Way to have git only recognize a file as changed if more than a specific set of lines change?

Question

In my current project, I am attempting to use git to version control text files that are utilized by software that generates code from them. This in itself isn't a problem, the problem is that every time I generate code, it automatically updates this file with properties such as the date the code was generated, as well as my name.

You can imagine it looking something like this:

SomeHeader{
  -SomeProperty : x
  -NameOfUserThatGenerateCode: myName
  -DateTimeCodeGenerated: 2013-07-23 06:28
  -SomeOtherProperty: y
}

What I want, is a way to tell git to say that "it is okay" if both the Name and CodeGeneration time to change (ie: Ignore that there was a change), but DO care if "SomeProperty" changes to say "z".

In that second case, it would commit the entire file (with the updated autogenerated files).

Is there anyway to do that? I recognize that git does changes at the "file" level, but I am hoping that there might be some sort of pre-processing hook that I can tie into that would only work when git attempts to compare file changes.

For those that care, this will enable me to properly version control rhapsody files.

Answer 1

I would recommend:

• keeping a copy of those files (as private files, meaning not versioned): your software would generate code in those copy
• a clean script, as a content filter driver, declared in a .gitattributes file (detailed in Git Pro book).

The idea is for that script to detect, on git add:

• the content of a 'SomeHeader' property file
• if the copy of that file has changed in way you need to keep (in which case you overwrite the versioned file with the content of that copy)
• if the copy of that file has not changed significantly, in which case, you don't modify the actual property file.

Note that the "keep a copy" part can be automated also by a content filter driver, with a smudge script activated automatically on git checkout.