mpd-parser 0.2.0

lxml based parser for DASH manifests (mpd files)

mpd-parser

Installation

$ python -m pip install mpd-parser

Note: mpd-parser requires Python 3.10 or newer.

Usage

Importing

from mpd_parser.parser import Parser

parse from string

with open("path/to/file.mpd", mode="r") as manifest_file:
    mpd_string = manifest_file.read()
    parsed_mpd = Parser.from_string(mpd_string)

parse from file

input_file = "path/to/file.mpd"
mpd = Parser.from_file(input_file)

convert back to string

mpd_as_xml_string = Parser.to_string(parsed_mpd)

Overview

A utility to parse mpeg dash mpd files quickly This package is heavily inspired by mpegdash package the main difference is that I choose to relay on lxml for parsing, and not the standard xml library.

The decision to implement it with lxml is for two reasons:

1. lxml is faster then minidom
2. lxml mimics the ElementTree API which is a more pythonic approach to XMLs

mpegdash package has two distinct advantages over this package:

1. it does not require third party libraries.
2. it uses the classic DOM approach to parsing XML files. it is a well known standard.

Benchmarks

You can run the benchmarks using the scripts in the benchmarks/ directory.

$ python benchmarks/bench_mpd_parser.py --file manifests/test_manifest_1mb.mpd
$ python benchmarks/bench_mpegdash.py --file manifests/test_manifest_1mb.mpd

Results

On a 1MB MPD file with 6,200 representations:

• mpd-parser: ~7 ms / parse
• mpegdash: ~174 ms / parse

(mpd-parser is ~24x faster than mpegdash on this payload).

Example manifests

Taken from https://ottverse.com/free-mpeg-dash-mpd-manifest-example-test-urls/ These are what I used to test and benchmark the package.

Missing unit-tests

1. tags
2. attribute parsers
3. full manifest testing

Contributing

TBA

Build locally

python -m build

Run pylint locally

I try to keep the pylint score above 9.

python -m pylint ./mpd_parser/

TODO

1. finish working on periods and sub tags
   1. periods
   2. adapt-sets
   3. segment bases
   4. segment lists
   5. segment templates
   6. asset ids
   7. event streams
   8. subsets
2. create package locally
3. test it
4. complete readme
   1. installation
   2. usage
   3. Benchmarks
   4. contributing
5. push to github
6. push package to pypi
7. add github actions
   1. pylint
   2. pytest
   3. build package
   4. push package
8. complete unit-tests
9. refactor tags to multiple files
10. Parsing:
    1. parsing from string
    2. parsing from file
    3. parsing from URL
11. save mpd object:
    1. object to string
    2. object to file