On Thu, May 16, 2024 at 11:24:36AM -0700, Junio C Hamano wrote: > Patrick Steinhardt <ps@xxxxxx> writes: > > > Reference backends have two callbacks `init` and `init_db`. The > > similarity of these two callbacks has repeatedly tripped myself whenever > > I was looking at those, where I always had to look up which of them does > > what. > > > > Rename the `init_db` callback to `create`, which should hopefully be > > clearer. > > Hmph, create() may be clearer than init_db(), but then I am not sure > what init() would do, differently from create(), so this rename > takes me back to the puzzled square one state X-<. > > I am guessing that create is about creating on-disk structure, while > init is about in-core structure out of an existing on-disk > structure? Once I understand the differences in these two things, > it is much less troublesome to tell them apart, regardless of what > they are called. Between .init and .init_db, it would be obvious > that the latter is about on-disk thing, without a rename done by > this step. On the other hand, contrast between <create, init> is > just as opaque as <init_db, init>---the names do not tell readers > that these two are about on-disk and in-core structures. > > Just my confused impression. I certainly wouldn't claim that `create()` vs `init()` is perfect, either. We could easily avoid that confusion if we were happy to make it more verbose, e.g. by calling it `create_on_disk()`. I don't really think that we also need to do the same for `init()`, mostly because it is a common idiom in our codebase to have `init()` initialize data structures. But that may also be my own personal bias. I'll go with `create_on_disk()` for now. It's more verbose, but given that this function isn't called all that much I think it's fine overall. Patrick
Attachment:
signature.asc
Description: PGP signature
