> Is hg as fast as git?

No. You still have to launch the Python process per command, which is "extremely" slow (relative to C programs). I air-quote extremely because the difference for a single command is small relative to human time, but it matters for automation that invokes many commands.

This post is from 2018, and maybe Hg's frontend is written in C or Rust now (I don't follow its development). But if it is still a Python-driven runtime, the same basic criticisms apply:

https://mail.python.org/pipermail/python-dev/2018-May/153296...

Personally, I like pushing with hg, but I prefer pulling with git. An HgHub service at the scale of GitHub would probably result in massive end-user UX issues but maybe happier developers and repo administrators.

This reminds how pleasant it was to use hg as a git front-end (via an extention[0]) - compared to using the git cli. I think I'll have to revisit and see if that's still the case, and if I can make it work with feature branches etc.

Also looks like there's finally python3 support in the recent mercurial 5.0 release.

[0] https://hg-git.github.io/

Git isn't really all that fast for what it is. The "git is fast" thing came about as a comparison to non-distributed version control systems, where changing what branch you were working on could involve coordinating with some (often overburdened) server on the other side of the planet.

That's not to say git is slow of course. It's just.. basically what you'd expect given it's architecture.