russell@unturf.com
db29a08762
B&W print-friendly diagrams + tinkerpop-0001 + wave-3 proof sections. Squash of 94 local commits onto remote master.
53 lines
2.4 KiB
Markdown
53 lines
2.4 KiB
Markdown
---
|
||
id: postgresql-0004
|
||
repo: postgres/postgres
|
||
severity: HIGH
|
||
status: DEFERRED
|
||
created: 2026-03-23
|
||
---
|
||
|
||
## Defect
|
||
|
||
**File:** `src/backend/optimizer/util/analyzejoins.c:1914`
|
||
**Pattern:** `list_member(toKeep->reltarget->exprs, node)` inside `foreach(lc, toRemove->reltarget->exprs)`
|
||
**Complexity:** O(|toRemove.exprs|² × |toKeep.exprs|) per join elimination pass
|
||
**Language:** C
|
||
|
||
## Description
|
||
|
||
`analyzejoins.c` implements join elimination — removing redundant joins from the query
|
||
plan when the optimizer can prove the join does not affect the result set. During join
|
||
relation merging, line 1914 calls `list_member()` inside `foreach(lc, toRemove->reltarget->exprs)`
|
||
to check whether each expression from the removed relation already exists in the kept
|
||
relation's target list.
|
||
|
||
Join elimination runs on every query that has self-joins, redundant outer joins, or
|
||
joins to views. For a query that eliminates a join between relations with E expressions
|
||
each, this is O(E²). Since `reltarget->exprs` contains all expressions needed from that
|
||
relation (not just the join key), complex queries with many projected columns trigger
|
||
quadratic behavior.
|
||
|
||
## Fix
|
||
|
||
**Replace:** `list_member(toKeep->reltarget->exprs, node)` — O(n) per check
|
||
**With:** Pre-build a pointer-identity hash set over `toKeep->reltarget->exprs` before
|
||
the foreach loop; check O(1) per expression.
|
||
**Data structure:** Simple array-backed hash set over `Node *` pointers; palloc'd for
|
||
the duration of the join elimination pass.
|
||
|
||
## Work required
|
||
|
||
- [ ] Patch in `defects/postgresql/patch/`
|
||
- [ ] Unit test — asserts exact operation counts before/after (in `defects/postgresql/unit/`)
|
||
- [ ] Integration test — query with self-join on table with C=5,10,20,50 projected columns (in `defects/postgresql/integration/`)
|
||
- [ ] Benchmark — EXPLAIN ANALYZE planning time, join elimination enabled vs disabled (in `defects/postgresql/bench/`)
|
||
- [ ] White paper section — `whitepaper/vectors/database/postgresql.rst`
|
||
|
||
## DEFERRED (2026-03-23)
|
||
|
||
PostgreSQL uses structural deep equality (`equal()`) for expression comparison.
|
||
No generic expression hash function exists in PostgreSQL core. A correct fix
|
||
requires either: (a) a structural hash based on Node type+fields, or (b) sort+merge
|
||
using a total order on Expr* — neither is available without significant framework
|
||
additions. PostgreSQL developers explicitly note the O(n²) cost in comments and
|
||
say "really you should be using some other data structure."
|