Lance format

Lance v9 reference

Lance is an open columnar format for multimodal AI - "a columnar data format that is 100x faster than Parquet for random access." It is not one format but a stack of interoperating specs: a file format, a table format, index formats, catalog specs, and a namespace client spec. The Rust workspace at lance-format/lance implements all of them plus Python (pylance) and Java bindings.

This skill tracks v9.1.0-beta.8 (the lance-format/lance git tag), the current development frontier. Pin against tags, not main - Lance ships beta tags every few days and next-format encodings can change. v9.0.0 never got a final tag; the line reached v9.0.0-rc.2 (2026-07-21) while main advanced to the 9.1.0-beta.* dev series. v8.0.0 final shipped 2026-07-01 - if you need a stable pin rather than the v9 dev betas, track v8.0.0 (the last final), whose format and API are the frozen predecessor of what this reference describes.

Three layers of reference, load what the task needs:

• references/lance-reference.md - the distilled deep reference. Any concrete schema, parameter, proto, or constraint.
• references/performance.md - ALL official performance guidance combined, plus field-verified remote-storage practices. Load for any performance, tuning, maintenance-cost, or "why is this slow" question.
• references/docs/ - a verbatim mirror of the official docs (docs/src at the tracked tag): every guide, quickstart, and format spec, unedited. Load when you need the full official text. Complete file map below.

This file is the orientation: read it first, then jump into what you need.

Lance vs LanceDB

These are two different things and conflating them produces wrong answers.

• Lance - the format and engine. The lance-format/lance repo; the lance /lance-* Rust crates; pylance. It gives you datasets, the file/table format, indexes, commits, scans. Consumed directly by DuckDB, Polars, Ray, Spark, PyTorch, DataFusion, or your own Rust/Python code. This skill is about Lance.
• LanceDB - a separate database product (lancedb/lancedb) built on top of Lance. It adds a query-builder API, an embedding registry, rerankers-as-API, multi-language SDK parity, and managed Cloud / Enterprise tiers. Not covered here.

If you are linking the lance crate in Cargo.toml, you are using Lance directly - use this skill. If a question is about LanceDB internals, the storage layer underneath it is still Lance, so this skill remains the authority for the format itself.

The crate workspace

26 crate directories under rust/. lance is the public entry point; the rest are layers beneath it. Full table with descriptions and citations in references/lance-reference.md section 2.

Crate	Role
lance	Public entry point - Dataset, scanner, indexes, commits
lance-table	Table format - manifest, feature flags, commit handlers, row IDs
lance-file	File format - file reader/writer
lance-encoding	Structural encodings, compression (internal, not for external use)
lance-index	Scalar / vector / FTS / system indexes
lance-index-core	Shared index primitives extracted from lance-index (new in v9.1, PR #7713)
lance-io	Object store, I/O schedulers
lance-core	Shared Error/Result, cache, datatypes
lance-datafusion	DataFusion glue (exec, expr, planner, UDFs)
lance-linalg	SIMD L2 / dot / cosine / hamming kernels
lance-select	Row-selection primitives - RowAddrMask, RowIdMask, IndexExprResult (extracted from lance-core/lance-index in v7.1.0-beta.2)
lance-tokenizer	FTS tokenizer stack (simple, ngram, jieba, lindera, stemmers)
lance-derive	#[derive(DeepSizeOf)] proc-macro for Arrow-aware memory accounting (new in v8, PR #6229; replaced the external deepsize crate)
lance-geo	Geospatial UDFs (feature-gated geo)
lance-namespace / -impls / -datafusion	Namespace trait, Directory/REST impls, DataFusion catalog bridge
lance-arrow, lance-tools, fsst, lance-bitpacking, ...	Arrow extensions, CLI, compression sub-crates

All share version = "9.1.0-beta.8" except lance-arrow-scalar, which is pinned at 58.0.0 to track Arrow. Workspace: edition 2024, rust-version = 1.91.0 (MSRV; the pinned build toolchain in rust-toolchain.toml moved to 1.97.0 in v9.1, PR #7712), resolver = "3"; notable deps arrow 58, datafusion 54 (bumped from 53 in v9.1, PR #7793), opendal 0.57, jieba-rs 0.10, lance-namespace-reqwest-client 0.8.6, itertools 0.14. Python bindings require Python 3.10+ (3.9 dropped in v9, PR #7345; 3.14 support added in v9.1, PR #7728).

