amberframework
Read JSON::Any Columns as `.to_json`'d String
Greetings 👋
Submitting this small PR as I recently ran into an issue when using import for a model. If I created the model one-off and used .new (add attribute values) then .save, the model would save just fine. If I instead set up a bunch of models then used .import, I'd get an error that stemmed down to this issue.
The short of it is that each adapter's import (pg's here) calls read_attribute under the hood and passes the then-read value along to the query builder but does so with its native type. If you ask a model for the value of a JSON::Any-typed column, you'll get a JSON::Any typed value. The query builder doesn't like that type though and will error.
This PR injects a .to_json in the process for any column whose type includes JSON::Any. This does indeed convert the value to a string and makes the query builder happy, but this PR may well be a proof of concept. There is definitely a better way of doing this, and this may overall be a stop-gap where correctly wiring up the Granite Converter(s) could be the right choice. Submitting this to explain my story, give some context, and hopefully inspire a better path forward!
Cheers!
Thanks for bringing this to my attention. I hadn't gotten notice of any PR's put on this repo, so I'm sorry about the late reply!
I'll have to dig into this more because I want to expand how Granite handles types and help improve it.
Have you joined our Discord yet? https://discord.gg/vwvP5zakSn
Closed by accident.
Ultimately this sounds like the import needs to be updated to handle bulk importing columns with JSON::Any types.