refactor!: fold lance-csr-catalog into lance-csr and pin namespace reads

.bumpversion.toml

@@ -33,11 +33,6 @@ filename = "crates/lance-csr/Cargo.toml"
 search = 'version = "{current_version}"'
 replace = 'version = "{new_version}"'
 
-[[tool.bumpversion.files]]
-filename = "crates/lance-csr-catalog/Cargo.toml"
-search = 'version = "{current_version}"'
-replace = 'version = "{new_version}"'
-
 [[tool.bumpversion.files]]
 filename = "crates/lance-csr-benches/Cargo.toml"
 search = 'version = "{current_version}"'

.github/workflows/python-publish.yml

@@ -217,8 +217,8 @@ jobs:
 
   publish-crates:
     # After PyPI, so the two registries never disagree. If only this job fails, rerun it on
-    # its own; the wheels don't need rebuilding. The crates.io trusted publishers for both
-    # crates are registered for this workflow and environment `crates-io`.
+    # its own; the wheels don't need rebuilding. The crates.io trusted publisher for the
+    # crate is registered for this workflow and environment `crates-io`.
     needs: publish
     runs-on: ubuntu-24.04
     timeout-minutes: 60
@@ -237,14 +237,13 @@ jobs:
           sudo apt-get install -y protobuf-compiler
 
       # Verifying compiles lance, so do it before getting the token, which lasts 30 minutes.
-      # Cargo packages lance-csr-catalog first, since lance-csr depends on it.
-      - name: Package and verify the crates
-        run: cargo package --locked -p lance-csr-catalog -p lance-csr
+      - name: Package and verify the crate
+        run: cargo package --locked -p lance-csr
 
       - uses: rust-lang/crates-io-auth-action@v1
         id: auth
 
       - name: Publish to crates.io
-        run: cargo publish --locked --no-verify -p lance-csr-catalog -p lance-csr
+        run: cargo publish --locked --no-verify -p lance-csr
         env:
           CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}

.github/workflows/python-test.yml

@@ -46,8 +46,7 @@ jobs:
           shared-key: "lance-csr-python"
           # maturin builds the bindings crate, which sits outside the root
           # workspace. The root workspace is listed too so the cache key hashes
-          # the manifests of lance-csr and lance-csr-catalog, which the bindings
-          # use by path; otherwise a dependency or feature change there keeps
+          # the manifest of lance-csr, which the bindings use by path; otherwise a dependency or feature change there keeps
           # restoring a stale cache. Its target directory is never built here.
           workspaces: |
             . -> target

AGENTS.md

@@ -2,7 +2,6 @@
 
 ## Project Structure & Module Organization
 - `crates/lance-csr/` hosts the Rust Cypher engine; keep new modules under `src/` and co-locate helpers inside `query/` or feature-specific submodules.
-- `crates/lance-csr-catalog/` provides the graph source catalog and `DirNamespace` (Unity Catalog and Delta Lake support were removed).
 - `crates/lance-csr-benches/` contains performance benchmarks in a dedicated unpublished crate; add new benchmarks here rather than in the main crate.
 - `crates/lance-csr-python/src/` contains the PyO3 bridge; `python/python/lance_csr/` holds the pure-Python facade and packaging metadata.
 - `python/python/tests/` stores functional tests; mirror new features with targeted cases here and in the corresponding Rust module.

Cargo.toml

@@ -1,7 +1,6 @@
 [workspace]
 members = [
     "crates/lance-csr",
-    "crates/lance-csr-catalog",
     "crates/lance-csr-benches",
 ]
 # Python package needs to be built by maturin.

RELEASE.md

@@ -1,7 +1,7 @@
 # Releasing lance-csr
 
 A release publishes the Python package to [PyPI](https://pypi.org/project/lance-csr/) and the
-`lance-csr-catalog` and `lance-csr` crates to [crates.io](https://crates.io/crates/lance-csr). PyPI publishing is
+`lance-csr` crate to [crates.io](https://crates.io/crates/lance-csr). PyPI publishing is
 automated from a GitHub release, and the crates follow once PyPI has the new version.
 
