Is there any script that lets one remember branch/commit seperatly for files in some directory so that one can simultaneously work on file1 on branch1 and file2 on branch2 in the same directory and have them commit appropriately. If not I'll implement it myself.
My plan is to have hidden checkout directories for various branches/repos and populating the apparent checkout with links to these files so that commits simply committed their respective hidden branches but advice would be appreciated. Thus one could do something like
mgit checkout branch1 filename/filegroup
mgit add filename (automatically to it's correct branch/repo)
mgit commit (automatically to it's correct branch/repo)
Before anyone tells me this is a bad idea or incorrect use of version control consider the application:
So often when writing academic documents in TeX you end up with loosely related files that you wish to keep in version control. The problem is that these files are related and can benefit from the history/merging info git provides but often different versions of the files need to be selected independently. Most commonly you simply want to work on variant1 of some paper while also working of variant2 of another but you also have situations where TeX combines several files so it's important to be able to put your choice of variants in the same folder.
In short you have a bunch of distinct files with distinct names which often derive from each other (thesis becomes paper), benefit from being in a single repo but it is unwieldy to have a dozen different checkouts to work on whatever versions of each file currently need attention.
No, I don't think this is natively supported by git.
May be a local hook like a pre-commit one hook might try to check and checkout the right branch before allowing the commit to proceed.
来源：https://stackoverflow.com/questions/7784825/file-level-tracking-in-git-files-from-multiple-branches-in-same-directory