Introduction

The line that starts with @@ in git diff or svn diff is the hunk header. It tells you where a block of changes belongs in the old file and in the new file, so both humans and patch tools can understand the edit.

Reading the Hunk Header

A typical unified diff hunk looks like this:

1@@ -12,4 +12,6 @@
2 line kept from the old file
3-old line
4+new line one
5+new line two
6 line kept after the change

The -12,4 part refers to the original file. It means the hunk starts at line 12 and covers 4 lines of old content.

The +12,6 part refers to the new file. It means the replacement block starts at line 12 and covers 6 lines in the updated version.

The header does not say that every line changed. It marks the range of the chunk, including unchanged context lines that surround the edit.

Why Diff Output Uses Hunks

Diff tools do not usually print the whole file. They print only the changed areas plus a few unchanged lines around each change. Each of those areas is a hunk.

That makes reviews readable and allows patch tools to apply the change later. If the target file shifted a little but the nearby context still matches, the patch can often still be applied.

Here is another example:

@@ -30,3 +30,3 @@
-return total / count
+return count == 0 ? 0 : total / count

This means the edit starts near line 30 in both versions. The old block has 3 lines and the new block also has 3 lines.

What the Prefixes Mean

Inside a hunk, line prefixes matter:

• A space means the line is unchanged context.
• '- means the line exists only in the old file.'
• '+ means the line exists only in the new file.'

Minimal example:

1@@ -8,2 +8,2 @@
2 const timeout = 10
3-const retries = 1
4+const retries = 3

Because one line was removed and one was added, the old and new counts both stay at 2 for that chunk.

Variations You Will See

Sometimes the count is omitted. For example, @@ -5 +5 @@ means one line in each file. The single-line count can be left out because it is implied.

Git may also append a label after the second @@:

@@ -42,7 +42,9 @@ function parseConfig(path) {

That trailing text is just extra context, often a function name or nearby code label. It is there to help the reader, not to change the patch semantics.

File-level headers appear above hunks and serve a different purpose:

--- a/app.js
+++ b/app.js

Those lines identify the old file and the new file. The @@ line describes the specific changed region within that file pair.

How This Helps When Applying Patches

Patch application depends on the hunk header plus the surrounding context. A patch tool does not blindly insert text at a line number. It uses the header as a guide, then checks whether the nearby context lines still match.

That is why a patch can sometimes apply even if line numbers drift slightly, and why it can fail if the surrounding code changed too much.

You can see the same structure when generating a patch file:

git diff > change.patch
git apply --check change.patch

The check step validates whether the hunks can be matched cleanly.

Common Pitfalls

A frequent misunderstanding is reading -12,4 as "four lines were deleted." It actually means the old side of the hunk spans four lines, including unchanged context.

Another mistake is confusing file headers with hunk headers. --- and +++ identify files; @@ identifies one changed chunk inside those files.

People also assume the trailing label after @@ affects how the patch is applied. It does not. It is explanatory text for humans.

Summary

• The @@ ... @@ line is the unified diff hunk header.
• '-start,count describes the old file range.'
• '+start,count describes the new file range.'
• Space, -, and + prefixes show unchanged, removed, and added lines.
• Trailing text after the second @@ is extra context, not patch logic.