Strict numpy dependency

[bjgbeelen]

Hi!

I'm running into an issue that this package requires a very specific numpy package 1.23.4. Aside from wondering why it needs numpy in the first place (I didn't dive into that), I'm wondering if this needs to be this specific.

Context:
We are using NixPkgs to provide reproducible environments. I'm trying to compile the latest dbt-databricks version and this is a dependency. From the latest stable NixPkgs I get numpy version 1.23.3. I can override it but building this specific version of numpy myself is quite heavy. And it makes me wonder if it can be solved in a different way by relaxing the constraint :-)

Thanks for thinking along!

[betodealmeida]

Same here, I'm having problems updating the Databricks driver in Apache Superset because of this. Can we relax the versioning here?

databricks-sql-python/pyproject.toml

Lines 22 to 25 in e3f3a6f

	numpy = [
	{version = "1.21.1", python = ">=3.7,<3.8"},
	{version = "1.23.4", python = ">=3.8"}
	]

[susodapop]

There's a PR for this here: #74 we're working on merging it ASAP.

[susodapop]

Also @betodealmeida we'd like to improve the Superset experience which currently doesn't use our built-in SQLAlchemy dialect (just released in v2.4.0). Are you interested in helping us test the new version?

[betodealmeida]

Also @betodealmeida we'd like to improve the Superset experience which currently doesn't use our built-in SQLAlchemy dialect (just released in v2.4.0). Are you interested in helping us test the new version?

Yeah, I'm happy to bump the version and test it! Will do it this weekend.

[susodapop]

@betodealmeida Unfortunately that won't work! Superset needs to be configured to use our in-built dialect. At present if you just bump the version of databricks-sql-connector nothing will change. Once we have concrete steps for testing I'll ping you here.