File format versions

The file format carries a single major.minor version. Selected per-dataset at creation via data_storage_version and fixed once the dataset exists (to change it, rewrite the dataset).

Version	Status	Notes
0.1 (legacy)	read-only	Original format; no longer writable
2.0	stable	Removed row groups; null support for lists/FSL/primitives
2.1	current default (stable)	Adaptive structural encodings; better integer/string compression; nulls in struct fields; better nested random access. Default since Lance 5.0.0
2.2	unstable	Map type, Blob v2, VariablePackedStruct, larger mini-blocks. Required for Map and Blob v2; the real experimental frontier - encodings may still change
2.3	unstable (next)	The current next alias target (V2_3 in the enum). No longer scaffolding - as of v9.1 it ships sparse structural pages (lance-encoding:structural-encoding=sparse, PR #7889); V2_3 references in lance-encoding jumped 6 -> 59. The docs version table now names it "sparse structural pages and other experimental encodings"

stable resolves to the default (2.1); next now resolves to 2.3 (not 2.2) in the running Lance release - pin an explicit number for deterministic behavior. In the version ladder 2.2 sits below next, so the code does not flag 2.2 as unstable. The docs version table (docs/src/format/file/versioning.md) lists 2.3 as the unstable row and no longer labels 2.2 unstable - and 2.3 now carries its own concrete experimental encoding (sparse pages), on top of the 2.2-era features (Map, Blob v2, VariablePackedStruct).

What's new in v9

v9.0 -> v9.1 (the current frontier)

The 9.1 dev line branched off main when v9.0.0-rc.1 was cut for stabilization - an automatic release-train bump, not a breaking change. Structural deltas from beta.18: 26 crates (new lance-index-core, #7713), 16 transaction ops (new DataOverlay, below), datafusion 53 -> 54 (#7793), and the build toolchain moved to Rust 1.97.0 (#7712; MSRV rust-version unchanged at 1.91.0). CommitConfig.num_retries still 20, file-format enum still next => 2.3 / default 2.1.

One breaking-labeled PR in the window: FTS/inverted-index creation takes a block_size param (compressed posting blocks; 128 or 256, default 128, 512 rejected, #7466). block_size=256 and the new code-analyzer tokenizer (#7681) require FTS on-disk format v3 (#7866) - readers must support v3 before such an index exists.

Net-new in 9.1:

• Data Overlay Files - attach new values for a subset of (row offset, field) cells to a fragment without rewriting its base data files (the upstream answer to cheap cell-level updates). New DataOverlay transaction op, feature flag 64, spec references/docs/format/table/data_overlay_file.md. Still unstable: env-gated by LANCE_ENABLE_UNSTABLE_DATA_OVERLAY_FILES, and release builds refuse overlay datasets (#7535, #7536).
• Sparse structural pages - the first real 2.3 encoding; represent flat/nested Arrow structure as slot-domain mappings instead of dense rep/def events (#7889).
• Exact IS NULL from zonemap and bloom-filter indexes via a new null_bitmap (was inexact / AtMost); nested-field FTS (index leaf fields like data.text, #7686); FTS impact-skip / MAXSCORE top-k / bulk-conjunction paths (#7602, #7603, #7624).
• OpenTelemetry metrics for Python (instrument_lance_metrics, pylance[otel], #7537); AWS creds via AssumeRoleWithWebIdentity to avoid role chaining (#7757); Python 3.14 support (#7728). Full delta in references/lance-reference.md section 14.

v8 -> v9

The v8 -> v9 boundary is a light major bump: structurally v9.0 was nearly identical to v8 (same 25 crates, 15 transaction ops at that point, file-format enum with next => 2.3 and default 2.1, CommitConfig.num_retries 20, arrow 58 / datafusion 53 / opendal 0.57 / jieba 0.10). The major version was auto-triggered by Lance's breaking-change-label detector (ci/check_breaking_changes.py), fired by two PRs: Python 3.9 was dropped (minimum now 3.10, #7345) and alter_columns now fails fast when you cast a column that has an index attached - you must drop_index() first instead of relying on the old silent drop/invalidate (#7158). A third breaking change rode the already-bumped series: the FM-Index proto message was renamed FMIndexIndexDetails -> FMIndexDetails (#7397), which makes existing FM indexes unreadable. One public Rust-API removal: as_vector_index is gone from the Index trait (#7392) - downcast via as_any(). A fourth breaking change landed later in the v9 beta line: FTS / inverted indexes now default to on-disk format v2 (#7512) - LANCE_FTS_FORMAT_VERSION no longer controls new indexes, pass format_version=1 if older Lance readers must read them (existing v1 indexes stay queryable, section 11.3).

