BE-428: HashQL: Simplify traversal tracking to path recording by indietyp · Pull Request #8498 · hashintel/hash

indietyp · 2026-03-02T20:59:32Z

🌟 What is the purpose of this PR?

Replaces the traversal extraction transform pass (which rewrote MIR bodies by splicing load statements and creating new locals) with a pure analysis pass that records which entity paths each block accesses as bitsets. Introduces per-block path transfer cost charging so the placement solver can account for data locality when assigning blocks to backends.

🔍 What does this change?

Traversal analysis (traversal/analysis/): walks the body, resolves vertex projections via EntityPath::resolve(), records accessed paths as TraversalPathBitSet. No MIR mutation.
Per-block cost analysis (cost/analysis.rs): new BasicBlockCostAnalysis combines per-statement costs with a path transfer premium. For each block, charges a transfer premium on targets that are not the natural origin for the accessed paths (e.g., Interpreter pays a premium for Vectors because Embedding is the origin). Premiums are charged once per block, not per statement.
BasicBlockCostVec: the placement solver now operates on precomputed per-block costs. PlacementSolverContext takes blocks: &BasicBlockCostVec instead of separate statement costs and assignment arrays.
BlockPartitionedVec<T>: generic offset-table-backed per-block storage, extracted from StatementCostVec internals.
TraversalLattice: lattice structure carrying VertexType for dataflow analysis over path bitsets.
EntityPath::origin(), EntityPath::estimate_size(): per-path origin backend and transfer size estimation.
TransferCostConfig: variable-size parameters (properties, embeddings, provenance sizes, per-target cost multiplier).
FiniteBitSet::negate() in hashql-core.
Removed pass/transform/traversal_extraction/ (the MIR transform and its tests).

Pre-Merge Checklist 🚀

🚢 Has this modified a publishable library?

This PR:

does not modify any publishable blocks or libraries, or modifications do not need publishing

📜 Does this require a change to the docs?

The changes in this PR:

are internal and do not require a docs change

🕸️ Does this require a change to the Turbo Graph?

The changes in this PR:

do not affect the execution graph

⚠️ Known issues

Path transfer premiums use pure over-approximation: every non-origin backend pays the full premium with no discounting for co-location. This can over-penalize non-origin assignment when multiple blocks access the same path on the same backend (cross-block dedup is not implemented). This is intentional: separable costs are a prerequisite for the PBQP solver planned in BE-436.

The ideal cost model is non-separable: the first block on a backend pays the full data fetch cost, subsequent co-labeled blocks pay zero ("shared-fetch" / pay-once-per-group). This maps to submodular clique costs in the optimization literature. Two frameworks formalize this:

Jegelka et al. 2014 (arXiv:1402.0240, Cooperative Graph Cuts): the "label cost" special case is mathematically identical to shared-fetch. Algorithms include polymatroidal flow (exact for separable surrogates) and semigradient methods. However, all are binary-only (k=2 labels).
Kolmogorov et al. 2010 (arXiv:1006.1990, Sum of Submodular Functions): general framework via submodular flow with capacity scaling. Also binary-only.

The multi-label extension via alpha-expansion (Boykov, Veksler, Zabih 2001) reduces k-label to a sequence of binary cooperative cuts (each solvable exactly), but only guarantees a local minimum. Our TransMatrix is not metric, so no global approximation factor holds. For k=3 backends (growing to 6-7), the overhead of iterative cooperative cut calls on ~100-node graphs is not justified when PBQP R1/R2 reductions on the separable over-approximation give exact solutions in microseconds.

Pure over-approximation does not distort the relative ranking between non-origin backends (the premium cancels in pairwise comparisons) but inflates the absolute cost of non-origin assignment, biasing toward origin backends. Intra-block dedup (implemented here) is the first mitigation; cross-block dedup remains an open problem for k > 2 on cyclic graphs.
:

🛡 What tests cover this?

