Hello there @Patrick-Beuks @lyrixx,

I'm just wondering:

Would this actually have to result in a deprecation warning?
For other tools using the DiffParser (like grumphp), we don't really care much about the raw index data.

See phpro/grumphp#1199

There are edge cases for parsing the diff that would result in an incorrect parse, not finding changed files, see issue #227.
I will check the use case as you described in your ticket to see if you would have hit the same edge cases tomorrow.

Oke, I have checked out your code.

You have a public interface on the diff in src/Git/GitRepository.php that returns the Gitonomy diff back.
Without the --raw this contains incorrect data and to protect consumers of this function from false information, I still think this deprecation is needed.

For example this this code for the spec/Git/GitRepositorySpec.php

    public function it_can_not_get_index_without_raw_without_changes(Repository $repository): void
    {
        $rawDiff = 'diff --git a/testfile b/testfile
old mode 100644
new mode 100755
';

        $diff = @$this->createRawDiff($rawDiff);
        $diff->shouldBeAnInstanceOf(Diff::class);
        /** @var File $file */
        $file = $diff->getFiles()[0];
        $file->getNewIndex()->shouldBe('');
        $file->getOldIndex()->shouldBe('');
    }

    public function it_can_get_index_with_raw(Repository $repository): void
    {
        $rawDiff = ':100644 100755 e69de29 e69de29 M        testfile

diff --git a/testfile b/testfile
old mode 100644
new mode 100755
';
        $diff = $this->createRawDiff($rawDiff);
        $diff->shouldBeAnInstanceOf(Diff::class);
        /** @var File $file */
        $file = $diff->getFiles()[0];
        $file->getNewIndex()->shouldBe('e69de29');
        $file->getOldIndex()->shouldBe('e69de29');

    }

    public function it_can_get_index_with_changes(Repository $repository): void
    {
        $rawDiff = 'diff --git a/testfile b/testfile
index e69de29..9daeafb 100755
--- a/testfile
+++ b/testfile
@@ -0,0 +1 @@
+test
';
        $diff = $this->createRawDiff($rawDiff);
        $diff->shouldBeAnInstanceOf(Diff::class);
        /** @var File $file */
        $file = $diff->getFiles()[0];
        $file->getNewIndex()->shouldBe('e69de29');
        $file->getOldIndex()->shouldBe('9daeafb');

    }

As you can see here without raw or changes the object is in an undesirable state.

(BTW thank you for letting me take another look at this, as I have found a bug with multiple files as fileIndex never increases)

Thanks for looking into it.
Sorry for my questions here, I'm just trying to understand :)

I'm trying to understand what is problematic about this, but can't really seem to find out why this should have negative impact on our application:
We don't use the new / old index anywhere. We only try to find the files from the git diff.

The way I see it:

• A diff can have different formats. When no --raw flag is passed, it will be missing some information indeed.
• A diff parser is responsible for parsing diffs in most common formats, but if some information is not available it just doesn't provide this information to the result of the parse.
• It's up to the piece of code consuming the diff to figure out it has all the information it needs.

An implementation could for example check if the indexes are set from the diff to check it is a file mode change or a difference in file content as well, without parsing the raw headers.

GrumPHP uses a diff as STDIN for the run command, so we don't always have control about the exact command the diff is created with. (we do have some presets of git hooks in which we have control, but users have the ability to overwrite those.

Without raw Diff contains most of the times the index information. This library even expects index information always to be available on the Diff object.

If you do not use the index information I would suggest to suppress the deprecation warning and wrap the Diff object in your own object and remove the index information to make is clear to users of your library that it is not accessible.

On the otherhand, if users can write their own diff, then it might be an idea to let the deprecation stand so that they can add the raw flag

Sidenote: this function does more then just get changed files, if you just need that there might even be faster option as this does parse the entire diff

(BTW I am just defending my justification for the deprecation as an end user of the library, in the end it is up to @lyrixx if this is a valid use case)