Net-new in v9: a hamming clustering utility for near-duplicate detection (SIMD union-find over 64-bit binary hashes, #7379); COUNT(*) pushdown now works on stable-row-id datasets (#7360); per-column blob size thresholds (lance-encoding:blob-inline-size-threshold / ...-dedicated-size-threshold, #7269); tunable 32k miniblock chunks via LANCE_MINIBLOCK_MAX_VALUES (#7356, default still 4096); an icu/split FTS tokenizer variant (#7474); distributed LabelList index builds (#7223); the ngram index now accelerates regex and infix LIKE (#7139); and cleanup-explain plus fragment-reuse remap are now exposed to Python and Java (#7248, #7438). Full delta in references/lance-reference.md section 14.

The v7 -> v8 boundary (the predecessor line) unified all index builds onto one segment-based lifecycle: bitmap migrated to the segment workflow (#6869), the standalone IndexSegmentBuilder API was removed (#6997), distributed BTree moved to the segmented framework (#7013), file writers' finish() began returning FileWriteSummary (#7096), and describe_indices()/list_indices() were reworked (#6903, #7129). v8 also added the lance-derive crate (#6229), the FM-Index scalar index, multi-bit IVF_RQ (num_bits 1..=9), the public vector-search approx_mode (fast/normal/accurate), and the Volcengine TOS (tos://) and feature-gated GooseFS (goosefs://) object stores. The v7 era - MemWAL, branches, the geo/RTree index, the lance-select crate, ICU FTS - all carries forward.

Navigating the reference

references/lance-reference.md is the full v9 reference, regrounded against the v9.1.0-beta.8 source (127 commits from v9.0.0-beta.18, 1 breaking change - delta in its section 14). Load the section for your task:

1. What Lance is - the lakehouse spec stack
2. Crate workspace - all 26 crates, what each does, the public entry point
3. File format - versions, container layout, structural encoding (mini-block / full-zip / constant / blob page types), compression schemes, blob encoding
4. Data types - Arrow type coverage, FixedSizeList for vectors, JSON (JSONB), blob, ML extension arrays (bfloat16, image types)
5. Table format - dataset directory layout, manifest contents, fragments, deletion files, base paths
6. Schema evolution - field IDs, zero-copy column add/drop/alter, why old rows read NULL
7. Versioning, tags, branches - manifest versions, time travel, tag pinning, branches
8. Row IDs - row address vs stable row ID, lineage, change-data-feed columns
9. Transactions and concurrency - the 16 transaction ops, OCC retry/rebase, commit handlers (conditional-put, DynamoDB), conflict resolution matrix
10. MemWAL - shards, MemTable/WAL/flush, the appender/tailer/flusher model, fencing
11. Indexes - vector (IVF/HNSW/PQ/SQ/RQ, multi-bit RQ), scalar (btree/bitmap/bloom/ labellist/ngram/zonemap/FM-Index), full-text (BM25, tokenizers), geo/RTree
12. Distributed write and indexing - two-phase commits, segment-based index builds
13. Object store - URI schemes, storage options, per-backend config
14. What changed - the full v7 -> v8 -> v9 delta
15. Capability matrix - what Lance can and cannot do
16. Source map - where each spec and proto lives in the repo

Performance questions

For anything performance-shaped - slow scans or searches, remote/object-storage cost, index maintenance cost, memory sizing, version bloat, benchmarking - load references/performance.md first. Part A collects every official performance recommendation in one place; Part B is field-verified practice from running Lance against S3-compatible storage, whose governing rule is: leave the store knobs (LANCE_IO_THREADS, LANCE_AIMD_*, timeouts, compression metadata) at their defaults and optimize by minimizing remote calls - fewer commits, fewer scans, fewer round trips.

Official docs mirror - file map

references/docs/ mirrors docs/src of lance-format/lance at the tracked tag, verbatim. Every file below is directly readable; pick by topic.

