etc/reports/25-02.md
Like the previous month was the first in 2025 with plenty of news to share, this month is a first as well: I seem to have nothing to write about.
Sure, there were a couple of smaller API improvements, but nothing that would inspire me.
So what did I do all this time? Probably this is the moment where there really is no other way but to talk about GitButler, something I avoided last month just because it doesn't seem to belong into the gitoxide newsletter.
GitButler is what I have been working on intensely for the last two months effectively, in a push to help it to unfold its true potential. Thus far it was mostly powered by git2, with gitoxide sprinkled in there, but the next iteration will be the inverse with git2 only being used where gitoxide is lacking a feature. That way, along with massive architectural changes, it will be able to cope with large repositories and be more compatible with various Git features, too.
I cannot wait to see all this work to finally come to fruition, and of course, to also see myself being pulled to a user interface that truly elevates my workflow and the workflow of other devs just like me who thus far preferred to stay on the command-line.
gix blameChristoph Rüßler kept working and managed to greatly improve the gix blame performance while increasing its conformance to Git at the same time. This means that now, depending on the sample, gix blame can be a bit faster than Git, but it typically is still up to 30% slower when commitgraph caches are used. Overall though, the performance is nothing to sneeze at, and it competes quite well except for in pathological cases.
Admittedly, I am quite blown away by the performance and have a feeling that last time I checked, I didn't use the latest version of the gix binary.
It's worth noting that rename-tracking still isn't support, but I also see no reason why it shouldn't be eventually, a feature that would make gix blame so much more useful in practice.
gix blame with experimental cacheA pretty slim PR shows how to use a special cache to greatly speedup blames, from ~300ms down to just ~4ms, a massive 75x speedup that would be very useful for editors and IDEs, or forges, I am sure! Of course, one first has to build such cache, and probably do so per file, but I am sure there are plenty of use-cases for it when it's driven by dev tooling.
With gix status now available I am planning to integrate it as soon as possible! That didn't happen yet, but… it will, the stack to work off before I can start this is pretty high though so it's unlikely to happen anytime soon.
Cheers Sebastian
PS: The latest timesheets can be found here (2025).