Unit tests for BasicBlockCostAnalysis covering: no vertex access, single origin/non-origin paths, multiple path accumulation, composite path expansion, restricted target domains, cross-block independence
Unit tests for TraversalAnalysisVisitor covering path resolution from MIR projections
Unit tests for EntityPathBitSet covering composite swallowing, lattice operations, normalization
Existing solver tests updated to use BasicBlockCostVec; new path_premiums_influence_placement and provenance_variants_produce_different_premiums integration tests
End-to-end mixed_postgres_embedding_interpreter test verifying three-backend splitting

❓ How to test this?

cargo nextest run --package hashql-mir
cargo test --package hashql-mir --doc

vercel · 2026-03-02T20:59:37Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
hash	Ready	Preview, Comment	Mar 20, 2026 8:19am
petrinaut	Ready	Preview, Comment	Mar 20, 2026 8:19am

2 Skipped Deployments

Project	Deployment	Actions	Updated (UTC)
hashdotdesign	Ignored	Preview	Mar 20, 2026 8:19am
hashdotdesign-tokens	Ignored	Preview	Mar 20, 2026 8:19am

augmentcode · 2026-03-02T20:59:41Z

This pull request is too large for Augment to review. The PR exceeds the maximum size limit of 100000 tokens (approximately 400000 characters) for automated code review. Please consider breaking this PR into smaller, more focused changes.

libs/@local/hashql/mir/src/pass/execution/cost/mod.rs

vercel · 2026-03-31T20:56:58Z

Deployment failed with the following error:

Invalid request: `attribution.gitUser` should NOT have additional property `isBot`.

Latest approval commit 2346ece is not an ancestor of 5e31131, indicating rewritten history after approval

cursor

Cursor Bugbot has reviewed your changes and found 2 potential issues.

Bugbot Autofix prepared fixes for both issues found in the latest run.

✅ Fixed: Unbounded premiums become impossible placements
- I changed unbounded transfer midpoint fallback from ApproxCost::INF to finite Cost::MAX-equivalent and added a regression test to ensure unbounded path premiums stay finite.
✅ Fixed: insert_range trusts inconsistent domain size
- I added an explicit capacity assertion in FiniteBitSet::insert_range to reject oversized domain_size inputs and added a panic regression test.

Or push these changes by commenting:

@cursor push 361796ea03

Preview (361796ea03)

diff --git a/libs/@local/hashql/core/src/id/bit_vec/finite.rs b/libs/@local/hashql/core/src/id/bit_vec/finite.rs
--- a/libs/@local/hashql/core/src/id/bit_vec/finite.rs
+++ b/libs/@local/hashql/core/src/id/bit_vec/finite.rs
@@ -230,6 +230,8 @@
         I: [const] Id,
         T: [const] FiniteBitSetIntegral,
     {
+        assert!(domain_size <= T::MAX_DOMAIN_SIZE as usize);
+
         let Some((start, end)) = inclusive_start_end(bounds, domain_size) else {
             return;
         };
@@ -620,6 +622,13 @@
     }
 
     #[test]
+    #[should_panic(expected = "assertion failed")]
+    fn insert_range_domain_size_exceeds_capacity_panics() {
+        let mut set: FiniteBitSet<TestId, u8> = FiniteBitSet::new_empty(8);
+        set.insert_range(.., 9);
+    }
+
+    #[test]
     fn union_combines_bits() {
         let mut a: FiniteBitSet<TestId, u8> = FiniteBitSet::new_empty(8);
         a.insert(TestId::from_usize(0));

diff --git a/libs/@local/hashql/mir/src/pass/execution/cost/analysis.rs b/libs/@local/hashql/mir/src/pass/execution/cost/analysis.rs
--- a/libs/@local/hashql/mir/src/pass/execution/cost/analysis.rs
+++ b/libs/@local/hashql/mir/src/pass/execution/cost/analysis.rs
@@ -1,6 +1,6 @@
 use core::alloc::Allocator;
 
-use super::{ApproxCost, StatementCostVec};
+use super::{ApproxCost, Cost, StatementCostVec};
 use crate::{
     body::basic_block::{BasicBlock, BasicBlockId, BasicBlockSlice, BasicBlockVec},
     pass::{
@@ -124,7 +124,7 @@
         let load = range
             .saturating_mul(config.target_multiplier[target].get())
             .midpoint()
-            .map_or(ApproxCost::INF, From::from);
+            .map_or_else(|| ApproxCost::from(Cost::MAX), ApproxCost::from);
 
