Importing BaseModel from aws_lambda_powertools.utilities.parser vs pydantic directly?

[xkortex]

I noticed in the docs some examples use from aws_lambda_powertools.utilities.parser import BaseModel. Normally, I'd just use from pydantic import BaseModel. In my understanding, it is fairly atypical for a 3rd-party library to re-export another library's objects. I know this library relies on pydantic and with v2 there are several breaking changes, so I'm wondering if it is necessary to use from aws_lambda_powertools.utilities.parser import BaseModel in production code, or if it makes no difference to import directly from pydantic? The latter is far preferable (it confused me when I first dug into my new team's codebase). Is this some sort of shim?

We are currently on pydanticV1 but I want our team to rip off the bandaid sooner rather than later to upgrade to v2.

This seems to be the commit where there's a comment on the wildcard import/export, which is a bit of a headscratcher.

[heitorlessa]

no need at all - it was a poor judgement's decision to help non-developers use models among other things.

In a future v4/v5 version, I'd love to a custom Python's import hook to start addressing this in a backwards compatible fashion, deprecate it, and get rid of it.

If you've bandwidth, we'd love help in updating the docs to use a pydantic import instead.

[xkortex]

Great to know! And yeah I'd love to help update docs but I need to check with my company.

get rid of it.

it meaning pydantic1?

Python has some of the most powerful import machinery of any language I've ever encountered, I'm sure there exists a way to polyfill ALP such that the pydantic version is (mostly) irrelevant.

[heitorlessa]

yeah, we've started working on the next major version (v3) - RFC - and we will remove pydantic v1 support (today we support both).

Couldn't agree more on import machinery. It is possible, just haven't had a chance to craft one to remap imports, as breaking change is a huge deal for us.