Re: [PATCH] wt-status: avoid quadratic insertion for untracked paths

From: Sahitya Chandra <sahityajb@gmail.com>
Date: Jul 17, 2026, 14:37 UTC
Message-ID: <CAP=WS+tb-HnMmOjH7i+ZY0EBOT0xfDPi4wsTboiH0mRtCCs=ng@mail.gmail.com>
In-Reply-To: <alnLPSnOt_Sf7cA5@pks.im>

On Fri, Jul 17, 2026 at 11:57 AM Patrick Steinhardt <ps@pks.im> wrote:

> Out of curiosity: is this something that you have encountered in the
> real world as inefficient, or is this rather a theoretical inefficiency?
> If the former it would be great to add a small benchmark to the commit
> message.

Thanks for asking. I do not have a real-world benchmark for this. After Jeff's reply, I agree that the O(n^2) claim is too strong for the current code path because fill_directory() already returns the entries sorted, so string_list_insert() should usually append at the end.

I have reworded v2 to avoid that performance claim and describe the change as making the collection strategy explicit instead.

> Instead of sorting and then deduplicating you can call
> `string_list_sort_u()`. It does the exact same thing as you do here, but
> I guess it makes sense to use that interface anyway.

Done in v2. Thanks.