[Pyparsing] How do fix this?

[Robin Siebler]

I have some text I am trying to parse, specifically the track data. I have
grammar defined and it works, however I am having the following problem:

If the first track found is an audio track, I can retrieve both the audio
and video tracks. However, if the first track is a video track, when I try
to retrieve the tracks, I get the video track twice and no audio track.


Here is my sample data:

===============================================================================
File:
file:///Users/test/Desktop/Media/Playback_Video_Format_Files/ipad_2_frontface_portrait.mov
5,674,352 bytes
File Type Info: MajorBrand 'qt ' MinorVersion 0x00000000 1 compatible
brands: 'qt '
Movie Data ('mdat'): 1 atom found, 5,665,002 data bytes
com.apple.coremedia.formatreader.quicktime-iso
Creation Date (date): 2011-06-14T16:06:30-0700
Movie Timescale: 600 Duration: 7444/600 00:00:12.407
Preferred rate: 1.00 Preferred Volume: 1.00 Movie Matrix: 1.0 0.0 0.0 / 0.0
1.0 0.0 / 0.0 0.0 1.0
Movie is self-contained. Not fast-start.
QT user data atom available. QT metadata atom available.

2 tracks present.

Track ID 1 vide (Video) Enabled Self-contained
Format vide/avc1 dimensions: video 640 x 480, presentation: 640 x 480
(pixelAspect+clean), cleanAperture: 640 x 480 @ 0,0 (originTopLeft)
Media Timescale: 600 Duration: 7447/600 00:00:12.412 MinSampleDuration:
19/600 AdvanceDecodeDelta: 0/600 00:00:00.000
Num data bytes: 5555600 Est. data rate: 3.581 Mbps Nominal framerate:
29.972 fps 372 samples
Included in auto selection. Language code <und>
Dimensions: 640 x 480 CleanAperture: 640 x 480 ProductionAperture: 640 x
480 EncodedPixels: 640 x 480 Track Matrix: 0.0 1.0 0.0 / -1.0 0.0 0.0 /
480.0 0.0 1.0
1 edit: Media start 0/600 00:00:00.000 dur 7444/600
00:00:12.407 Track start 0/600 00:00:00.000 dur 7444/600
00:00:12.407
QT metadata atom available.

Track ID 2 soun (Audio) Enabled Self-contained
Format soun/aac 44100 Hz aac FormatFlags: 0x00000000 Bytes/Pkt: 0
Frames/Pkt: 1024 Bytes/Frame: 0 Chan/Frame: 1 Bits/Chan: 0 Reserved:
0x00000000
ChannelLayout: Mono
Media Timescale: 44100 Duration: 549888/44100 00:00:12.469
MinSampleDuration: 1024/44100 AdvanceDecodeDelta: 0/44100 00:00:00.000
Num data bytes: 99464 Est. data rate: 63.815 kbps Nominal framerate: 43.066
fps 537 samples
Track volume: 1
Included in auto selection. Language code <und>
Dimensions: 0 x 0 Track Matrix: 1.0 0.0 0.0 / 0.0 1.0 0.0 / 0.0 0.0 1.0
1 edit: Media start 0/44100 00:00:00.000 dur 7444/600
00:00:12.407 Track start 0/600 00:00:00.000 dur 7444/600
00:00:12.407
"""

Here is a snippet of my grammar:

# Start Track Info Block
# Track Info
self.crap = Suppress(Optional('[file]'))
self.track_id = 'Track ID' + self.plain_number + self.word + '(' +
self.word.setResultsName('type') + restOfLine

# Track Format
self.audio_track_format = 'Format' + Combine(self.word + '/' +
self.word).setResultsName('track_format') + \
Combine(self.plain_number + self.word,
adjacent=False).setResultsName('frequency') + \
self.word.setResultsName('codec') + restOfLine
self.subtitle_track_format = 'Format' + Combine(self.word + '/' +
self.word).setResultsName(
'track_format') + restOfLine
self.vtrack_format = 'Format' + Combine(self.word + '/' +
self.word).setResultsName('track_format')
self.vtrack_dimensions = 'dimensions: video' + Combine(self.plain_number +
'x' + self.plain_number,

adjacent=False).setResultsName('video')
self.vtrack_presentation = Suppress(', ') + 'presentation:' + \
Combine(self.plain_number + 'x' +
self.plain_number + '(' + self.word + '+' +
self.word + ')' + Suppress(','),
adjacent=False) \
.setResultsName('presentation')
self.vtrack_cleanaperture = 'cleanAperture:' + Combine(self.plain_number +
'x' + self.plain_number +
restOfLine,
adjacent=False) \
.setResultsName('cleanAperture')
self.video_track_format = self.vtrack_format + self.vtrack_dimensions + \
self.vtrack_presentation +
self.vtrack_cleanaperture

