v0.2.0

[0.2.0] - 2026-09-14

🚀 Features

• (kermit-bench) Move tuple generators into generation module
• (kermit-bench) Add graph generation stub with GraphModel enum
• (kermit-bench) Add exponential and factorial benchmark configs
• (kermit) Add SpaceMeasurement module for Criterion space benchmarks
• (kermit) Add bench suite subcommand for running named benchmark workloads
• (bench) Replace synthetic benchmarks with cloud-based YAML system
• (bench) Support multiple named queries per benchmark
• (bench) Add list subcommand to show available benchmarks
• (bench) Add Oxford .tbl to Parquet conversion script
• (bench) Add Oxford DSI benchmark definitions
• (algos) Add SingletonTrieIter for Const-view rewrite
• (algos) Add TrieIterKind dispatch enum
• (algos) Add const_rewrite for paper-canonical atom filtering
• (kermit) Wire Const-view rewrite into DatabaseEngine::join
• (watdiv) Preprocessor dictionary builder
• (watdiv) Preprocessor predicate partitioner
• (watdiv) Preprocessor SPARQL → Datalog translator
• (watdiv) Preprocessor YAML emitter and end-to-end driver
• (watdiv) Preprocessor emits expected.json from .desc files
• (watdiv) Add generic-workload driver (sparql-preprocess CLI)
• (watdiv) Add stress-100/1000 benchmark YAMLs
• (kermit) Wire SpaceMeasurement through bench ds + bench run
• (kermit) Unified bench metadata writer + BytesFormatter integration
• (kermit) Add --report-json channel for bench subcommands
• (kermit) Tighten bench output and CLI surface
• (kermit-rdf) Scaffold crate with empty stub modules
• (kermit-rdf) Add RdfError enum
• (kermit-rdf) Add RdfValue enum
• (kermit-rdf) Add streaming N-Triples parser
• (kermit-rdf) Add Dictionary type
• (kermit-rdf) Add predicate sanitization
• (kermit-rdf) Add triple partitioning by predicate
• (kermit-rdf) Add Parquet writers for dict + relations
• (kermit-rdf) Add SPARQL parser wrapper
• (kermit-rdf) SPARQL→Datalog translator with golden tests
• (kermit-rdf) Emit BenchmarkDefinition YAML
• (kermit-rdf) Harvest watdiv .desc cardinalities into expected/ CSVs
• (kermit-rdf) Add TempStagingDir for watdiv driver
• (kermit-rdf) Driver invokes watdiv -d/-s/-q
• (kermit-rdf) Driver entry point and DriverInputs
• (kermit-rdf) Pipeline orchestrator and meta.json schema
• (kermit-rdf) Vendor watdiv binary + support files
• (kermit-bench) Discovery walks workspace + cache roots
• (kermit) Add bench watdiv-gen subcommand
• (kermit) List/fetch/run walk workspace + cache roots
• (kermit) Drop criterion auto-plot rendering
• (kermit) Bump bench-report schema to v2 with axes map
• (kermit-plot) Add python plotter for bench reports
• (kermit) Always write json bench report
• (kermit-lab) Add DataFrame loading API
• (kermit-lab) Plot functions return Figure; CLI is a thin wrapper
• (kermit-lab) Add summary, compare, and stats functions
• (kermit-rdf) Add lubm benchmark generation pipeline
• (kermit) Add bench lubm-gen subcommand
• (kermit) Add declarative generator yaml for bench run
• (kermit) Sweep --indexstructure and --algorithm via 'all' selectors
• (kermit-lab) Scaffold full-timeline notebook with prereq cells
• (kermit-lab) Notebook phase 1 — bench gen lubm
• (kermit-lab) Notebook phases 2-3 — bench run time and space
• (kermit-lab) Notebook phase 4 — load reports into DataFrame
• (kermit-lab) Notebook phases 5-6 — plots + outro
• (kermit-lab) Bar_queries accepts multi ds/algo with stacked layout
• (iters) Add hash_attribute for hash-trie iterator family
• (iters) Add HashTrieIterator trait
• (iters) Add HashTrieIterable trait
• (iters) Re-export HashTrieIterator family from crate root
• (ds) Scaffold HashTable primitive
• (ds) HashTable::get/get_mut with linear probing
• (ds) HashTable::entry_or_insert_with (pre-resize)
• (ds) HashTable resize on load factor > 0.7
• (ds) HashTable::iter over occupied buckets
• (ds) HashTrieNode recursive enum
• (ds) HashTrie struct + Relation::new
• (ds) HashTrie::insert via recursive insert_at
• (ds) HashTrie::from_tuples and insert_all
• (ds) HashTrie::collect_tuples for projection and tests
• (ds) HashTrie HeapSize impl
• (ds) HashTrie Projectable impl
• (ds) HashTrieIter scaffolding with open
• (ds) HashTrieIter::{key,next,at_end,size}
• (ds) HashTrieIter::lookup with HashTable::index_of
• (ds) HashTrieIter::up
• (ds) HashTrieIter::leaf_tuples
• (ds) HashTrie HashTrieIterable impl
• (ds) Expose HashTrie via IndexStructure and crate root
• (algos) SingletonHashTrieIter skeleton
• (algos) SingletonHashTrieIter::leaf_tuples
• (algos) HashTrieIterKind dispatch enum
• (algos) Hash_triejoin query indexing helpers
• (algos) Verify_and_construct for leaf-level emission
• (algos) Enumerate + emit_leaf — Algorithm 3 verbatim
• (algos) HashTriejoin entry point
• (cli) Add HashTrie/HashTriejoin to selectors
• (cli) IndexStructureSelector::supports_algorithm
• (cli) Hash_join free function for HashTrie family
