website repo is very big

[ntrel]

Just thought I'd mention git cloning the website seems like it will download > 1GB of data! I'm writing this as it clones. Is this really necessary? It's very surprising.

[ntrel]

I cancelled the clone and used --depth 1. Apparently that makes git fetch slower though.

[Verdagon]

Oh wow. You're not kidding, that's ridiculous!

I think I know what's going on: until recently, we've had all past release binaries as part of this repo (~80mb per release per OS) and once it was getting too big I removed them, but I bet they're still in the git history!

I'll find a way to cut them out of the history, and if all else fails, I'll just nuke the git history for this repo.

Thanks for reporting this! I likely wouldn't have found this for a while, since I already have a clone on my machine.

[spartanatreyu]

There's two options that come to mind:

1. Rip out all the past binaries from the repo
2. Keep the binaries, but migrate them out of the git repo in to git-lfs

Option 1: Binaries begone

If you want to rip out all the old binaries then I'd suggest using: git-filter-repo.

You can use it to automatically rewind each commit back to the first commit, then replay each commit with a few changes that you specify as CLI options.

That could be things like:

• Dropping each file larger than 8 MB
• Renaming a path from ddocs to docs
• Replacing the string SuperSecretApiKey to Redacted

Option 2: Binaries yeet

If you still want to keep the binary files but not keep them inside the repo, you can use git-lfs-migrate. This will upload the binaries up to a git-lfs storage host and replace the binary files in your repo with tiny pointer files that contain a link in the file on the git-lfs storage host.

[Verdagon]

I had no idea that git-lfs-migrate was a thing! I think that's exactly what we need here. Thanks for the tip!