         BasicBlockTargetCost { base, load }
     }
@@ -511,6 +511,45 @@
         assert!(interpreter_cost > interpreter_base);
     }
 
+    /// Unbounded transfer-size estimates should stay expensive but finite, so they remain
+    /// valid candidates for solver search.
+    #[test]
+    fn unbounded_path_premium_stays_finite() {
+        let heap = Heap::new();
+        let interner = Interner::new(&heap);
+        let env = Environment::new(&heap);
+
+        let body = body!(interner, env; [graph::read::filter]@0/2 -> ? {
+            decl env: (), vertex: [Opaque sym::path::Entity; ?], val: ?;
+            @proj properties = vertex.properties: ?;
+
+            bb0() {
+                val = load properties;
+                return val;
+            }
+        });
+
+        let targets = make_targets(&body, all_targets());
+        let targets = BasicBlockSlice::from_raw(&targets);
+
+        let costs: TargetArray<StatementCostVec<Global>> =
+            TargetArray::from_fn(|_| StatementCostVec::from_iter([1].into_iter(), Global));
+
+        let analysis = BasicBlockCostAnalysis {
+            vertex: VertexType::Entity,
+            assignments: targets,
+            costs: &costs,
+        };
+
+        let result = analysis.analyze_in(&default_config(), &body.basic_blocks, Global);
+        let bb0 = BasicBlockId::new(0);
+
+        let interpreter_cost = result.cost(bb0, TargetId::Interpreter);
+        let interpreter_base = costs[TargetId::Interpreter].sum_approx(bb0);
+        assert!(interpreter_cost > interpreter_base);
+        assert!(interpreter_cost.is_finite());
+    }
+
     /// Paths across multiple blocks are analyzed independently per block.
     #[test]
     fn paths_across_blocks_independent() {

_{This Bugbot Autofix run was free. To enable autofix for future PRs, go to the Cursor dashboard.}

cursor · 2026-03-31T21:09:00Z

libs/@local/hashql/mir/src/pass/execution/mod.rs

+        }
+        .analyze_in(
+            &TransferCostConfig::new(InformationRange::full()),
+            &body.basic_blocks,


Unbounded premiums become impossible placements

High Severity

ExecutionAnalysis builds TransferCostConfig with InformationRange::full(), so unbounded paths like EntityPath::Properties produce None from midpoint() and are converted to ApproxCost::INF. Those infinite block costs can make valid cyclic assignments look infeasible during solver search, causing false unsatisfiable placement results.

Additional Locations (1)

libs/@local/hashql/mir/src/pass/execution/cost/analysis.rs#L123-L128

cursor · 2026-03-31T21:09:00Z

libs/@local/hashql/core/src/id/bit_vec/finite.rs

+        I: [const] Id,
+        T: [const] FiniteBitSetIntegral,
+    {
+        let Some((start, end)) = inclusive_start_end(bounds, domain_size) else {


insert_range trusts inconsistent domain size

Medium Severity

FiniteBitSet::insert_range now accepts an external domain_size but never validates it against T::MAX_DOMAIN_SIZE. If a larger domain is passed, end can exceed the backing width and the T::MAX_DOMAIN_SIZE - 1 - end math underflows, leading to panic or incorrect masking instead of a bounded range insert.