• (cli) Wire HashTrie + HashTriejoin into bench dispatch
• (iters) Add optimization trait family
• (iters) Add HashStrategy trait + Sip/FxHash strategies
• (ds) Implement HasOptimizationAxes on HashTrie
• (cli) Add HasherChoice and LayoutChoices CLI scaffolding
• (cli) Bench-report axes include ds_layout_hasher
• (rdf) Vendor WatDiv Basic Testing query templates (L/S/F/C)
• (bench) Add watdiv-basic GeneratorSpec variant
• (rdf) Add drive_basic for the WatDiv Basic Testing workload
• (rdf) Add run_basic_pipeline; parameterize process_artifacts meta kind
• (cli) Dispatch watdiv-basic generator to run_basic_pipeline
• (bench) Add watdiv-basic and watdiv-stress-default benchmarks
• (kermit-lab) Add optimization-axis defaults registry
• (kermit-lab) Wildcard-wide optimization-axis ingest in loader
• (kermit-lab) Column-aware colour/marker/linestyle + HashTrie colour
• (kermit-lab) Aesthetic-map encoding + scalar series resolution
• (kermit-lab) Bar + line geom drawers
• (kermit-lab) Facet grid + shared-legend helper
• (kermit-lab) Kl.plot entry point (bar + line)
• (kermit-lab) Scatter + violin geoms and tradeoff metric-x path
• (kermit-lab) Presets over the general plot engine
• (kermit-lab) Rewire public API to engine + presets
• (kermit-lab) Preset-based render-all with ablation auto-discovery
• (kermit-lab) General plot subcommand + preset routing in CLI
• (bench) Add lubm-reference declarative benchmark
• (kermit-lab) Reference notebook — preflight and profile knob
• (kermit-lab) Reference notebook — correctness phase
• (kermit-lab) Reference notebook — benchmark sweep phase
• (kermit-lab) Reference notebook — load phase
• (kermit-lab) Reference notebook — comparison figures
• (kermit-lab) Reference notebook — statistical analysis
• (kermit-lab) Reference notebook — summary and outro
• (kermit-ds) Cardinality trait with O(1) TreeTrie tuple counter
• (kermit-ds) ColumnTrie Cardinality impl via matched-existing insert signal
• (kermit-ds) HashTrie Cardinality impl (multiset stored-tuple count)
• (kermit-algos) Optimiser module with shared query analysis (hoisted from build_variable_index)
• (kermit-algos) QueryPlan with validation and rank-parameterised topological ordering
• (kermit-algos) QueryOptimiser trait, CatalogStats, LexicographicOptimiser default
• (kermit-algos) CardinalityOptimiser — smallest-relation-first variable ordering
• [breaking] JoinAlgo::join_iter consumes a QueryPlan; engine plans via QueryOptimiser
• (cli) --optimiser flag on join/bench join/bench run with 'optimiser' report axis
• (bench) Add end-to-end metric timing build + K queries per sample
• (bench) Let a benchmark commit its relation file via path:
• (ds) Add HashTrieConfig with the singleton_pruning flag
• (ds) Add ConfigurableRelation and carry HashTrieConfig on HashTrie
• (ds) Singleton pruning at HashTrie build time (paper Fig. 5)
• (ds) Emulate pruned levels in HashTrieIter with singleton frames
• (ds) Add Configured<R, P> to lift a Config value to a type for test suites
• (kermit-lab) Back-fill ds_config_singleton_pruning=false for pre-Config reports
• (kermit) Carry HashTrieConfig on the HashHtj execution cell
• (kermit) Add --ds-config with singleton-pruning for hash-trie benches
• (ds) Add the PruningPolicy Layout markers with an uninhabited off payload
• (ds) Make the HashTable load-factor cap a HashTrieConfig value (ds_config_load_factor)
• (kermit-lab) Ds_layout_pruning and ds_config_load_factor defaults; fixtures name real axes
• (kermit) Accept --ds-config load-factor and report ds_config_load_factor
• (ds) [breaking] Singleton pruning becomes the P Layout parameter of HashTrie (ds_layout_pruning)
• (kermit) [breaking] --ds-layout-pruning and the hasher × pruning Layout dispatch
• (kermit-ds) Add IndexStructure::axis_value as the pinned identity label
• (kermit-algos) Add JoinAlgorithm::axis_value as the pinned identity label
• (kermit) Resolve the workspace root at runtime with a KERMIT_WORKSPACE override
• (kermit) Add ReportSink for atomic per-cell report rewrites
• (kermit) Write bench reports incrementally through ReportSink
• (kermit) Add bench::Workload, the runner's input for bench run and bench join
• (kermit) Run bench join through the generic runner with metrics, axes and --ds-config
• (kermit-bench) Optional expected result cardinality on query definitions
• (kermit-rdf) Write expected cardinalities into benchmark.yml and drop the sidecars
• (kermit) Bench run --verify checks result cardinalities before timing
• (kermit-lab) Verified column and a real-binary contract test
• (kermit-bench) Optional sha256 on relations, checked on download and by verify_integrity
• (kermit) Bench fetch verifies declared sha256 digests; pin triangle's relation
• (scripts) Add WatDiv stress snapshot installer and per-template sampler
• (algos) Selection rewrite and equality-selection views for repeated variables

