fix: Optimize query plan in iter_filtered_chunks #939
Merged
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
Centril
commented
Mar 7, 2024
| spacetimedb_vm::dsl::query(schema.as_ref()).with_select(filter_to_column_op(&schema.table_name, filter)); | ||
| let query = spacetimedb_vm::dsl::query(schema.as_ref()) | ||
| .with_select(filter_to_column_op(&schema.table_name, filter)) | ||
| .optimize(&|table_id, table_name| stdb.row_count(table_id, table_name)); |
crates/sats/src/relation.rs
Outdated
Comment on lines
274
to
296
| .or_else(|| { | ||
| self.fields | ||
| .get(*field) | ||
| .filter(|c| c.field.table() == table) | ||
| .map(|_| *field) | ||
| }) |
There was a problem hiding this comment.
What could trigger the or_else scenario? It should not happened
There was a problem hiding this comment.
The implementation of this method hasn't actually changed other than checking against the table name (which is why one test case had to change). If you look at the old pub fn column(...) implementation below, it was just scanning self.fields for an equal FieldName, but this came up surprisingly empty. So the || *field == pos bit was needed to make it work. The reason it was required is that the header stores field names but the query stores positions.
2024-03-07T18:52:35.195880Z ERROR crates/vm/src/expr.rs:891: select_best_index, header
= Header { table_name: "LocationState", fields: [
Column { field: Name { table: "LocationState", field: "entity_id" }, algebraic_type: Builtin(U64), col_id: ColId(0) },
Column { field: Name { table: "LocationState", field: "chunk_index" }, algebraic_type: Builtin(U64), col_id: ColId(1) },
Column { field: Name { table: "LocationState", field: "x" }, algebraic_type: Builtin(I32), col_id: ColId(2) },
Column { field: Name { table: "LocationState", field: "z" }, algebraic_type: Builtin(I32), col_id: ColId(3) },
Column { field: Name { table: "LocationState", field: "dimension" }, algebraic_type: Builtin(U32), col_id: ColId(4) }],
constraints: [([ColId(0)], Constraints { attr: ColumnAttribute(INDEXED | UNIQUE | PRIMARY_KEY) }), ([ColId(1)], Constraints { attr: ColumnAttribute(INDEXED) }), ([ColId(2), ColId(3), ColId(4)], Constraints { attr: ColumnAttribute(INDEXED) })] }
2024-03-07T18:52:35.195907Z ERROR crates/vm/src/expr.rs:892: select_best_index, op
= Cmp { op: Logic(And), lhs: Cmp { op: Logic(And), lhs: Cmp { op: Cmp(Eq), lhs: Field(Name(Pos { table: "LocationState", field: 2 })), rhs: Field(Value(I32(80))) }, rhs: Cmp { op: Cmp(Eq), lhs: Field(Name(Pos { table: "LocationState", field: 3 })), rhs: Field(Value(I32(80))) } }, rhs: Cmp { op: Cmp(Eq), lhs: Field(Name(Pos { table: "LocationState", field: 4 })), rhs: Field(Value(U32(80))) } }
crates/vm/src/expr.rs
Outdated
Comment on lines
149
to
226
| /// Flattens a nested conjunction of AND expressions. | ||
| /// | ||
| /// For example, `a = 1 AND b = 2 AND c = 3` becomes `[a = 1, b = 2, c = 3]`. | ||
| /// | ||
| /// This helps with splitting the kinds of `queries`, | ||
| /// that *could* be answered by a `index`, | ||
| /// from the ones that need to be executed with a `scan`. | ||
| pub fn flatten_ands_ref(&self) -> ColumnOpRefFlat<'_> { | ||
| fn fill_vec<'a>(buf: &mut ColumnOpRefFlat<'a>, op: &'a ColumnOp) { | ||
| match op { | ||
| ColumnOp::Cmp { | ||
| op: OpQuery::Logic(OpLogic::And), | ||
| lhs, | ||
| rhs, | ||
| } => { | ||
| fill_vec(buf, lhs); | ||
| fill_vec(buf, rhs); | ||
| } | ||
| op => buf.push(op), | ||
| } | ||
| } | ||
| let mut buf = SmallVec::new(); | ||
| fill_vec(&mut buf, self); | ||
| buf | ||
| } |
There was a problem hiding this comment.
Inlined into extract_fields, thereby avoiding a potential allocation.
Centril
force-pushed
the
centril/fix-multi-col-index-query
branch
from
fb0be32
to
95808f1
Compare
mamcx
approved these changes
Mar 7, 2024
bfops
added
include-in-perf-test
test-with-bots
Centril
force-pushed
the
centril/fix-multi-col-index-query
branch
3 times, most recently
from
211be03
to
8e33bfb
Compare
bfops
added
release-any
jdetter
approved these changes
Mar 15, 2024
joshua-spacetime
force-pushed
the
centril/fix-multi-col-index-query
branch
from
8e33bfb
to
ea76fb8
Compare
joshua-spacetime
changed the title
query!(...) : Fix use of multi-col indices
joshua-spacetime
force-pushed
the
centril/fix-multi-col-index-query
branch
from
ea76fb8
to
c23dfcb
Compare
2 tasks
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.
Description of Changes
There were two bugs that prevented us from using multi-col indexes in
query!.None of these were in
find_sargable, rather, they were in:InstanceEnv::iter_filtered_chunks: it didn't callq.optimize(...)(so we may potentially see other improvements now).Header::columnwould not find theColumndue toFieldName::Posbeing passed.I've verified that the query ends up being optimized to a
Query::IndexScanand that this actually reaches the index with the right name inBTreeIndex::seek.API and ABI breaking changes
None
Expected complexity level and risk
2 - There are other users of
Header::column_pos_idx.