github-actions · 2026-03-31T22:20:34Z

Benchmark results

@rust/hash-graph-benches – Integrations

policy_resolution_large

Function	Value	Mean	Flame graphs
resolve_policies_for_actor	user: empty, selectivity: high, policies: 2002	$$27.3 \mathrm{ms} \pm 210 \mathrm{μs}\left({\color{gray}-3.658 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: empty, selectivity: low, policies: 1	$$3.41 \mathrm{ms} \pm 26.4 \mathrm{μs}\left({\color{gray}0.435 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: empty, selectivity: medium, policies: 1001	$$12.3 \mathrm{ms} \pm 100.0 \mathrm{μs}\left({\color{gray}-4.343 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: seeded, selectivity: high, policies: 3314	$$42.1 \mathrm{ms} \pm 298 \mathrm{μs}\left({\color{gray}-2.339 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: seeded, selectivity: low, policies: 1	$$13.8 \mathrm{ms} \pm 108 \mathrm{μs}\left({\color{gray}-3.676 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: seeded, selectivity: medium, policies: 1526	$$23.6 \mathrm{ms} \pm 198 \mathrm{μs}\left({\color{gray}-3.171 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: high, policies: 2078	$$28.2 \mathrm{ms} \pm 157 \mathrm{μs}\left({\color{gray}-1.527 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: low, policies: 1	$$3.69 \mathrm{ms} \pm 22.8 \mathrm{μs}\left({\color{gray}0.075 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: medium, policies: 1033	$$13.5 \mathrm{ms} \pm 108 \mathrm{μs}\left({\color{gray}-2.962 \mathrm{\%}}\right) $$	Flame Graph

policy_resolution_medium

Function	Value	Mean	Flame graphs
resolve_policies_for_actor	user: empty, selectivity: high, policies: 102	$$3.73 \mathrm{ms} \pm 17.6 \mathrm{μs}\left({\color{gray}-1.013 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: empty, selectivity: low, policies: 1	$$2.93 \mathrm{ms} \pm 16.3 \mathrm{μs}\left({\color{gray}-0.018 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: empty, selectivity: medium, policies: 51	$$3.28 \mathrm{ms} \pm 17.8 \mathrm{μs}\left({\color{gray}-1.253 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: seeded, selectivity: high, policies: 269	$$5.12 \mathrm{ms} \pm 29.3 \mathrm{μs}\left({\color{gray}-0.890 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: seeded, selectivity: low, policies: 1	$$3.48 \mathrm{ms} \pm 15.9 \mathrm{μs}\left({\color{gray}-0.590 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: seeded, selectivity: medium, policies: 107	$$4.07 \mathrm{ms} \pm 23.4 \mathrm{μs}\left({\color{gray}-0.966 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: high, policies: 133	$$4.42 \mathrm{ms} \pm 32.0 \mathrm{μs}\left({\color{gray}-1.117 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: low, policies: 1	$$3.33 \mathrm{ms} \pm 13.0 \mathrm{μs}\left({\color{gray}-1.248 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: medium, policies: 63	$$3.99 \mathrm{ms} \pm 17.6 \mathrm{μs}\left({\color{gray}-2.461 \mathrm{\%}}\right) $$	Flame Graph

policy_resolution_none

Function	Value	Mean	Flame graphs
resolve_policies_for_actor	user: empty, selectivity: high, policies: 2	$$2.65 \mathrm{ms} \pm 10.1 \mathrm{μs}\left({\color{gray}-1.441 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: empty, selectivity: low, policies: 1	$$2.62 \mathrm{ms} \pm 15.6 \mathrm{μs}\left({\color{gray}-0.861 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: empty, selectivity: medium, policies: 1	$$2.71 \mathrm{ms} \pm 13.3 \mathrm{μs}\left({\color{gray}-1.374 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: high, policies: 8	$$2.99 \mathrm{ms} \pm 15.5 \mathrm{μs}\left({\color{gray}-0.703 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: low, policies: 1	$$2.78 \mathrm{ms} \pm 12.7 \mathrm{μs}\left({\color{gray}-0.570 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: medium, policies: 3	$$3.07 \mathrm{ms} \pm 13.2 \mathrm{μs}\left({\color{gray}-2.216 \mathrm{\%}}\right) $$	Flame Graph