🐛 Bug Fixes

• Address code review findings
• (bench) Reject HTTP errors on download and use distinct placeholder URLs
• (watdiv) Thread URI→name map through preprocessor to prevent predicate aliasing
• (watdiv) Make emitter fail loud on unknown predicates; document head-atom scope
• (watdiv) Include literal-object triples in predicate partitions
• (watdiv) Absorb query-only URIs into dictionary instead of raising
• (watdiv-gen) Apply code review feedback for on-the-fly rollout
• (kermit-rdf) Align watdiv -s/-q driver with vendored binary CLI
• (kermit-rdf) Canonicalize watdiv binary path before staging
• (kermit-plot) Handle slashed group names and null slope from criterion
• (kermit-lab) Accept glob patterns in load(); preserve CLI exit codes
• (kermit-lab) Make quick-start notebook robust to single-scale data
• (kermit-lab) Document data-shape requirements and unhardcode DS names in notebooks
• (kermit-lab) Correct --report-json flag position in watdiv notebook
• (kermit-lab) Correct watdiv benchmark names in 04 notebook
• (kermit) Address code review issues for declarative generator yaml
• (kermit-lab) Stream run_phase output and align launch command
• (kermit-ds) Update interval_index in duplicate-key descent path
• (iters) Seek works on unstarted iterator
• (rdf) Inverse rule skips blank-node objects
• (ds) Guard ColumnTrieIter::open against post-exhaustion descent
• (ds) Assert header arity in from_tuples
• (kermit) Report missing relation name in DB diagnostics
• (kermit) Propagate cache-dir errors and dedupe cache lookups
• (rdf) Tighten subprocess and path error handling
• (algos) Enforce valid global attribute order in triejoins
• (rdf) Seed empty relations for predicates absent from basic-workload data
• (kermit-lab) Deterministic stable colour fallback, preserve committed-colour separation
• (kermit-lab) Dedup group cols, neutral fallback colour, error-path tests for encoding
• (kermit-lab) Generalize tradeoff orientation + keys, strengthen plot tests, drop redundant path comments
• (kermit-lab) Clean-break API rewire — revert preset compat shims, drop obsolete top-level-API tests
• (kermit-lab) Reference notebook — per-config sweeps, F5/T2 labels, citation fixes
• (kermit-lab) Reference notebook — align intro sweep phrasing with per-config invocations
• (kermit) Make bench run sweeps enumerate valid cells, unify runners
• (kermit) Build the bench run insertion metric with the family's config
• (ds) Plain doc comment on define_config_provider! marker so rustfmt is idempotent
• (kermit-algos) Make a failed LFTJ descent atomic
• (kermit) Resolve the workspace root once per process
• (rdf) Seed empty relations for absent predicates in the stress workload too
• (kermit-rdf) Describe RdfError::Expected by all the stages that raise it
• (kermit-lab) Contract test runs in a temp dir and surfaces kermit's output
• (db) Enforce repeated variables via selection views; suite runs through real entry points
• (bench) Reject downloads that are not Parquet files before caching
• (bench,lab) Resolve Criterion results by id, refuse truncating sweeps

💼 Other

• Add algorithm and data-structure docs
• (cli) Optimiser flag hardening — axis guard test, tempfile hygiene, naming

🚜 Refactor

• (kermit-bench) Remove download infrastructure, rewrite config types
• (kermit-ds) Remove Criterion benchmarks
• (kermit) Remove Benchmarker stub and Commands::Benchmark
• (watdiv) Extract run_pipeline from main
• (watdiv) Decouple expected.py from watdiv-stress-* glob
