Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

db: refactor table cache newIters #3247

Merged
merged 1 commit into from
Jan 31, 2024

Conversation

jbowens
Copy link
Collaborator

@jbowens jbowens commented Jan 26, 2024

This commit refactors the interface of the table cache newIters method to support creation of any of a sstable's iterators, including its point iterator, range deletion iterator and range key iterator. This is motivated by the refactor planned in #2863, which in turn requires refactoring some of the newIters call sites that determine whether ingested sstables overlap existing tables.

There's also a minor benefit that a compaction setup no longer needs to open point iterators and then immediately close them when initializing the range deletion iterators.

goos: linux
goarch: amd64
pkg: github.com/cockroachdb/pebble
cpu: Intel(R) Xeon(R) CPU @ 2.80GHz
              │ old-NewItersAlloc.txt │       new-NewItersAlloc.txt        │
              │        sec/op         │   sec/op     vs base               │
NewItersAlloc             427.0n ± 1%   412.1n ± 1%  -3.47% (p=0.000 n=10)

              │ old-NewItersAlloc.txt │     new-NewItersAlloc.txt      │
              │         B/op          │    B/op     vs base            │
NewItersAlloc              0.000 ± 0%   0.000 ± 0%  ~ (p=1.000 n=10) ¹
¹ all samples are equal

              │ old-NewItersAlloc.txt │     new-NewItersAlloc.txt      │
              │       allocs/op       │ allocs/op   vs base            │
NewItersAlloc              0.000 ± 0%   0.000 ± 0%  ~ (p=1.000 n=10) ¹
¹ all samples are equal

@cockroach-teamcity
Copy link
Member

This change is Reviewable

@jbowens jbowens requested review from a team and itsbilal January 26, 2024 22:04
Copy link
Collaborator

@sumeerbhola sumeerbhola left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

:lgtm:

Reviewed 20 of 20 files at r1, all commit messages.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @itsbilal and @jbowens)


table_cache.go line 166 at r1 (raw file):

	}
	if iter == nil {
		iter = emptyKeyspanIter

if the caller is going to place this in an iterSet can't we just return nil, since the iterSet handles the nil => non-nil empty iter transformation itself?


table_cache.go line 546 at r1 (raw file):

			// Return an empty iterator. This iterator has no mutable state, so
			// using a singleton is fine.
			return filteredAll, err

same question about letting iterSet handle the nil iter case.


table_cache.go line 1286 at r1 (raw file):

// bitwise-OR iterPointKeys, iterRangeDeletions and/or iterRangeKeys together to
// represent a set of desired iterator kinds.
type iterKinds int8

nit: how about uint8, given it is a bitmap.

Copy link
Member

@RaduBerinde RaduBerinde left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewable status: all files reviewed, 6 unresolved discussions (waiting on @itsbilal and @jbowens)


level_iter.go line 50 at r1 (raw file):

// all returns an iterator set containing iterators over all the keys that meta
// contains.
func (f tableNewIters) all(

This method on a function is a bit dubious. I'd consider adding an iterAll constant and call it directly with that instead.


level_iter.go line 54 at r1 (raw file):

) (iterSet, error) {
	var kinds iterKinds
	if file.HasPointKeys {

[nit] Feels like this optimization belongs inside the implementation, not here


table_cache.go line 1248 at r1 (raw file):

}

// RangeDeletion returns the contained range deletion iterator. If there is no

[nit] Will something like HasRangeDeletion() bool be useful for fast paths?

Copy link
Collaborator Author

@jbowens jbowens left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TFTRs!

Reviewable status: 16 of 21 files reviewed, 4 unresolved discussions (waiting on @itsbilal, @RaduBerinde, and @sumeerbhola)


level_iter.go line 50 at r1 (raw file):

Previously, RaduBerinde wrote…

This method on a function is a bit dubious. I'd consider adding an iterAll constant and call it directly with that instead.

Done.


level_iter.go line 54 at r1 (raw file):

Previously, RaduBerinde wrote…

[nit] Feels like this optimization belongs inside the implementation, not here

Done.


table_cache.go line 166 at r1 (raw file):

Previously, sumeerbhola wrote…

if the caller is going to place this in an iterSet can't we just return nil, since the iterSet handles the nil => non-nil empty iter transformation itself?

I removed this entire func; this was a remaining top-level table cache entry point that did not use iterSet in order to adapt to the function signature of keyspan.TableNewSpanIter. I replaced it with a new func tableNewRangeKeyIter that takes a newIters and returns a function with the appropriate signature.


table_cache.go line 546 at r1 (raw file):

Previously, sumeerbhola wrote…

same question about letting iterSet handle the nil iter case.

Added a comment here. This one is necessary because of the MaybeFilteredKeys() interface, but it will be go away with #2863.


table_cache.go line 1248 at r1 (raw file):

Previously, RaduBerinde wrote…

[nit] Will something like HasRangeDeletion() bool be useful for fast paths?

Hm, yeah, I think it might—I'll defer it for now but after #2863 we can look at the usages and adapt. Left a TODO


table_cache.go line 1286 at r1 (raw file):

Previously, sumeerbhola wrote…

nit: how about uint8, given it is a bitmap.

Good call, done.

This commit refactors the interface of the table cache `newIters` method to
support creation of any of a sstable's iterators, including its point iterator,
range deletion iterator and range key iterator. This is motivated by the
refactor planned in cockroachdb#2863, which in turn requires refactoring some of the
newIters call sites that determine whether ingested sstables overlap existing
tables.

There's also a minor benefit that a compaction setup no longer needs to open a
point iterator and then immediately close it when initializing the range
deletion iterators.

```
goos: linux
goarch: amd64
pkg: github.com/cockroachdb/pebble
cpu: Intel(R) Xeon(R) CPU @ 2.80GHz
              │ old-NewItersAlloc.txt │       new-NewItersAlloc.txt        │
              │        sec/op         │   sec/op     vs base               │
NewItersAlloc             427.0n ± 1%   412.1n ± 1%  -3.47% (p=0.000 n=10)

              │ old-NewItersAlloc.txt │     new-NewItersAlloc.txt      │
              │         B/op          │    B/op     vs base            │
NewItersAlloc              0.000 ± 0%   0.000 ± 0%  ~ (p=1.000 n=10) ¹
¹ all samples are equal

              │ old-NewItersAlloc.txt │     new-NewItersAlloc.txt      │
              │       allocs/op       │ allocs/op   vs base            │
NewItersAlloc              0.000 ± 0%   0.000 ± 0%  ~ (p=1.000 n=10) ¹
¹ all samples are equal
```
@jbowens jbowens merged commit fea6cbb into cockroachdb:master Jan 31, 2024
11 checks passed
@jbowens jbowens deleted the tablecache-refac branch January 31, 2024 16:59
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

4 participants