# Audio Channel Layout
self.channel_layout = 'ChannelLayout:' +
restOfLine.setResultsName('channel_layout')

# Audio Track Volume
self.audio_track_volume = 'Track volume:' +
self.plain_number.setParseAction(lambda t: int(t[0])) \
.setResultsName('audio_volume')

# Frame reordering
self.frame = Optional(Combine('Frame') + restOfLine)

# Included
self.included = Combine('Included' + restOfLine)

# Media Timescale
self.timescale = Combine(self.word + 'Timescale:' + self.plain_number,
adjacent=False)
self.duration = Combine('Duration:' + self.div_num + self.time,
adjacent=False)
self.sample_duration = Combine('MinSampleDuration:' +
self.div_num.setResultsName('sample_duration'),
adjacent=False)
self.decode_delta = Combine('AdvanceDecodeDelta:' + Combine(self.div_num +
restOfLine)
.setResultsName('decode_delta'), adjacent=False)
self.media_timescale = self.timescale + self.duration +
self.sample_duration + self.decode_delta

# Track Data
self.data_bytes = 'Num data bytes:' +
self.plain_number.setResultsName('data_bytes')
self.data_rate = 'Est. data rate:' + Combine(self.decimal_number +
self.word, adjacent=False) \
.setResultsName('estimated_data_rate')
self.frame_rate = 'Nominal framerate:' + Combine(self.decimal_number +
self.word, adjacent=False) \
.setResultsName('fps')
self.samples = self.plain_number.setResultsName('samples')
self.track_data = self.data_bytes + self.data_rate + self.frame_rate +
self.samples + restOfLine

# Track Dimensions
self.audio_dimensions = 'Dimensions:' + Combine(self.plain_number + 'x' +
self.plain_number, adjacent=False) \
.setResultsName('Dimensions')
self.audio_track_matrix = 'Track Matrix:' +
restOfLine.setResultsName('track_matrix')
self.audio_track_dimensions = self.audio_dimensions +
self.audio_track_matrix
self.subtitle_dimensions = 'Dimensions:' + Combine(self.plain_number + 'x'
+ self.plain_number, adjacent=False) \
.setResultsName('Dimensions')
self.subtitle_track_matrix = 'Track Matrix:' +
restOfLine.setResultsName('track_matrix')
self.subtitle_track_dimensions = self.subtitle_dimensions +
self.subtitle_track_matrix
self.video_dimensions = 'Dimensions:' + Combine(self.plain_number + 'x' +
self.plain_number, adjacent=False) \
.setResultsName('Dimensions')
self.video_clean_aperture = 'CleanAperture: ' + Combine(self.plain_number +
'x' + self.plain_number,

adjacent=False).setResultsName('CleanAperture')
self.video_production_aperture = 'ProductionAperture:' +
Combine(self.plain_number + 'x' + self.plain_number,

adjacent=False).setResultsName(
'ProductionAperture')
self.video_encoded_pixels = 'EncodedPixels:' + Combine(self.plain_number +
'x' + self.plain_number,

adjacent=False).setResultsName('EncodedPixels')
self.video_track_matrix = 'Track Matrix:' +
restOfLine.setResultsName('TrackMatrix')
self.video_track_dimensions = self.video_dimensions +
self.video_clean_aperture + \
self.video_production_aperture +
self.video_encoded_pixels + self.video_track_matrix

# Edits
self.num_edits =
self.plain_number.setResultsName('number_of_edits').setParseAction(lambda
t: int(t[0])) + \
Optional(Literal('edit:') | Literal('edits:'))
self.media_start = 'Media start' + Optional(Group(self.div_num + self.time)
| 'INVALID TIME') \
.setResultsName('MediaStart')
self.media_duration = 'dur' + Group(self.div_num +
self.time).setResultsName('MediaDuration')
self.track_start = 'Track start' + Group(self.div_num +
self.time).setResultsName('TrackStart')
self.track_duration = 'dur' + Group(self.div_num +
self.time).setResultsName('TrackDuration')
self.media_edit = Group(self.media_start + self.media_duration +
self.track_start + self.track_duration +
restOfLine + LineEnd().suppress())
self.edits = self.num_edits +
OneOrMore(self.media_edit).setResultsName('edits')

