Switching to YAML for config

[brokkr]

I believe either JSON or YAML would allow for significant code simpification (there's a lot of lines just parsing the XML). And I don't believe we really depend on the things that XML brings to the table (namespaces etc.)?

Advantages of JSON: Built-in (one less dependency), fast
Advantages of YAML: Easy to read, supplanting pickle?

Really depends on how well either one deals with the heavily nested structure of say, a subscription with filters and metadata and renames... Also how would attributes translate?

[brokkr]

Leaving aside the question of attributes, YAML seems like an easy choice.

• It's easy to read and write
• Recreating the poca.xml structure is easy
• Parsing it is two lines of code and I have the entire dictionary in hand

Example:

defaults:
  metadata:
    genre: podcast
settings:
  base_dir: /mnt/media/
  filenames: permissive
  id3removev1: true
  id3v2version: 3
  useragent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101 Firefox/78.0
subscriptions:
- filters:
    title: natur
  max_number: 2
  metadata:
    artist: Det:Er
  rename:
  - title
  - episode_title
  title: "Er h\xE5bet stadig gr\xF8nt?"
  url: https://www.dr.dk/mu/feed/er-habet-stadig-groent.xml?format=podcast
- filters:
    filename: CCEpSpecial
  max_number: 2
  title: The Crate and Crowbar
  url: http://crateandcrowbar.com/category/episode/feed/

A few issues:

• Dictionaries are randomly rearranged. Apparently this can be avoided using sortkeys=False when dumping? (https://stackoverflow.com/a/64902048) but it doesn't seem to work ootb with safe_dump.
• yaml dumping is by default encoded latin-1. Apparently it can use UTF-8 instead by using the allow_unicode=True setting (https://stackoverflow.com/a/29600111) Also works for safe_dump (tested)
• Biggest one: How to validate input? YAML doesn't come with schemas so some options are:
  • Build your own: https://stackoverflow.com/a/45812573
  • Use third party: schema or Cerberus or jsonschema
  • drop validation
• read up on how safe is safe_load? and if we safe_load does it matter if we don't safe_dump?

[brokkr]

Also: pathlib.Path (Python 3.4) for all Paths class operations

[brokkr]

Note: using os.access on pathlike object (e.g. pahlib's Path class) requires 3.6

[brokkr]

Still to do:

• input validation
• safe_loading/safe_dumping
• updating 301
• save state.yaml

[brokkr]

We can now save and read state.yaml including converting time_structs and pathlib.Paths back and forth. Still to do in addition to those mentioned above:

• consolidate handling of yaml in one submodule
• save 410 codes to state