On 11/19/26 17:11, Laurenz Albe wrote: > On Thu, 2026-11-19 at 08:57 -0300, Luís Roberto Weck wrote: >> I've been working on extracting more information from Postgres statistics to >> better approximate some estimation errors I've faced in the past. >> >> I've never submitted a patch before, so before I go and do that, I'd love >> for some feedback on it from the hackers lurking here. >> >> Patch is in https://github.com/lrweck/postgres/pull/1 >> >> I am also working on other patches that I might share soon. > > +2450 lines? > > The right approach is to come up with a design first and to discuss it on > pgsql-hackers. > > Your monster patch seems to implement several features at once. Don't do > that. > Select a single, self-contained feature and do that. > Pick a simple one for your first patch. > > Make sure that you review and understand every single line of code that your > chatbot > generated and can answer questions about it. Dumping generated code on human > reviewers is impolite; people would resent the waste of their time. > > Read the pages with advice for beginning hackers on the PostgreSQL Wiki. > > Best hang around on pgsql-hackers for a while and review a couple of patches > in a commitfest, that will get you in tune with how things work. >
That's a bit harsh, but I agree the patch needs to be split into smaller pieces, to make review easier. It also makes it possible to commit one piece that's ready, and keep working on the rest. The ticket even splits it into 5 different cases, so why not to split the patch like that? It's possible an initial commit will introduce some shared infrastructure, or maybe the patches will have dependencies, but that's OK. If you split it like that, I'll take a look ... regards -- Tomas Vondra