• (watdiv) Move WatDiv CLI into drivers/watdiv
• (kermit-lab) Hoist PATHS and CRITERION_ROOT to constants in notebooks
• (kermit) Nest gen subcommands under bench gen
• (kermit-algos) Clarify LFTJ naming and state machine
• (kermit-ds) Tighten Relation trait and clarify trie internals
• (kermit-rdf) Extract shared timestamp helpers
• (kermit-rdf) Name parquet column strings as constants
• (kermit) Name default group and byte-unit constants
• (kermit) Drop unused PhantomData in DatabaseEngine
• (kermit-ds) Rename TreeTrieIter pos to sibling_idx
• (kermit-algos) Rename LFTJ iterator_pool to idle_iterators
• (kermit) Replace anyhow rewrap boilerplate with bare ?
• (kermit) Introduce CacheState enum in materialize
• (kermit-bench) Extract try_load_cache_subdir helper
• (kermit-rdf) Split LUBM entailment loop into per-rule fns
• (kermit-ds) Extract LayerStep enum from internal_insert
• (kermit) Extract add_space_bench helper for space metrics
• (ds) Extract project_via_trie_iter helper
• (rdf) Lift sha256_file helper to crate root
• (bench) Expose base_cache_dir and load_cached_benchmark
• (derive) Validate IntoTrieIter input generics
• (kermit) Parameterize DatabaseEngine name from caller
• (iters) Remove legacy hash_attribute function
• (ds) Make HashTrie generic over H: HashStrategy
• (ds) Make HashTrieIter generic over H: HashStrategy
• (algos) SingletonHashTrieIter takes precomputed hash
• (kermit) Make hash_join generic over HashStrategy
• (kermit-lab) Polish optimization-axis loader per review
• (kermit-lab) Delete legacy plots/ package and its tests
• (kermit-lab) Final polish — drop dead axis_mapping fns, boolean dtype, README + test gaps
• (kermit-algos) Share Const_ predicate prefix between rewrite and stats; harden docs+tests
• Readability pass — document contracts, clarify names, name magic values
• (cli) Decompose main dispatch into per-subcommand handlers
• (rdf) Consolidate duplicated watdiv/lubm pipeline logic
• (ds) Reduce trie internal complexity, behavior-preserving
• (algos) LFTJ literature alignment — uniform ring indexing
• (rdf) Share WatDiv/LUBM post-driver orchestration via Generator trait
• (kermit-rdf) Drop unused kermit-parser and kermit-ds dependencies
• (ds) Tighten pruning invariant walker and insert_at readability
• (ds) Close the Configured<R, P> drift hole and pin its contract
• (ds) Panic on Singleton-in-Table frames and tighten HashTrieIter docs/tests
• (ds) Expose read_csv/read_parquet so callers can build with a config
• (kermit) Make build_relation the single construction site and pin the load path
• (ds,kermit) Drop the redundant singleton-frame depth, tighten Layout docs and tests
• (kermit) Move optimisation-axis CLI options into options.rs
• (kermit) Derive HashHtj's report labels from its Layout types
• (kermit) Pin the Layout macro pairings and tighten options docs
• (kermit) Collapse run_ds_bench/run_ds_bench_hash onto ExecutionFamily
• (kermit) Split RelationFamily out of ExecutionFamily for bench ds
• (kermit) Collapse the DB layer into family-symmetric join functions
• (kermit) Derive identity labels from axis_value instead of Debug
• (algos) Drop dead dependency, dead spelling, unreachable pub
• (algos) Split src/ by iterator family into sorted/ and hash/
• (algos) Fix planner/executor layering; split optimiser/mod.rs
• (kermit) Move Metric, Criterion helpers and the bench ds runner into bench/
• (kermit) Run bench run through bench::run on a Workload
• (kermit) Group the runner's per-invocation inputs into RunSettings