policy_resolution_small

Function	Value	Mean	Flame graphs
resolve_policies_for_actor	user: empty, selectivity: high, policies: 52	$$3.04 \mathrm{ms} \pm 13.0 \mathrm{μs}\left({\color{gray}-2.025 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: empty, selectivity: low, policies: 1	$$2.74 \mathrm{ms} \pm 14.1 \mathrm{μs}\left({\color{gray}-1.216 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: empty, selectivity: medium, policies: 25	$$2.91 \mathrm{ms} \pm 16.1 \mathrm{μs}\left({\color{gray}-0.537 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: seeded, selectivity: high, policies: 94	$$3.48 \mathrm{ms} \pm 18.6 \mathrm{μs}\left({\color{gray}-0.679 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: seeded, selectivity: low, policies: 1	$$2.99 \mathrm{ms} \pm 16.8 \mathrm{μs}\left({\color{gray}-1.947 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: seeded, selectivity: medium, policies: 26	$$3.24 \mathrm{ms} \pm 18.9 \mathrm{μs}\left({\color{gray}-1.706 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: high, policies: 66	$$3.37 \mathrm{ms} \pm 15.5 \mathrm{μs}\left({\color{gray}-0.992 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: low, policies: 1	$$2.99 \mathrm{ms} \pm 14.9 \mathrm{μs}\left({\color{gray}-2.861 \mathrm{\%}}\right) $$	Flame Graph
resolve_policies_for_actor	user: system, selectivity: medium, policies: 29	$$3.32 \mathrm{ms} \pm 23.9 \mathrm{μs}\left({\color{gray}0.419 \mathrm{\%}}\right) $$	Flame Graph

read_scaling_complete

Function	Value	Mean	Flame graphs
entity_by_id;one_depth	1 entities	$$44.2 \mathrm{ms} \pm 181 \mathrm{μs}\left({\color{gray}-0.381 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;one_depth	10 entities	$$81.9 \mathrm{ms} \pm 457 \mathrm{μs}\left({\color{gray}-1.087 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;one_depth	25 entities	$$48.1 \mathrm{ms} \pm 214 \mathrm{μs}\left({\color{gray}-1.649 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;one_depth	5 entities	$$51.2 \mathrm{ms} \pm 335 \mathrm{μs}\left({\color{gray}-1.744 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;one_depth	50 entities	$$60.9 \mathrm{ms} \pm 377 \mathrm{μs}\left({\color{gray}0.254 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;two_depth	1 entities	$$45.9 \mathrm{ms} \pm 180 \mathrm{μs}\left({\color{gray}0.284 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;two_depth	10 entities	$$420 \mathrm{ms} \pm 970 \mathrm{μs}\left({\color{gray}-1.728 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;two_depth	25 entities	$$99.4 \mathrm{ms} \pm 535 \mathrm{μs}\left({\color{gray}-2.960 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;two_depth	5 entities	$$89.8 \mathrm{ms} \pm 360 \mathrm{μs}\left({\color{gray}-1.009 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;two_depth	50 entities	$$288 \mathrm{ms} \pm 892 \mathrm{μs}\left({\color{gray}-1.272 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;zero_depth	1 entities	$$19.0 \mathrm{ms} \pm 106 \mathrm{μs}\left({\color{gray}-3.149 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;zero_depth	10 entities	$$19.5 \mathrm{ms} \pm 105 \mathrm{μs}\left({\color{gray}-2.216 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;zero_depth	25 entities	$$19.8 \mathrm{ms} \pm 105 \mathrm{μs}\left({\color{gray}-2.587 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;zero_depth	5 entities	$$19.9 \mathrm{ms} \pm 147 \mathrm{μs}\left({\color{gray}-0.546 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id;zero_depth	50 entities	$$23.5 \mathrm{ms} \pm 145 \mathrm{μs}\left({\color{gray}-3.358 \mathrm{\%}}\right) $$	Flame Graph

read_scaling_linkless

Function	Value	Mean	Flame graphs
entity_by_id	1 entities	$$19.2 \mathrm{ms} \pm 116 \mathrm{μs}\left({\color{gray}-2.919 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	10 entities	$$19.1 \mathrm{ms} \pm 94.5 \mathrm{μs}\left({\color{gray}-1.202 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	100 entities	$$19.0 \mathrm{ms} \pm 100 \mathrm{μs}\left({\color{gray}-2.950 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	1000 entities	$$19.5 \mathrm{ms} \pm 116 \mathrm{μs}\left({\color{gray}-1.761 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	10000 entities	$$25.4 \mathrm{ms} \pm 157 \mathrm{μs}\left({\color{gray}-3.563 \mathrm{\%}}\right) $$	Flame Graph

