Add support for reference cloning

[johnslavik]

(offtop) Hi! First of all, I've been searching exactly this tool for a while now. I'm glad to have had the luck to run into it by accident and that I'm not the only one who had this exact workflow in mind -- thank you!

---

I assume that the main audience of this project is people who already have a CPython repository cloned locally.

I'm proposing adding support for reference cloning, which tells Git to borrow objects from the existing local repository instead of pulling them from the internet. The speedup can be quite big; let me know if you're interested in more data on that.

I'd mainly use every-python to attach snippets running uvx every-python commands in CPython tickets so that others could quickly copy-and-paste them to reproduce differences in behaviors between revisions. Most people reading things there already have a CPython repo clone locally, so it would be especially cool if every-python could use that optimization 🚀

A user would have to indicate where their existing repo is, possibly like every-python --from-cpython-repo . (or whatever is least confusing), and then in _ensure_repo() we'd pass --reference-if-able path/to/cpython, and ideally --dissociate to make sure the borrowed objects become standalone copies so that when the reference repository gets deleted, the clones of every-python remain intact. --dissociate is a no-op without a reference repository, so no worries if the reference repository isn't found but the clone proceeds because of --reference-if-able. Since this flag causes incorrect reference repository directories to not throw a terminal failure, I'm tempted to propose that the current working directory is the default for --reference-if-able, but I'll understand if that's not preferred to expose such implicit behavior without a better reason than my personal workflow.

I've already used uvx every-python to demonstrate something in python/cpython#86153. I can attach --from-cpython-repo . by default, and note at all times that the command best runs in the CPython clone, to achieve the desired effect.

And of course, I can work on this patch if you think that it's worth it (I'm OK with a "no", sometimes simpler is better).

[savannahostrowski]

Oh, this is an interesting idea. I didn't know about --reference-if-able so that's neat.

I'm definitely open to the idea, especially if it makes the tool more useful to others! I wonder if this could be something configurable so you don't have to pass --reference every time. Maybe in a config file and/or able to be set via env var?

[johnslavik]

I wonder if this could be something configurable so you don't have to pass --reference every time. Maybe in a config file and/or able to be set via env var?

Do you mean Git or every-python?

I think it would be very useful to support an environment variable in every-python for producing --reference-if-able argv to the underlying Git clone command. I think it would be easier to start with an environment variable first (say, CPYTHON_REFERENCE_REPO) and not bother with a configuration file yet. We can add a configuration file as needed, ideally when someone with a particular use case requests it (so that our configuration file format would be easy for them to set up). What do you think?

If the question was about Git, I don't think it has an option to configure reference clones through environment variables or configuration files. Any configuration relevant to the alternates mechanism is only for changing behavior of the gc and submodules.