📚 Documentation

• Update CLAUDE.md and benchmarking architecture spec for consolidation
• Update README benchmarking section for consolidated CLI
• Update CLAUDE.md for cloud-based benchmark system
• Fix inaccuracies and add missing context to CLAUDE.md
• Document bench list subcommand in CLAUDE.md
• (bench) Add benchmarks/README.md documenting YAML schema
• (iters) Add README, crate-level example, clean JoinIterable doc
• (derive) Add README and expand crate-level example
• (parser) Add README, crate-level doctests, # Errors
• (ds) Add # Errors sections and expand trait docs
• (ds) Expand TreeTrie and ColumnTrie struct docs
• (ds) Add README and deny missing_docs
• (algos) Add leapfrog_triejoin module doc and variant doc
• (algos) Add README and deny missing_docs
• (bench) Add module docs, field docs, and # Errors sections
• (bench) Add doctests, README, and deny missing_docs
• (kermit) Expand lib.rs and db module docs
• (kermit) Add main.rs crate doc
• (kermit) Add README and deny missing_docs
• (plans) Add WatDiv benchmark integration design
• Note Const-view rewrite and WatDiv generated YAMLs
• (watdiv) Document generic driver
• (watdiv) Document literal partitioning and dict/YAML coupling
• (benchmarking) Refresh architecture spec for current CLI
• Add gotchas for rustfmt noise and bench CLI quirks
• (watdiv) Spec on-the-fly generation pipeline
• (plans) Add watdiv on-the-fly implementation plan
• Note miri's fchmod limitation in gotchas
• (specs) Add design and schema specs for bench-report v2
• (kermit) Document bench --report-json + kermit-plot in readme
• Add usage guide with end-to-end examples
• Document kermit-plot uv workflow and new path
• (kermit-lab) Add notebook templates and reposition as analysis library
• Fix USAGE.md --format claim and add bench watdiv-gen section
• Document kermit-rdf as 8th crate and lubm-gen subcommand
• (benchmarks) Add lubm benchmark reference
• (benchmarks) Add watdiv benchmark reference
• Polish claude.md after lubm work
• Clarify rustfmt and cache-yaml gotchas in claude.md
• (benchmarks) Add minimal declarative generator examples
• Update CLI references for bench gen subcommands
• Note iter_custom calibration trap in space-bench gotcha
• Refresh user-facing docs for kermit-rdf and bench surface
• (specs) Add design for example jupyter notebook full timeline
• (plans) Add implementation plan for full-timeline notebook
• (kermit-lab) Correct phase 1 prose accuracy
• (kermit-lab) Point at 00_full_timeline.ipynb as the entry point
• Point root readme at the full-timeline notebook
• (benchmarks) Unify LUBM and WatDiv docs under shared schema
• (kermit-rdf) Document VarOrder first-appearance invariant
• (kermit-algos) Document LFTJ invariants and K-var naming
• Address review nits from readability refactor
• (kermit-ds) Clarify LayerStep semantics for duplicate-key path
• (kermit) Fix misleading doc-link target in add_space_bench
• (rdf) Clarify StressParams forwarded vs informational fields
• (ds) Fix HashTrie::collect_tuples doc link path
• (data-structures) Hash-trie.md per Priorities item 3
• (algorithms) Hash-triejoin.md per Priorities item 3
• (specs) Add optimization-standard.md (normative)
• (claude.md) Document optimization standard in 'Extending the System'
• (specs) Document standard axis prefixes in bench-report-schema
• (data-structures) Add Optimizations section to hash-trie.md
• (specs,plans) Restore hash-triejoin design + plan
• (specs,plans) Restore optimization-standard design + plan
• (benchmarks) Document WatDiv default workloads, templates, and absent-predicate tolerance
• (plans) Add WatDiv default workloads implementation plan
• Add human-facing optimization-standard guide
• (specs) Design for optimization-axis-aware kermit-lab plotting
• (plans) Implementation plan for optimization-axis-aware plotting
• (kermit-lab) Document the general plot engine + ablation
• (kermit-lab) Update example notebooks for the new plotting API
