ivy-completion-in-region: correctly calculate the length of replacement for file category

[kiennq]

• Ensure accurate calculation of the common length for completion items. The string face may contain only a single completions-first-difference, rather than a list of them.
• Additionally, when dealing with file category completions, calculate the prefix offset from the file name directory, not the entire input string. This adjustment is necessary because fuzzy completion styles can lead to incorrect common length values. The prefix offset ensures that we replace only the relevant portion of the input when calculating completion items.

[basil-conto]

Thanks! I'll try to have a closer look at this soon.

In the meantime, I'm curious: does this relate at all to the issues described in #1755?

[kiennq]

Thanks! I'll try to have a closer look at this soon.

In the meantime, I'm curious: does this relate at all to the issues described in #1755?

Yes, it should address that issue. The first item is actually used to fix that.

Ensure accurate calculation of the common length for completion items. The string face may contain only a single completions-first-difference, rather than a list of them.

[TristanCacqueray]

Hello, this fixes #1755 and #3056 for me too after updating to the emacs-30 branch. Thanks!

[offby1]

Hello, this fixes #1755 and #3056 for me too after updating to the emacs-30 branch. Thanks!

Same here; I've been using it for a few days with GNU Emacs 30.0.91 (build 1, aarch64-apple-darwin24.0.0, NS appkit-2566.00 Version 15.0.1 (Build 24A348)) of 2024-10-22 and it's fine.