How does git diff tell if a line has been modified or added?

Question

Given two files how does git diff or any diff tool tell which lines have been added, modified or deleted?

Are there any edge cases where it may fail to recognise the correct status of the edited line?

EDIT: This question is more concerned with the actual method of how the diffing happens

_Does_ git show lines as "changed", ever? Or are they deleted and added? — CodeCaster, Jan 06 '20 at 14:36
Each commit references a tree of files. To output diff between commits A and B, git compares commit A's tree and commit B's tree. — Romain Valeri, Jan 06 '20 at 14:38
Have a look at a [related post](https://unix.stackexchange.com/questions/498581/how-to-tell-the-difference-between-changed-and-added-lines). — Axel Kemper, Jan 06 '20 at 14:39
This doesn't really have anything to do with `git`, but with diff algorithms in general. `git` just supplies the input files to whatever `diff` tool you use. — chepner, Jan 06 '20 at 15:09

score 4 · Accepted Answer · answered Jan 06 '20 at 18:49

Git stores snapshots: given any two snapshots of your choice, Git extracts the files you'd like compared¹ to a temporary area² and then runs a "diff engine" on the two files. The result of this engine is, or should be, a series of instructions that, when applied, will convert the left-side file to the right-side file.

Git has several built-in diff engines, which Git calls:

myers: This uses an algorithm by Eugene W Myers with some slight modifications. There are numerous StackOverflow questions and answers about Myers diff: Myers diff algorithm vs Hunt–McIlroy algorithm; Diff Algorithm? (multiple answers but several about Myers diff); Myers' diff: Why V[k − 1] < V[k + 1] guarantee to choose the further D-path?
minimal: same as myers but without some modifications done to go faster instead of more-minimal when the divide part of the divide-and-conquer starts behaving badly due to very long edit scripts. In a sense, this is the real Myers algorithm.
patience: Bram Cohen's algorithm; see the accepted answer to Where can I find the diff algorithm?
histogram: a modification of patience aimed at not accidentally synchronizing on things like brace-only lines.

It can also run your chosen external diff engine.

¹Or at least, the ones it wants to compare. If you are comparing files that are not actually in commits, they may already be extracted.

²Git mostly does this in-memory, but when running an external diff, really does use temporary files.

See also [chapter 3](http://web.torek.net/torek/tmp/book.pdf), by the way. — torek, Jan 07 '20 at 02:43

score 0 · Answer 2 · answered Jan 06 '20 at 14:58

git keeps the current committed version of every single file in the repo. If a file is modified, then the committed version is generated and compared with the current version of your file. It is then the same as diff'ing two files on the hard drive. If the diff tool can diff two ordinary files, then it can diff the version that git produced and the version that you have.

Unless if there is some really obscure bug, the diff tool should be able to catch any difference, as the diff tool compares the two files line by line.

How does git diff tell if a line has been modified or added?

2 Answers2

Linked