• Add BENCHMARKING.md methodology guide; refresh USAGE.md plotting
• (specs) Consolidate optimization-standard guide and normative spec
• (specs) Add LUBM reference comparison notebook design
• (plans) Add LUBM reference notebook implementation plan
• (specs,plans) Correct false claim that bench run all×all skips incompatible pairs
• Audit and correct documentation across the workspace
• (specs) Query optimiser design — planner trait, QueryPlan, cardinality policy
• (plans) Query optimiser implementation plan — 12 TDD tasks
• (specs) Correct Cardinality seam — lives in kermit-ds, stored-tuple (multiset-aware) semantics
• (kermit) Tighten db.rs plan-flow comments; note measured-path planning cost
• Optimiser component docs, extension recipe, architecture and schema updates
• (optimisers) House-style headers, See-also cross-links, CLAUDE.md trim
• Surface --optimiser in README, USAGE, and benchmarking-architecture
• Fix hasher-ablation example — hash-trie pairs with hash-triejoin
• Sync ARCHITECTURE.md and CLAUDE.md with the current architecture
• Correct ARCHITECTURE.md sections overtaken by the #56 fix
• (specs) Design singleton pruning as the first Config consumer ([#58](#58))
• (plans) Implementation plan for singleton pruning as first Config consumer ([#58](#58))
• (ds) Tighten ConfigurableRelation contract and HashTrie constructor docs
• Document singleton pruning, ConfigurableRelation, and the with_config suite
• Correct optimization-standard walkthrough and stale hash-trie text
• Record measured singleton-pruning overhead and the open design decision
• (specs) Amendment 1 — pruning is a Layout, the load-factor cap is the Config consumer
• (plans) Implementation plan for Amendment 1 (pruning Layout, load-factor Config)
• Pruning as a Layout, load factor as the Config; classification rule in the standard
• Correct layout-guarantee wording and stale claims after the Layout switch
• Record Amendment 1 acceptance measurements
• (specs) Mark the pre-amendment optimisation specs superseded and update their axis examples
• Point the option-group references at kermit/src/options.rs
• Record the FxHash pivots in the Amendment 1 acceptance record
• (claude) Require the LUBM hook when adding a query optimiser
• Name the merged bench ds families correctly
• (architecture) Name the HashHtj cell's Layout and Config fields
• Finish renaming the HashHtj cell and runner references after the merge
• (kermit) Describe the family-symmetric join layer; bump kermit to 0.2.0
• (benchmarks) Explain why scripts/ is retained
• (specs) Sweep-hardening design (sub-project A)
• (plans) Sweep-hardening implementation plan
• (specs) Mark sweep-hardening as implemented
• Note incremental reports and runtime workspace root
• (algos) Rewrite README for the reorganised crate
• (algos) Record why the triejoin tests stay inline
• (specs) Design for seeding absent predicates in the WatDiv stress workload ([#63](#63))
• (plans) Implementation plan for WatDiv stress absent-predicate seeding ([#63](#63))
• (benchmarks) Absent predicates are seeded in both WatDiv workloads; note 1-second seed granularity
• (specs) Runner-consolidation design (sub-project B)
• (specs) Fix criterion_groups count in runner-consolidation test bullet
• (plans) Runner-consolidation implementation plan
• Describe bench join's generic-runner surface and the bench/ module
• Fix stale bench join and module-location sentences
• (specs) Answer-verification design (sub-project C)
• (specs) Fold sub-projects D and E into the answer-verification spec
• (plans) Answer-verification, Rust–Python contract and download-integrity plan
• Expected cardinalities, --verify, sha256 integrity, and the kermit-lab contract test
• (benchmarks) Keep the static-fields schema table contiguous
• Note sha256 verification in bench fetch and disambiguate the run-time hash
• Selection rewrite for repeated variables; 14 standard join patterns

