bcongdon / bolero Goto Github PK

Currently, permissions is handled at the tracker implementation level (check_auth is set as a request preprocessor in Tracker.create_api()).

This should be changed so that tracker permissions can be set at the tracker configuration level (i.e. config.json)

Per #1, trackers will be objects instead of modules.

Thus, tracker loading will need to be updated from the naive importlib.import_module approach.

It will probably be sufficient to do something like youtube-dl's mass-import, and then selectively authenticate / activate the enabled trackers.

Trackers should be objects, instead of formless modules. They behave more like objects already, and it's more idiomatic to call duck-typed functions of a subclass to Tracker than to a module.

Trackers should be able to do the following:

• Authenticate the API client
  • handleAuthentication()
• Do some type of regular scraping function
  • scrape()
• Do a complete 'refresh' of their object's database
  • backfill()

Trackers should be distinct from their associated models, and should only have simple CRUD tasks on the actual data.

Plugins should be more like plugins than internal modules.
It's fine to have a list of modules as strings in a configuration file and import them via __import__.

The idea is to have bolero be a relatively lean framework, and push a lot of the complexity / implementation out to the tracker plugins.

Next big push is to get the actual tests working.
#3 reduced the dependencies between the app and trackers, so it should be possible to unit test the trackers without having to load / startup the app. (If it isn't, then that's something to do too)

Not sure how much testing is necessary for an external api driven project like this, but something better than running the updateall cli command would be good. 👍

Should have a simple "howto" guide on how to implement a new Tracker

Implement a Todoist tracker with similar functionality to the Wunderlist tracker.

Need to have some type of explanation as to what problem (API data warehousing) this project is trying to solve in the README.

MyFitnessPalTracker.backfill should find the user's earliest logged day and start backfilling from there.

This is a bit difficult due to the way that the python module used to connect with MFP is structured.
Right now, we're just naively grabbing the last 1 year of history.