 for file, exc in broken:
1692 moan('%s: %s' % (file, exc))
1693 SYS.exit(1)
1694
1695 ## This is basically a successful completion: we did what we were asked to
1696 ## do. It seems polite to report a message, though.
1697 ##
1698 ## Why don't we have a nonzero exit status? The idea would be that a
1699 ## calling script would be interested that we used up all of our time, and
1700 ## not attempt to convert some other directory as well. But that doesn't
1701 ## quite work. Such a script would need to account correctly for time we
1702 ## had spent even if we complete successfully. And if the script is having
1703 ## to watch the clock itself, it can do that without our help here.
1704 if KILLSWITCH.is_set():
1705 moan('killed by timeout')
1706
1707 ###----- That's all, folks --------------------------------------------------