Apply binary search filter expressions directly on the block metadata of Index Scans
#1619
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Co-authored-by: Johannes Kalmbach <[email protected]>
Co-authored-by: Johannes Kalmbach <[email protected]>
Co-authored-by: Johannes Kalmbach <[email protected]>
Co-authored-by: Johannes Kalmbach <[email protected]>
…neral implementation
Co-authored-by: Johannes Kalmbach <[email protected]>
Co-authored-by: Johannes Kalmbach <[email protected]>
…ng correct prefix in Constants.h
joka921
requested changes
Nov 21, 2024
There was a problem hiding this comment.
This is very nice and was quite a lot of work.
Mostly missing are Tests that the Prefilter is applied in the indexScan etc.
A good way (also for manual debugging) is to add the applied prefilters to the runtimInfo() as a detail, then they appear in the UI, and this can also be used for testing.
joka921
requested changes
Nov 27, 2024
joka921
reviewed
Nov 29, 2024
joka921
requested changes
Nov 29, 2024
Conformance check passed ✅No test result changes. |
|
joka921
approved these changes
Dec 2, 2024
joka921
changed the title
PrefilterExpressionIndex in IndexScanIndex Scans
realHannes
added a commit
to realHannes/qlever
that referenced
this pull request
Dec 2, 2024
… of `Index Scan`s (ad-freiburg#1619) With this PR, filter expressions that can be evaluated via binary search on a sorted input are directly evaluated on the block metadata of an IndexScan. For example in a query that contains `{ ?s ?p ?o FILTER (?o > 3)`} only the blocks of the full index scan (sorted by the object) are read from disk that according to their metadata might contain values `> 3`. Currently this mechanism has the following limitations: 1. It can only be applied if the IndexScan directly is the child of the FILTER clause 2. It can only be applied to logical expressions (AND/OR/NOT) and to relational expressions (greater than, equal to, etc.) between a variable and a constant. Currently the constant can not yet be an IRI or Literal.
joka921
pushed a commit
that referenced
this pull request
Dec 6, 2024
With this PR, the prefilter expressions implemented in #1619 also apply to literals and IRIs. For example the following query only extracts the relevant, prefiltered blocks from the `IndexScan`: ``` SELECT * { ?s ?p ?o FILTER (?o >= "hallo" && ?o <= "hello") } ```
hannahbast
pushed a commit
that referenced
this pull request
Dec 12, 2024
Since #1619, the size estimate for an index scan always involved one or several copies of the block metadata, which incurred a significant query planning cost for most queries. Now, such a copy is only made for an index scan followed by a `FILTER` and only the metadata of those blocks is copied, which remain after the `FILTER` (in which case the two operations are expensive anyway).
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
With this PR, filter expressions that can be evaluated via binary search on a sorted input are directly evaluated on the block metadata of an IndexScan. For example in a query that contains
{ ?s ?p ?o FILTER (?o > 3)} only the blocks of the full index scan (sorted by the object) are read from disk that according to their metadata might contain values> 3.Currently this mechanism has the following limitations: