mpp or "mpodder python", is a re-implementation of the perl-based mpodder program using Python 3 and a more modern approch to configuration and data storage. The intention is to create broadly the same end-user experience in terms of program usage, but with a more robust program which is easier to maintain and performs better than the original mpodder.

Non-exhaustive list of goals:

1. Use Python's multi-processing module to parallelize downloads of RSS feeds.
2. Use JSON format to store configuration.
3. Faster handling of large feeds.

• Python 3 (come on people, let's not be staring new projects with Python 2!)
• Python modules (pip3 package names):
  • veryprettytable
  • python-dateutil

Optional development dependencies:

• python-nose (for running tests)

Podcast objects can be created from a URL or loaded from a file. They contain limited information about a podcast:

• the feed URL
• podcast title
• a list of episodes

Each Episode object stores:

• the episode download url
• episode title
• published date
• media path (where the downloaded media file is stored locally)
• listened flag

When a Podcast object is created from a URL, the URL is downloaded, and the title set. Episode objects are created with listened=False.

SKIPPED     MEDIA_PATH is None  MEDIA_PATH_EXISTS   STATUS

False       True                -                   new
False       False               True                downloaded
False       False               False               listened
True        True                -                   skipped
True        False               True                skipped (needs cleaning)
True        False               False               skipped

Expected status changes over time:

new -> skipped new -> downloaded -> skipped new -> downloaded -> listened

list [--url] [--path] [filter]

Print a list of podcasts which are known to mpp. If a filter is specified the resulting list of filtered to include only podcasts whose title matches the filter. if --url is specified, only the title of the podcast and the url of it's feed are listed. If --path is specified, only the title of the podcast and the path to the json feed file are listed.

ls

add <url>

Adds a podcast from a feed URL. All episodes are set to ready to download.

show [filter]

Dump all available information about one of more podcasts. This includes the URL.

info

remove <filter>

Remove podcasts whose title matches filter. This will also remove any downloaded episodes of the podcast, and all information regarding which episodes have been listened to.

rm

catchup [--leave n] <filter>

Sets all episodes to state "listened" for podcasts matching filter. If the --leave option is specified, n episodes will be left with ready status.

rename <filter> <title> [title ...]

Rename a podcast matching filter. If more than one podcast matches filter, an error is raised and mpp quits. If multiple title arguments are specified, they are joined with the space character.

mv

update [--parallel=p] filter

Updates the feed data for all podcasts that match filter. If --parallel is specified, this is done with up to p parallel downloading processes.

up

download [--parallel=p] [--max=m] filter

Download up to m new episodes for each podcast which matches filter, using at most p parallel download processes.

dl

fetch [--parallel=p] [--max=m] filter

Equivalent to running update and then download.

get, f

episodes [--first=f] [--last=l] [--path] [--status=s] filter 

Lists the episodes for podcasts matching filter. By default only new and ready episodes are displayed, and the output consistes of: podcast title, episode title, episode status.

The --first option may be used to show only the first f results (sorted in feed order). The --last option shows only the las l results. If --status is specified, only episodes having status s are displayed (possible statuses are: new, ready, skipped, cleanedm and all, which selects episodes with any status. If the --path option is used, only new podcasts with the media downloaded are displayed, and the output is just the path to the downloaded files.

ep, lse

renew [--since=YYYYMMDD] [--last=l] [--first=f] [--status=s] filter 

For podcasts that match filter, change the status from 'skipped' or 'cleaned' to 'new'.

If --status is specified, only affect episodes which have status s. If --last is specified, only affect the last l episodes. If --first is specified, only affect the first f episodes. If --sine is specified, only affect episodes published since the date defined by YYYYMMDD (inclusive).

The original URL supplied when adding the podcast may not be the same as the one the feed replies with, causing the hash to change between the original addition and subsequent saves.

Solution: instead of basing the filename on a hash of the URL, each podcast should be allocated a unique. Challenge: detecting duplicates on addition

Limit of multiprocessing... because each download process gets it's own copy of the Podcast object, writing will only ever save one episode... the last one to be written.

Temporary work around - only download one episode from a given podcast per invokation.

Proper solution:

1. Have save_to_file include locking mechanism (file locks), and re-load Podcast object before saving (ugly)
2. Have processes send back updated episodes to calling thread and save at end of all downloaders (risks unsaved changes)
3. Threaded (rather than multiprocessing based) DownloadManager, which spawns downloader threads, sharing Podcast objects and using a thread lock to mediate updates and then perform saves.