representative_read_entity

Function	Value	Mean	Flame graphs
entity_by_id	entity type ID: `https://blockprotocol.org/@alice/types/entity-type/block/v/1`	$$34.1 \mathrm{ms} \pm 291 \mathrm{μs}\left({\color{gray}-2.342 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	entity type ID: `https://blockprotocol.org/@alice/types/entity-type/book/v/1`	$$33.4 \mathrm{ms} \pm 316 \mathrm{μs}\left({\color{gray}-4.870 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	entity type ID: `https://blockprotocol.org/@alice/types/entity-type/building/v/1`	$$35.0 \mathrm{ms} \pm 321 \mathrm{μs}\left({\color{gray}2.88 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	entity type ID: `https://blockprotocol.org/@alice/types/entity-type/organization/v/1`	$$35.3 \mathrm{ms} \pm 283 \mathrm{μs}\left({\color{gray}3.97 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	entity type ID: `https://blockprotocol.org/@alice/types/entity-type/page/v/2`	$$33.7 \mathrm{ms} \pm 288 \mathrm{μs}\left({\color{gray}-2.269 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	entity type ID: `https://blockprotocol.org/@alice/types/entity-type/person/v/1`	$$33.7 \mathrm{ms} \pm 297 \mathrm{μs}\left({\color{gray}-1.283 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	entity type ID: `https://blockprotocol.org/@alice/types/entity-type/playlist/v/1`	$$34.2 \mathrm{ms} \pm 300 \mathrm{μs}\left({\color{gray}3.55 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	entity type ID: `https://blockprotocol.org/@alice/types/entity-type/song/v/1`	$$35.4 \mathrm{ms} \pm 312 \mathrm{μs}\left({\color{red}6.44 \mathrm{\%}}\right) $$	Flame Graph
entity_by_id	entity type ID: `https://blockprotocol.org/@alice/types/entity-type/uk-address/v/1`	$$36.0 \mathrm{ms} \pm 305 \mathrm{μs}\left({\color{red}7.20 \mathrm{\%}}\right) $$	Flame Graph

representative_read_entity_type

Function	Value	Mean	Flame graphs
get_entity_type_by_id	Account ID: `bf5a9ef5-dc3b-43cf-a291-6210c0321eba`	$$8.27 \mathrm{ms} \pm 39.8 \mathrm{μs}\left({\color{gray}0.177 \mathrm{\%}}\right) $$	Flame Graph

representative_read_multiple_entities

Function	Value	Mean	Flame graphs
entity_by_property	traversal_paths=0	0	$$94.1 \mathrm{ms} \pm 666 \mathrm{μs}\left({\color{gray}0.051 \mathrm{\%}}\right) $$
entity_by_property	traversal_paths=255	1,resolve_depths=inherit:1;values:255;properties:255;links:127;link_dests:126;type:true	$$144 \mathrm{ms} \pm 664 \mathrm{μs}\left({\color{gray}-2.305 \mathrm{\%}}\right) $$
entity_by_property	traversal_paths=2	1,resolve_depths=inherit:0;values:0;properties:0;links:0;link_dests:0;type:false	$$102 \mathrm{ms} \pm 694 \mathrm{μs}\left({\color{gray}0.280 \mathrm{\%}}\right) $$
entity_by_property	traversal_paths=2	1,resolve_depths=inherit:0;values:0;properties:0;links:1;link_dests:0;type:true	$$110 \mathrm{ms} \pm 689 \mathrm{μs}\left({\color{gray}-0.302 \mathrm{\%}}\right) $$
entity_by_property	traversal_paths=2	1,resolve_depths=inherit:0;values:0;properties:2;links:1;link_dests:0;type:true	$$119 \mathrm{ms} \pm 721 \mathrm{μs}\left({\color{gray}0.035 \mathrm{\%}}\right) $$
entity_by_property	traversal_paths=2	1,resolve_depths=inherit:0;values:2;properties:2;links:1;link_dests:0;type:true	$$124 \mathrm{ms} \pm 669 \mathrm{μs}\left({\color{gray}-0.749 \mathrm{\%}}\right) $$
link_by_source_by_property	traversal_paths=0	0	$$100 \mathrm{ms} \pm 555 \mathrm{μs}\left({\color{gray}-2.161 \mathrm{\%}}\right) $$
link_by_source_by_property	traversal_paths=255	1,resolve_depths=inherit:1;values:255;properties:255;links:127;link_dests:126;type:true	$$129 \mathrm{ms} \pm 712 \mathrm{μs}\left({\color{gray}0.709 \mathrm{\%}}\right) $$
link_by_source_by_property	traversal_paths=2	1,resolve_depths=inherit:0;values:0;properties:0;links:0;link_dests:0;type:false	$$108 \mathrm{ms} \pm 683 \mathrm{μs}\left({\color{gray}-0.877 \mathrm{\%}}\right) $$
link_by_source_by_property	traversal_paths=2	1,resolve_depths=inherit:0;values:0;properties:0;links:1;link_dests:0;type:true	$$116 \mathrm{ms} \pm 497 \mathrm{μs}\left({\color{gray}0.001 \mathrm{\%}}\right) $$
link_by_source_by_property	traversal_paths=2	1,resolve_depths=inherit:0;values:0;properties:2;links:1;link_dests:0;type:true	$$118 \mathrm{ms} \pm 583 \mathrm{μs}\left({\color{gray}-0.216 \mathrm{\%}}\right) $$
link_by_source_by_property	traversal_paths=2	1,resolve_depths=inherit:0;values:2;properties:2;links:1;link_dests:0;type:true	$$118 \mathrm{ms} \pm 551 \mathrm{μs}\left({\color{gray}-0.495 \mathrm{\%}}\right) $$

