Re: [PATCH v2 0/5] format-rev: add --abbrev, --color, and --date
- From
- Kristoffer Haugsbakk <kristofferhaugsbakk@fastmail.com>
- Date
- Nov 20, 2026, 14:34 UTC
- Message-ID
- <e6debee7-8645-43c3-b556-b39da7df6b64@app.fastmail.com>
- In-Reply-To
- <CALnO6CB9LKjJ+NeSacJCUNrhWLNj6CM2Oo+JR9c7ZdohQPJWow@mail.gmail.com>
On Sat, Nov 7, 2026, at 21:22, D. Ben Knoble wrote:
Show 40 quoted lines
> On Tue, Nov 3, 2026 at 5:46 AM Kristoffer Haugsbakk > <kristofferhaugsbakk@fastmail.com> wrote: >> >> On Mon, Nov 2, 2026, at 17:49, Ben Knoble wrote: >> >> Le 2 nov. 2026 à 05:51, Kristoffer Haugsbakk <kristofferhaugsbakk@fastmail.com> a écrit : >> >> >> >> On Mon, Nov 2, 2026, at 09:51, Junio C Hamano wrote: >> >>> Junio C Hamano <gitster@pobox.com> writes: > >> >> For you all’s information, what was originally totally unrelated >> >> research has made me doubt whether this command is needed.[1] >> >> git-diff-tree(1) cannot do what this command does right out of the box, >> >> but it looks like some modifications can make it a replacement for `git >> >> format-rev --stdin-mode=revs`.[2] >> > >> > I saw these come through, and I’m not sure where you landed. >> >> That git-diff-tree(1) just needs to learn to optionally NUL-terminate >> diff and log output.[1] >> >> I looked into it after that and the code has turned out to be too >> difficult for me to make sense of. >> >> [1] <238e5652-e95b-4d44-a900-e16aa6ab81f0@app.fastmail.com>; >> https://lore.kernel.org/git/238e5652-e95b-4d44-a900-e16aa6ab81f0@app.fastmail.com/ >> >> > I was mostly interested in the stdin filter mode that turns hashes >> > into references; is that supported by your alternative? >> >> No it’s not. `git format-rev --stdin-mode=text` still has its niche. >> >> >> [1]: <ce69d94e-7329-449d-b3c3-47d4f3aa608c@app.fastmail.com>; >> >> https://lore.kernel.org/git/ce69d94e-7329-449d-b3c3-47d4f3aa608c@app.fastmail.com/ >> >> [2]: <238e5652-e95b-4d44-a900-e16aa6ab81f0@app.fastmail.com>; >> >> https://lore.kernel.org/git/238e5652-e95b-4d44-a900-e16aa6ab81f0@app.fastmail.com/ > > Hm. In that case, what if we made diff-tree do the things you want > from, IIUC, the "revs" mode, and made format-rev work only in "text" > mode? That would make invoking it [say, from an editor ;)] nicer, and > also cut down its scope a bit.
Oh, yeah. I’m fine with keeping `format-rev --stdin-mode=text`. And we should of course simplify it to just `format-rev` if we remove `revs`.
So command removal or option removal: could go either way as far as I am concerned.
> Or would you still want the somewhat higher-level interface of revs > mode to be an option?
Being able to either do `revs` or `text` doesn’t seem to buy us much in itself.
The only thing that I don’t like about the git-diff-tree(1) approach is that it is completely undiscoverable. At least to me. I was only able to find it by happenstance. It makes no sense to be to use git-diff-tree(1) as a commit log plumber.
1. You need `-v` or `--format=...` to get a commit 2. But you also need `--always` in order to countermand the “no diff so no output” for *merge* commits
So this:
git diff-tree -v --always --no-patch
Reads like: do a diff verbosely, always, without a diff. Huh?
But again, I am totally fine with removing the command. It would depend on if anyone else would, perhaps, want to keep it as a wrapper/skin around these options. For me, I know now, so using the above or making wrappers is fine.