Versioning when binary AND text files are involved

问题:

I have a project where I need to maintain changes to both the text and binary files. I have a couple of options:

1. Use patches
2. Use a versioning system like git or hg.

For my purposes, patches are a better option if it was only text files. However, since there are images that might be replaced/added/deleted, which is the best way to go?

Is there a clean diff/patch utility that can take care of binary differences as well (without me having to specify it is binary -- I should be able to diff the entire directory and not individual files, which I can't with bash's diff in binary mode) and use them as patches? If not, which versioning system is a cleaner option when it comes to binary files?

回答1:

Mercurial has the notion of patch queue (with the mq extension enabled), largely inspired by quilt.
This allows you to manage your patches in several ways:

• Use a single patch queue and have all your patch order sequentially in it
• Use several patch queues, with the qqueue command, and have patches grouped by whatever criteria fits you.

As a bonus, as patches queues are really patch files, you can easily change their order and even move/copy them from one queue to another (have a look in your .hg directory to find the patch queues).

You can find a lot more of usefull information about managing patch queues at Steve Losh's tutorial on mercurial queue.

回答2:

Both Git and Mercurial can easily handle both text and binary files. Use whichever you prefer. And yes, a VCS is the right choice.

回答3:

If you don't like a VCS as such, try using a ftp server containing patches. Version numbers can be used as directory names. This system can be backed with a script, which applies patches sequentially based on directory names.

In the project I'm working on, database patches are applied in this manner. However, they are maintained inside a VCS. A file can be used to maintain the current version, which will be read by the script to decide which patch to apply.