Git为哪些二进制文件保存增量？含大量二进制的大型项目Git迁移疑问

Great question! Let's break this down step by step since dealing with large binary files in Git can get tricky, especially when you can't exclude them.

How Git Stores Binary File Deltas #

By default, Git handles text and binary files very differently when it comes to versioning:

• For text files, Git automatically uses delta compression: it stores only the differences (deltas) between successive versions, which saves a ton of space.
• For binary files, Git defaults to storing the entire file every time it changes. That said, Git can apply delta compression to binaries—you just need to explicitly enable it for specific file types. When you do, Git uses its built-in xdelta-based algorithm to calculate the binary differences between versions, just like it does for text.

Which Binary Files Don't Benefit from Delta Compression? #

Not all binaries are good candidates for delta storage. Here are the main ones to avoid:

• Pre-compressed files: Things like *.zip, *.7z, *.rar, or even media files like *.jpg, *.mp4—these are already tightly compressed, so even small changes to the underlying content result in massive binary differences. Storing deltas here usually wastes space instead of saving it.
• Large files with frequent, full updates: Think database backups, disk images, or large video files where every change replaces most of the file's content. Deltas won't help here because there's almost no overlap between versions.
• Binaries with non-deterministic builds: Some *.dll or *.exe files might include timestamps or build IDs that change every time you compile, even if the functionality is identical. The delta would be almost as big as the full file, so you're better off either fixing the build to produce deterministic outputs or using Git LFS instead.

Do You Need to Explicitly List File Types in .gitattributes? #

Git does have a small set of default rules for identifying text vs binary files (it checks for null bytes or unusual line endings), but it does not enable delta compression for any binaries by default. So if you want Git to use delta storage for specific binary types (like *.dll), you must define this in your .gitattributes file.

For example:

# Enable delta compression for DLL files
*.dll delta

# Use Git LFS for ZIP files (better for large/compressed binaries)
*.zip filter=lfs diff=lfs merge=lfs -text

Explicitly defining these rules is also critical for consistency across your team—different developers might have different Git configurations that could misclassify files, leading to bloated repositories or unexpected behavior.

内容的提问来源于stack exchange，提问作者carlspring