Download the binary for your platform and place it somewhere within your $PATH. The git-p statically links libc so the same binary can be used across linux distributions.
git poly depends on git. So you must have git installed.
At the moment there is only a linux pre-built binary. Get in contact if you can provide machines for building on osx or windows.
Alternatively, you can build it quite easily yourself using cargo from rust.
| Platform | Version | Instruction Set | Download |
|---|---|---|---|
| Linux | Kernel 4.19.x | x86_64 | here |
A tool to help developing across 50+ git repos easier. It's a simple, light weight tool to parallelize working with many git repos.
Make working with many git repos feel like working in a single git repo for most day to day operations, while still allowing those repos to be completely seperate.
- Very fast!
- Almost all operations are done in parallel
- Multi-platform (mac, linux, windows)
- Searching for git repos is done asynchronously, the moment we have found a git repo a new thread is created to process it.
- Regex based find and replace
- There's no config/manifest file.
- Aligns very closely with git, only four more very straight forward commands are added.
- Track repo dependencies
- Manage sub repo / parent repo histories
- Specify how to organise your repos
You can use these along with 'git poly'.
What's the point of git-poly. Why don't I just use one of the above? git-poly fills a specific need. An unintrusive, simple way to make changes to multiple repos at once.
If you have lots of little repos that depend on one and other, and you find that you are often changing a few of them at a time when you develop a new feature. You want the histories to remain independant, tags and releases to be independant but you still need to change a few of them at a time.
git-poly is written in rust. You can find the source code here:
git poly doesnt have a manifest or configuration file, so it's up to you to manage that.
The simplest way to clone the repos is git clone.
git clone https://github.com/openssl/openssl.git
git clone https://github.com/libjpeg-turbo/libjpeg-turbo.gitIf you already have a setup in place that tracks the repos you are working with, then you can take advantage of 'git p clone' to download everything in parallel.
Imagine for example you have a yml configuration file that looks like this.
repos:
- https://github.com/openssl/openssl.git
- https://github.com/libjpeg-turbo/libjpeg-turbo.gitThen on linux to clone them all in parallel you can use grep.
grep --color=never --only-matching 'https.*\.git' config.yml | git p cloneIt's important to note that 'git p clone' will maintain the folder path in the url. This is different from git 'clone' which will clone into a single folder.
git poly will show you the status of all the files in the repos you have checked out. It combines them to make it appear as if you are working in a single git repo.
To perform a parallel 'git status' across all the repos and display the combined results you use 'git p status'
git p statusFor example imagine I edit both repos. Here I'm adding an additional comment to a header file and a c file, one in libjpeg and one in openssl.
echo '// extra comment' >> libjpeg-turbo/libjpeg-turbo/jsimd.h
echo '// extra comment' >> openssl/openssl/ssl/ssl_rsa.cNow if I run 'git p status' this is the result I'll see.
on branch master
Changes not staged for commit:
(use "git add <file>..." to include in what will be committed)
modified: ./libjpeg-turbo/libjpeg-turbo/jsimd.h
modified: ./openssl/openssl/ssl/ssl_rsa.c
Changes are grouped by the branches currently checked out. In the example above both repos are on the master branch.
If we switch libjpeg onto a branch called develop, git poly status will display the results differently.
Here we're running the git command 'git checkout -b develop' on all the repos which have 'libjpeg' in the path name. In this case, its only one repo.
git p --path libjpeg go checkout -b develop
Now if I run 'git p status', I get a different result.
on branch develop
Changes not staged for commit:
(use "git add <file>..." to include in what will be committed)
modified: ./libjpeg-turbo/libjpeg-turbo/jsimd.h
on branch master
Changes not staged for commit:
(use "git add <file>..." to include in what will be committed)
modified: ./openssl/openssl/ssl/ssl_rsa.c
To add those changes.
git p add ./libjpeg-turbo/libjpeg-turbo/jsimd.h
git p add ./openssl/openssl/ssl/ssl_rsa.cor
git p add -uTo commit
git p commit -m "Ive made a change to some files"Finally, there isnt a first class 'push' subcommand. So you have to go via the go subcommand.
git p go push
git poly can run 'git grep' in parallel over multiple repos. The benefit to using 'git p grep' over 'git p go grep' is that the resulting file paths are displayed with respect to the current directory.
git p grep hel
Similarly to 'git p grep', there is first class support for 'git ls-files', with resulting file paths correctly adjusted.
git p ls-files
This sub command is a utility for moving files from one repo to another.
git p mv ./openssl/openssl/ssl/ssl_rsa.c ./libjpeg-turbo/libjpeg-turbo/ssl_rsa.c
You can reset everything you have staged with 'git p reset'.
git p reset
Then if you want to throw away everything that's changed.
git p go checkout .
You can perform a parallel find and replace, using 'git p replace'. This will run a thread per repo and then a thread per file.
git p replace cat dog
This runs a git command across the given repos in parallel.
For example
git p go add -u
will stage any unstaged changes in any of the repos you are working in.
The same as 'go' command, only executes shell commands.