# Define Track Info Block
# Audio Block
self.audio_track_info = Group(self.crap + self.track_id +
self.audio_track_format + self.channel_layout +
self.media_timescale + self.track_data +
self.audio_track_volume +
self.included + self.audio_track_dimensions +
OneOrMore(self.edits) + Optional(self.qt_user_data))

# Subtitle Block
self.subtitle_track_info = Group(self.crap + self.track_id +
self.subtitle_track_format + \
self.media_timescale + self.track_data +
self.included + \
self.subtitle_track_dimensions +
OneOrMore(self.edits) + Optional(self.qt_user_data))

# Video Block
self.video_track_info = Group(self.crap + self.track_id +
self.video_track_format + self.media_timescale +
self.track_data + self.frame + self.included +
self.video_track_dimensions +
OneOrMore(self.edits) +
Optional(self.qt_user_data))

self.tracks = ZeroOrMore(self.audio_track_info | self.video_track_info |
self.subtitle_track_info) \
.setResultsName('tracks')
------------------------------------------------------------------------------

[Paul McGuire]

First of all, it is not necessary (and probably not even helpful) to define
every element of your grammar as an instance variable, preceded by 'self.'.
In nearly all cases, when I define a grammar within a parsing class, I'll do
something like this:

class BobParser(object):
def __init__(self):
expr1 = Literal("Bob's your")
expr2 = oneOf("uncle aunt brother sister father mother")
self.parser = expr1 + expr2('relation')

def who_is_it(self, string):
return self.parser.parseString(string).relation


All the sub-expressions can just be local variables, and save the "self."
business for just the top-level parser.

This really looks like you just started at the beginning of your input text
and started writing pyparsing expressions to it. Many times there's nothing
wrong with that, but in your case, there is a lot of complexity and
structure to your input. And more importantly, many internal patterns that
are repeated - these can be defined once and then reused by name over and
over.

I often encourage people to write a BNF for complex data like this. Or at
*least* look at the input at an overall level for which bits are common and
can be reused. There are many small pieces that can be defined and reused in
larger parts, and these will help simplify your grammar. For instance,
there are many places where you use this:

Combine(self.plain_number + 'x' + self.plain_number,
adjacent=False).setResultsName('something')

which really is challenging to the eye to see what is the parser and what is
the meta-information (adjacency, results name). And you repeat this whenever
you need a "0 x 0" or "640 x 480", so it makes things very messy. If you
define for yourself this reusable expression:

number_x_number = Combine(plain_number + 'x' + plain_number, adjacent =
False)

Then you can use it in multiple places as:

number_x_number("clean_aperature")
number_x_number("Dimensions")
etc.

I also strongly recommend using the short-cut version of
expr.setResultsName('xyz'), as expr('xyz'). This notation can really clean
up your grammar definition, and make it easier to see the overall parser
without being distracted by all the function calls.

Finally, you make heavy use of "Combine(something + something + something,
adjacent=False)". Please consider using Group instead. It's clearer to
follow, it implicitly allows whitespace (so adjacent=False is not
necessary), and it allows you to define results names within the group,
making for a useful substructure (as in sample_time below):

number_x_number = Group(plain_number + 'x' + plain_number)
number_sl_number = Group(plain_number + '/' + plain_number)
word_sl_word = Gruop(word + '/' + word)
timestamp = Regex(r'\d\d:\d\d:\d\d\.\d\d\d')
sample_time = Group(number_sl_number('sample') + timestamp('time'))

So just going through and stripping out all the "self." stuff and using some
of these repeated sub-expressions might make things easier to follow, and
your overall intent and structure will be clearer. Here is one area of your
parser with these changes:

audio_dimensions = 'Dimensions:' + number_x_number('Dimensions')
audio_track_matrix = 'Track Matrix:' + restOfLine('track_matrix')
audio_track_dimensions = audio_dimensions + audio_track_matrix
subtitle_dimensions = 'Dimensions:' + number_x_number('Dimensions')
subtitle_track_matrix = 'Track Matrix:' + restOfLine('track_matrix')
subtitle_track_dimensions = subtitle_dimensions + subtitle_track_matrix
video_dimensions = 'Dimensions:' + number_x_number('Dimensions')
video_clean_aperture = 'CleanAperture: ' +
number_x_number('CleanAperture')
video_production_aperture = 'ProductionAperture:' +
number_x_number('ProductionAperture')
video_encoded_pixels = 'EncodedPixels:' +
number_x_number('EncodedPixels')

Now you can answer some other questions for yourself, like "why do I repeat
the 'Dimensions:' expression?", and further simplify your grammar.

Good luck,
-- Paul


---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus


------------------------------------------------------------------------------