scenarios

Function	Value	Mean	Flame graphs
full_test	query-limited	$$130 \mathrm{ms} \pm 444 \mathrm{μs}\left({\color{lightgreen}-7.644 \mathrm{\%}}\right) $$	Flame Graph
full_test	query-unlimited	$$143 \mathrm{ms} \pm 583 \mathrm{μs}\left({\color{lightgreen}-5.305 \mathrm{\%}}\right) $$	Flame Graph
linked_queries	query-limited	$$104 \mathrm{ms} \pm 592 \mathrm{μs}\left({\color{gray}-1.019 \mathrm{\%}}\right) $$	Flame Graph
linked_queries	query-unlimited	$$586 \mathrm{ms} \pm 3.09 \mathrm{ms}\left({\color{gray}4.16 \mathrm{\%}}\right) $$	Flame Graph

indietyp added 26 commits February 28, 2026 18:48

feat: move to shared infrastructure

3e7176e

feat: traversal path

d8a0ec0

feat: move analysis traversal pass (WIP)

2208fe7

feat: checkpoint

e03fcf6

feat: checkpoint

18491b4

feat: checkpoint

db7c41f

feat: checkpoint

4d69417

feat: rework tests

1566891

feat: define lattice over the entity

e2eb7b5

feat: checkpoint

db39aa6

feat: define lattice over the entity

3e2e118

feat: terminator placement

d4960b4

feat: terminator placement

fb3a6f7

chore: tests

c53de7e

feat: re-enable unused lints

4acdb2c

feat: checkpoint

0bb42a6

fix: remove the old traversals intermediary

8121899

fix: remove the old traversals intermediary

7b3f458

chore: remove old version

e7c0282

feat: sketch out the cost vec

f1a00c0

feat: checkpoitn

3502d44

feat: checkpoint

8de1462

feat: checkpoint

d93c19a

feat: checkpoint

abbe3ab

fix: doc cleanup

79e0f09

fix: suggestions from code review

7dd5915

indietyp mentioned this pull request Mar 2, 2026

BE-434: HashQL: Refactor entity storage path resolution with centralized schema mapping #8496

Merged

3 tasks

github-actions bot added the area/libs Relates to first-party libraries/crates/packages (area) label Mar 2, 2026

vercel bot deployed to Preview – hash March 2, 2026 21:31 View deployment

graphite-app bot requested review from a team March 2, 2026 21:39

fix: ordering

7433fee

vercel bot temporarily deployed to Preview – petrinaut March 3, 2026 09:55 Inactive

indietyp force-pushed the bm/be-434-hashql-entity-path-resolution-and-storage-mapping branch from 3d9cc9d to 885a006 Compare March 8, 2026 17:03

