Plan

Idea

• Use pydantic to (de)structure the requests and responses
• Raw api wrapper called API which handles authorization, wrapping each call in a function. Inputs and outputs are python types
• Higher level wrapper which will have objects representing the implicit objects in the API e.g. Task, List, Filter

Authorization layer

• Stores the key, secret etc
• Generates the api sig
• Makes generic authorized/unauthorized calls. Inputs are able to be coerced to strings. Outputs are dictionaries that come out of the json parsing

Wrappers for the specific RTM calls

• Inputs are proper types like datetime, enums, lists
• Outputs are parsed out into the same types (datetime, enums, lists etc)
• Should it throw RTM errors rather than return parsed fail objects? Probably, since it's possible with complete fidelity and fits with the way the code has to be written

Task objects

• Ordering of start/due dates
• Hide whether they're dates or datetimes
• Hiding of "no tag"/tag type inconsistency
• Coalesce sending of different attributes to the server with an explicit call - have to do that for start/due dates anyway
• Validate the repeat input values

List objects

Sometimes you only get the listid back, could hide the expansion of the other attributes

Filter objects

• Construct from string
• Construct from logical combination of conditions
• Output strings to the server using the pydantic stuff

Client layer

• First entry point
• Search for tasks
• CRUD for lists
• Holds the API object

Future?

Make it sansio, so that we can use other than requests