sqeleton vendoring PoC

[vvkh]

Basically

mkdir vendor
cd vendor
git clone git@github.com:datafold/sqeleton.git
rm -rf sqeletion/.git
rm -rf sqeleton/.github

+ updates in the pyproject.toml file

[erezsh]

What is the point of this change?

[vvkh]

@erezsh we briefly discussed the idea with @wvanbergen with the primary goal of improving time to market and simplifying the release process.

[wvanbergen]

Because we shelved our plans for the other tool that would use the sqeleton library, it doesn't really make much sense to have it in a separate repository anymore. It makes contributing to datadiff harder than it has to be, and makes releases a more complicated process. So we decided to merge these repositories back into a single codebase.

[nolar]

Thank you for taking care of it. Luckily, we have no rush on this merging, so I suggest that we take a different route — the one that preserves the git history of changes and embeds sqeleton as one extra top-level package into the library — instead of bringing it as a vendored subdirectory which requires import-path injections.

Essentially, that is git remote add sqeleton … + git merge sqeleton/master --allow-unrelated-histories, and then resolve the conflicts on the same-named files (mostly CI & metadata). I am already working on it.

[nolar]

On a side note: I wonder, will this approach also package the files properly into tarballs & wheels and install it in a clean environment that is not git-based (i.e. pip install data-diff)?

[nolar]

Here is an alternative proposal with the git history preserved: #485

[williebsweet]

@nolar @wvanbergen Should we close this?