Guides (references/docs/guide/)

File	Covers
read_and_write.md	CRUD, merge_insert semantics, cleanup_old_versions + automatic cleanup
performance.md	The official performance guide (also embedded in references/performance.md)
object_store.md	URI schemes, credentials, storage_options per backend (S3/GCS/Azure/...)
distributed_write.md	Two-phase distributed writes - fragments on workers, single commit
distributed_indexing.md	Segment-based distributed index builds, merge, finalize
json.md	JSON columns, json_get_* / json_extract, JSON scalar index
tokenizer.md	FTS tokenizer configuration - language, stemming, jieba/lindera/icu
data_types.md	Arrow type coverage, FixedSizeList vectors, JSON, blob
data_evolution.md	Zero-copy add/drop/alter columns, backfills
blob.md	Blob columns - storing and reading large binary
arrays.md	ML extension arrays (bfloat16, image types)
tags_and_branches.md	Managing tags and branches
migration.md	Migration guides across Lance versions
observability.md	Logging, trace events, object-store metrics

Quickstarts (references/docs/quickstart/)

File	Covers
index.md	First dataset - core table operations end to end
vector-search.md	ANN index tutorial - build, tune, filtered search
full-text-search.md	FTS tutorial incl. index maintenance + performance tips
versioning.md	Time travel, restore, cleanup basics

Format specs (references/docs/format/)

File	Covers
index.md	The spec-stack overview
file/index.md	File-format container spec
file/encoding.md	Structural encodings and compression strategy
file/versioning.md	File-format versions (2.0 / 2.1 / 2.2 / 2.3)
table/index.md	Table-format overview
table/data_overlay_file.md	Data Overlay Files - cell-level (offset, field) updates without base-file rewrite (feature flag 64, unstable)
table/layout.md	Dataset directory layout
table/schema.md	Schema and field-metadata spec
table/transaction.md	Commit protocol, transaction ops, conflict-resolution matrix
table/versioning.md	Manifest versioning and feature flags
table/row_id_lineage.md	Stable row IDs, lineage, change-data-feed columns
table/branch_tag.md	Branch and tag spec
table/mem_wal.md	MemWAL / streaming-write spec (experimental)

Index specs (references/docs/format/index/)

File	Covers
index.md	Index lifecycle - creation, fragment coverage, compaction interplay (with diagrams)
vector/index.md	Vector indices - IVF / PQ / SQ / RQ / HNSW concepts and storage layout
scalar/fts.md	Inverted (FTS) index - tokenizers, posting lists, memory/disk costs
scalar/fmindex.md	FM-Index - substring / regex search over raw bytes
scalar/ngram.md	N-gram index - contains() and LIKE acceleration
scalar/btree.md	BTree - range queries, sorted access
scalar/bitmap.md	Bitmap - low-cardinality equality
scalar/bloom_filter.md	Bloom filter index
scalar/label_list.md	Label-list index for array_has_any/all
scalar/zonemap.md	Zone maps - per-zone min/max/null-count stats
scalar/rtree.md	R-Tree geospatial index
system/frag_reuse.md	Fragment Reuse Index - compaction without index remap
system/mem_wal.md	MemWAL system index entry

Integrations (references/docs/integrations/)

File	Covers
datafusion.md	SQL over Lance via DataFusion, incl. JSON functions

Maintenance

Citations in references/lance-reference.md are path:line relative to the lance-format/lance repo; build a permalink as https://github.com/lance-format/lance/blob/v9.1.0-beta.8/<path>.

To refresh: git -C ~/pjv/lance-format/lance fetch --tags, check out the newest v9* tag (or the v8.0.0 rc/final line for a stable pin), then:

1. Re-copy the docs mirror: the .md files of docs/src/{guide,quickstart}, docs/src/format (plus the format/index/*.svg diagrams), and docs/src/integrations/datafusion.md into references/docs/, preserving the tree. Update this file's file map if docs were added or removed.
2. Rebuild Part A of references/performance.md from the new guide/performance.md and the other perf sections it cites. Part B (field-verified practices) is experience-derived - only edit it with new measured results, never speculation.
3. Re-verify the crate workspace and re-read the format spec for references/lance-reference.md, then bump metadata.upstream plus every current-tag version reference. Line numbers in citations drift between tags - treat them as approximate.