It's a bot that assumes control of a Trello account with a configurable API key and token, and it automatically adds a random cover to a card when assigned or mentioned.
It's also written in Haskell, and it's also the first non-trivial thing I've written in Haskell. It's bad.
- Iterates all boards you're assigned to
- Keeps track of the last action scanned per board, and only processes new actions
- Persists these action tail markers to disk
- When added to card, adds a random cover and removes self
- When mentioned on card, takes the message (or card name if message is blank) and adds a random cover
That's like, feature completion right there.
- Development. Commenting out chunks of code so that it runs once and doesn't persist the tail marker is a little frustrating when trying to test.
- Error handling parsing json responses. I abuse
fromJustall over the place. - Error handling with non-200 status codes from Trello. Especially when it comes to advancing the action tail marker.
- Parsing the jpg.to response. Yeah, it's html, but man. I'm dropping the first 75 characters right now and taking everything up to the first double quote. That's first-class clown town.
- Modularity. I can split a lot of logic into separate files, even.
- Doing things right. The action triggers smell pretty bad.
- Better model representation.
Action ObjectId ActionType ActionDatais not a good way to represent an action. One would say it's not even close. I think the answers is lenses somewhere. - A full Haskell Trello client. Or a part of a Haskell Trello client. Something like that.
- Pull the majority of the core logic into a package. Let the world write their own action-tailing bots.
- Tests. I think those are a thing.
- Concurrency? To some extent? When testing it, it was running a lot slower than I would have liked. The requests seemed like the likely reason.
- Learning! More learning.