⚡ Performance

• (kermit) Skip per-iter trie rebuild in space benchmarks
• (kermit) Load each parquet relation once in run_benchmark
• (ds) Keep HashTrieNode table accessors inlined despite the Singleton panic arm

🎨 Styling

• Apply nightly rustfmt formatting
• Apply nightly rustfmt formatting
• Apply nightly rustfmt formatting
• Apply nightly rustfmt formatting
• Apply nightly rustfmt formatting
• Apply nightly rustfmt formatting
• Apply nightly rustfmt to documentation edits
• (kermit) Apply nightly rustfmt to recently-touched files
• (kermit-rdf) Wrap long doc comments in error.rs to satisfy rustfmt
• (kermit) Apply nightly rustfmt to bench_report
• Apply nightly rustfmt formatting pass
• (kermit-ds) Regroup hex literal in column_trie test seed
• (iters) Cargo fmt + doc/param naming consistency
• (ds) Cargo fmt + clippy fixes after Phase 2
• (ds) Cargo fmt after Phase 3
• (ds) Cargo fmt after Phase 4
• (algos) Cargo fmt after Phase 7
• (iters) Cargo fmt after Phase 1
• (cli) Cargo fmt after Phase 4
• Apply rustfmt to watdiv default-workload changes
• Cargo fmt
• Cargo fmt
• Cargo fmt
• Run rustfmt over the dependency-bump edits
• Rustfmt and align axis_value test imports with the optimiser precedent

