On 10/17/2021 1:58 AM, Junio C Hamano wrote: > Derrick Stolee <stolee@xxxxxxxxx> writes: > >> I think you are slightly mistaken here: If index.sparse=true, then a >> full index will be converted to one on write, but not immediately upon >> read. This means that subsequent commands will read a sparse index, and >> they will either benefit from that or not depending on whether they are >> integrated with the sparse index or not. >> >> The new behavior here is that if index.sparse=false, then we convert >> a sparse index to a full one upon read, avoiding any chance that a >> Git command is operating on a sparse index in-memory. > > And if index.sparse=true, then we convert a full on-disk index to a > sparse one in-core upon reading, right? My comment was solely on > that side of the picture, not on the "index.sparse is set to false > so we automatically expand" case. > >> The simplest summary I can say is here: >> >> * If index.sparse=false, then a sparse index will be converted to >> full upon read. >> >> * If index.sparse=true, then a full index will be converted to sparse >> on write. > > Oh, I see, so yes I was very much misunderstanding what you guys are > trying to do. I somehow thought that sparse-to-full and > full-to-sparse conversions (1) already happen on the write codepath, > and (2) this patch makes them both happen also on the read codepath. > > IOW: > > * If index.sparse=false, a sparse index will be written as full, > and if it is true, a non-sparse index will be written as > sparse, even before these patches. This statement is true. > * In addition, with these patches, if index.sparse=false, a > sparse index will be expaned to full upon reading, and if it > is true, a non-sparse index will be shrunk to sparse upon > reading This is only half true. If "index.sparse=false", then a sparse index will be expanded to full upon reading. If "index.sparse=true" then nothing special will happen to an index on reading. > What your summary above is saying is very much different. Yes, these are different things. It also gets around the concern that "if we have a non-integrated command, then index.sparse=true would convert a full index to a sparse one, only to be expanded again". That doesn't happen if we only convert from full to sparse on write. Thanks, -Stolee
