feat: implement DenseSet::Shrink #6094
Conversation
vyavdoshenko
requested review from
BorysTheDev,
dranikpg,
kostasrim and
romange
dranikpg
left a comment
dranikpg
left a comment
There was a problem hiding this comment.
It seems to me that all those internal calls to Delete() are now not safe anymore. The safeguards are not a substitute for actually reviewing the code around it and its usage
|
@dranikpg I've already discussed with @vyavdoshenko all these problems; he rewrites all this logic. |
vyavdoshenko
changed the title
![augmentcode[bot]](https://avatars.githubusercontent.com/in/1027498?s=60&v=4){kind=small}
vyavdoshenko
force-pushed
the
bobik/shrink_dense_set
branch
from
038c970 to
c23eba4
Compare
dranikpg
left a comment
dranikpg
left a comment
There was a problem hiding this comment.
After multiple rounds of review by Borys I don't have much comments 🙂 Please don't forget to set set_time (see comments)
| if (has_ttl && ObjExpireTime(obj) <= time_now_) { | ||
| ObjDelete(obj, true); | ||
| --size_; | ||
| continue; | ||
| } |
There was a problem hiding this comment.
this will currently never happen because you never call set_time() from the SHRINK command, you can look at container_utils::GetStringMap for an example
| << CI{"LATENCY", CO::NOSCRIPT | CO::LOADING | CO::FAST, -2, 0, 0, acl::kLatency}.HFUNC( | ||
| Latency) | ||
| << CI{"MEMORY", kMemOpts, -2, 0, 0, acl::kMemory}.HFUNC(Memory) | ||
| << CI{"SHRINK", CO::WRITE | CO::FAST, 2, 1, 1, acl::kMemory}.HFUNC(Shrink) |
There was a problem hiding this comment.
haven't seen what it was before, but I assumed it has to be part of MEMORY. It won't be as easy to a transaction then, because you'll need to create one manually
| continue; | ||
| } | ||
|
|
||
| uint32_t new_bid = BucketId(obj, 0); |
There was a problem hiding this comment.
based on the hash functon, it looks like 2*i and i are mapped to -> i. Not sure if its worth dcheck'ing or commenting, because the code looks more generic than it already is
Closes: #6044
Add
DenseSet::Shrink()to reduce the hash table size when sets/hashes have excess capacity after deletions.