🧪 Testing

• (kermit-bench) Add tests for tuple generators and benchmark configs
• (kermit) Add trimmed WatDiv fixture for correctness checks
• (kermit) Integration test — WatDiv mini fixture
• (kermit-rdf) Integration test for stages 4-6
• (kermit-rdf) E2e SF=1 against real watdiv binary
• (kermit) CLI smoke test for bench watdiv-gen
• (kermit-rdf) Gate driver fs tests behind cfg(miri)
• (kermit-rdf) Add lubm pipeline integration tests
• (kermit-ds) Move tree_trie tests to a sibling file
• (kermit-ds) Add randomised round-trip test for column_trie inserts
• (kermit) Pin bench run JSON-schema with triangle benchmark
• (kermit) Address review findings on bench-run schema test
• Miri-gate e2e_watdiv and de-collide cli_join temp path
• Sort both sides in test_join for multiset equality
• Define_multiway_join_test_suite!(HashTrie, HashTriejoin)
• (cli) Unit tests for validate_layout_choices
• Split HashTrie suite by Layout variant via type aliases
• (lubm) Add cardinality regression test
• (rdf) E2e test for the WatDiv Basic Testing pipeline
• (kermit-ds) Harden HashTrie multiset counter — from_tuples duplicate pin + desync guard comment
• (kermit-algos) Pin repeated-variable analysis edge; state head-index guarantee
• (kermit-algos) Pin out-of-range plan rejection; document rank call frequency
• (kermit-algos) Pin CardinalityOptimiser min-fold semantics; drop redundant tie-break
• Run the 11-pattern multiway join suite under both optimisers (88 tests)
• (ds) Add HashTrie iterator-contract, Parquet, and collision suites
• (ds) Run HashTrie iterator, Parquet and collision suites with pruning on
• (kermit) Add define_multiway_join_test_suite_with_config! and pruning-on suites
• (kermit) CLI smoke test for --ds-config singleton-pruning
• (kermit) CLI smoke test for --ds-layout-pruning
• (kermit) Share the CLI smoke-test helpers in tests/common/cli.rs
• (kermit) Run the LUBM cardinality check under every optimiser
• (kermit) Add a deep-dead-end pattern to the standard join suite
• (kermit) Partial bench run sweep keeps finished benchmarks' reports
• (rdf) Pin seed_missing_predicates with a deterministic unit test
• (rdf) Accept seeded relations in e2e_watdiv and assert YAML/seeded relations reach Parquet
• (kermit) Pin the Python schema version to REPORT_SCHEMA_VERSION
• Reference the head-projection issue from the diagonal tests
• (rdf) Pin pipeline artifact contents on hand-crafted input
• (rdf) Assert the exact entailment closure of interacting rules
• (kermit) Java-free miniature LUBM oracle
• (rdf) Tighten pipeline fixture; document partition tuple order
• (kermit) Mini LUBM oracle covers every optimiser and constant

🔧 Build

• (flake) Expose python and ld_library_path for kermit-plot
• (kermit-plot) Migrate to uv with hatchling backend
• (flake) Add uv to dev shell
• (flake) Add jdk 8 for lubm-uba jar runtime
• (kermit-lab) Sync uv.lock with 0.2.0 package version
• (flake) Track CI's nightly and fix the drift it exposes
• (deps) Bump dirs to 7 and reqwest to 0.13
• (deps) Bump sha2 to 0.11 and hex-encode digests explicitly
• (deps) Bump syn to 3
• (deps) Bump winnow to 1.0
• (deps) Bump arrow and parquet to 59
• (deps) Bump oxrdf, oxttl and spargebra
• (deps) Bump criterion to 0.8

⚙️ Miscellaneous Tasks

• Add .worktrees/ to .gitignore
• (watdiv) Scaffold Python preprocessor skeleton
• (kermit) Fail loud on missing relation in join; ignore watdiv-data/
• (kermit-rdf) Drop redundant tempfile dev-dep + premature kermit-algos dep
• Add watdiv binary to .gitignore
• Ignore .loom/ output directory
• (kermit-plot) Move project from scripts/ to python/
• (kermit-lab) Rename from kermit-plot
• Ignore log.txt files emitted by lubm-uba jar
• (kermit-rdf) Vendor lubm-uba jar with provenance
• Ignore .serena/ MCP server config dir
• Ignore .claude.local.md
• (release) Bump crate versions and add kermit-rdf changelog
• Remove dead Node trait and stale commented test block
• Update CLAUDE.md
• (kermit) Stub HashTrie arms in CLI dispatch until phase 9
• (release) Bump workspace crate versions
• (python) Ignore Jupyter notebook checkpoint directories
• (kermit-lab) Resave timeline notebook with kermit-lab kernelspec
• Resolve architecture-review cleanups ([#60](#60))
• (algos) Drop four dependencies the crate never used
• Run kermit-lab tests with the real-binary contract test
• Cache uv in the kermit-lab job
• (release) Stop publishing to crates.io