indietyp force-pushed the bm/be-428-hashql-simplify-traversal-tracking-to-path-recording branch from 7014842 to e6439e0 Compare March 8, 2026 17:03

vercel bot deployed to Preview – petrinaut March 8, 2026 17:06 View deployment

cursor bot reviewed Mar 8, 2026

View reviewed changes

libs/@local/hashql/mir/src/pass/execution/cost/mod.rs Show resolved Hide resolved

vercel bot deployed to Preview – hash March 8, 2026 17:14 View deployment

Base automatically changed from bm/be-434-hashql-entity-path-resolution-and-storage-mapping to main March 10, 2026 20:41

indietyp force-pushed the bm/be-428-hashql-simplify-traversal-tracking-to-path-recording branch from e6439e0 to 2346ece Compare March 20, 2026 08:11

vercel bot deployed to Preview – petrinaut March 20, 2026 08:14 View deployment

vercel bot deployed to Preview – hash March 20, 2026 08:19 View deployment

This was referenced Mar 26, 2026

BE-455: HashQL: Introduce evaluation orchestrator #8586

Draft

BE-474: HashQL: Take into account terminator target eligibility in execution backend placement #8587

Draft

indietyp requested a review from TimDiekmann March 26, 2026 09:35

TimDiekmann previously approved these changes Mar 26, 2026

View reviewed changes

This was referenced Mar 28, 2026

BE-482: HashQL: Remove logical not from MIR and fix postgres boolean lowering #8595

Draft

BE-494: HashQL: Loop-breaker selection for recursive inlining #8600

Draft

fix: library features

5e31131

indietyp force-pushed the bm/be-428-hashql-simplify-traversal-tracking-to-path-recording branch from 2346ece to 5e31131 Compare March 31, 2026 20:56

cursor bot reviewed Mar 31, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

BE-428: HashQL: Simplify traversal tracking to path recording#8498

BE-428: HashQL: Simplify traversal tracking to path recording#8498
indietyp wants to merge 29 commits intomainfrom
bm/be-428-hashql-simplify-traversal-tracking-to-path-recording

indietyp commented Mar 2, 2026 •

edited

Loading

Uh oh!

vercel bot commented Mar 2, 2026 •

edited

Loading

Uh oh!

augmentcode bot commented Mar 2, 2026

Uh oh!

Uh oh!

vercel bot commented Mar 31, 2026

Uh oh!

cursor bot left a comment •

edited

Loading

Uh oh!

cursor bot Mar 31, 2026

Uh oh!

cursor bot Mar 31, 2026

Uh oh!

github-actions bot commented Mar 31, 2026

@rust/hash-graph-benches – Integrations

policy_resolution_large

policy_resolution_medium

policy_resolution_none

policy_resolution_small

read_scaling_complete

read_scaling_linkless

representative_read_entity

representative_read_entity_type

representative_read_multiple_entities

scenarios

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

Conversation

indietyp commented Mar 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🌟 What is the purpose of this PR?

🔍 What does this change?

Pre-Merge Checklist 🚀

🚢 Has this modified a publishable library?

📜 Does this require a change to the docs?

🕸️ Does this require a change to the Turbo Graph?

⚠️ Known issues

🛡 What tests cover this?

❓ How to test this?

Uh oh!

vercel bot commented Mar 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

augmentcode bot commented Mar 2, 2026

Uh oh!

Uh oh!

vercel bot commented Mar 31, 2026

Uh oh!

cursor bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cursor bot Mar 31, 2026

Choose a reason for hiding this comment

Unbounded premiums become impossible placements

Uh oh!

cursor bot Mar 31, 2026

Choose a reason for hiding this comment

insert_range trusts inconsistent domain size

Uh oh!

github-actions bot commented Mar 31, 2026

Benchmark results

@rust/hash-graph-benches – Integrations

policy_resolution_large

policy_resolution_medium

policy_resolution_none

policy_resolution_small

read_scaling_complete

read_scaling_linkless

representative_read_entity

representative_read_entity_type

representative_read_multiple_entities

scenarios

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

indietyp commented Mar 2, 2026 •

edited

Loading

vercel bot commented Mar 2, 2026 •

edited

Loading

cursor bot left a comment •

edited

Loading