 The whole release takes about 50 minutes: the macOS wheel build, then the crates. Don't add steps that build the
@@ -26,9 +26,9 @@ cargo update --workspace --manifest-path crates/lance-csr-python/Cargo.toml
 (cd python && uv lock)
 ```
 
-`bump-my-version` updates `.bumpversion.toml`, the four crate `Cargo.toml` files and `python/pyproject.toml`. It
+`bump-my-version` updates `.bumpversion.toml`, the three crate `Cargo.toml` files and `python/pyproject.toml`. It
 doesn't update the three lockfiles that also carry the version, which is what the other three commands are for.
-The diff should be 9 files and 15 lines each way, touching only version lines. If `uv lock` or `cargo update`
+The diff should be 8 files and 11 lines each way, touching only version lines. If `uv lock` or `cargo update`
 moves a dependency, undo that and update only the workspace packages. Add `--allow-dirty` to `bump-my-version` if
 untracked files in your checkout stop it.
 
@@ -68,12 +68,12 @@ Publishing it starts `.github/workflows/python-publish.yml`:
 | `build-macos` | about 35 min | macOS arm64 wheel, built natively on `macos-15` |
 | `test` | under 1 min each | each wheel on Python 3.12, 3.13 and 3.14 |
 | `publish` | seconds | checks the tag matches `python/pyproject.toml`, then uploads through PyPI trusted publishing |
-| `publish-crates` | about 10 to 15 min | after `publish`: packages and verifies `lance-csr-catalog` and `lance-csr`, then uploads them through crates.io trusted publishing |
+| `publish-crates` | about 10 to 15 min | after `publish`: packages and verifies `lance-csr`, then uploads it through crates.io trusted publishing |
 
 The build jobs run in parallel. Nothing reaches PyPI unless every build and test job passes, and nothing
 reaches crates.io unless PyPI publishing succeeds. The `pypi` and `crates-io` environments only accept `v*`
 tags. The PyPI trusted publisher is registered for workflow `python-publish.yml` and environment `pypi`, and
-the crates.io trusted publishers for both crates for the same workflow and environment `crates-io`, so don't
+the crates.io trusted publisher for `lance-csr` for the same workflow and environment `crates-io`, so don't
 rename any of them.
 
 **Don't run a separate `mode=dry_run` before a release.** The release run builds and tests the same wheels
@@ -94,11 +94,8 @@ If a build or test job failed, nothing has been published, so nothing needs undo
 PyPI already has the release, so don't release again. Rerun just that job with `gh run rerun <run id> --failed`;
 the wheels aren't rebuilt.
 
-Cargo publishes `lance-csr-catalog` first, since `lance-csr` depends on it. If it published the catalog and
-then failed on `lance-csr`, a rerun fails because the catalog version already exists. Publish the rest by hand
-from the release tag: a crate owner runs `cargo publish -p lance-csr`. If **Require trusted publishing** is on
-for the crate, crates.io rejects API tokens, so turn it off for the manual publish and back on afterwards.
-`lance-csr-python` and `lance-csr-benches` are `publish = false`.
+`lance-csr-python` and `lance-csr-benches` are `publish = false`. Up to 0.7.1, the catalog types shipped
+in a separate `lance-csr-catalog` crate; it is no longer published.
 
 ## 4. Check the release
 

crates/lance-csr/Cargo.toml

@@ -33,7 +33,6 @@ datafusion-expr = "54"
 datafusion-sql = "54"
 datafusion-functions-aggregate = "54"
 futures = "0.3"
-lance-csr-catalog = { path = "../lance-csr-catalog", version = "0.7.1" }
 lance = "13.0.0"
 lance-linalg = "13.0.0"
 lance-namespace = "13.0.0"

crates/lance-csr/README.md

@@ -157,12 +157,13 @@ clauses, patterns, expressions and functions the engine runs.
 - `query` – High level `CypherQuery` API and runtime.
 - `csr_index`, `csr_expand` – The CSR adjacency index and the traversal operators that use it.
 - `error` – `GraphError` and result helpers.
-- `namespace` – Namespace helpers (re-exported from `lance-csr-catalog`).
-- `source_catalog` – Catalog helpers for looking up table metadata (re-exported from `lance-csr-catalog`).
+- `mapping` – `GraphMapping`, the JSON graph descriptor that maps labels to tables, and
+  `GraphSnapshot`, which pins each table to a Lance version and opens them for `CypherEngine`.
+- `namespace` – `DirNamespace`, a Lance namespace for a directory of datasets.
+- `source_catalog` – `GraphSourceCatalog`, which resolves labels to DataFusion table sources.
 
-`lance-csr` re-exports the catalog and namespace types from the `lance-csr-catalog` crate for
-API compatibility. You can depend on `lance-csr-catalog` directly if you only need catalog or
-namespace utilities.
+Up to 0.7.1 the `namespace` and `source_catalog` types lived in a separate `lance-csr-catalog` crate.
+They now live here, at the same root paths (`lance_csr::DirNamespace` and so on).
 
 ## Error Handling
 

crates/lance-csr/src/apply.rs

@@ -506,7 +506,7 @@ mod tests {
     #[test]
     fn call_bypasses_snapshot_templates_in_every_reading_stage() {
         let (config, _) = graph();
-        let catalog = Arc::new(lance_csr_catalog::InMemoryCatalog::new());
+        let catalog = Arc::new(crate::InMemoryCatalog::new());
         for text in [
             "CALL { RETURN 1 AS n } RETURN n",
             "WITH 1 AS x CALL { WITH x RETURN x AS n } RETURN n",

crates/lance-csr/src/datafusion_planner/config_helpers.rs

@@ -10,7 +10,7 @@ use super::analysis::PlanningContext;
 use super::DataFusionPlanner;
 use crate::config::{NodeMapping, RelationshipMapping};
 use crate::error::Result;
-use lance_csr_catalog::GraphSourceCatalog;
+use crate::GraphSourceCatalog;
 use std::sync::Arc;
 
 impl DataFusionPlanner {

crates/lance-csr/src/datafusion_planner/join_ops.rs

@@ -14,11 +14,11 @@ use crate::ast::{PropertyValue, RelationshipDirection};
 use crate::case_insensitive::qualify_column;
 use crate::config::{NodeMapping, RelationshipMapping};
 use crate::error::Result;
+use crate::GraphSourceCatalog;
 use datafusion::common::Column;
 use datafusion::logical_expr::{
     col, BinaryExpr, Expr, JoinType, LogicalPlan, LogicalPlanBuilder, Operator, TableSource,
 };
-use lance_csr_catalog::GraphSourceCatalog;
 use std::collections::HashMap;
 use std::sync::Arc;
 

crates/lance-csr/src/datafusion_planner/mod.rs

@@ -37,8 +37,8 @@ pub(crate) use scalar_functions::sql_expr;
 use crate::config::GraphConfig;
 use crate::error::Result;
 use crate::logical_plan::LogicalOperator;
+use crate::GraphSourceCatalog;
 use datafusion::logical_expr::LogicalPlan;
-use lance_csr_catalog::GraphSourceCatalog;
 use std::sync::Arc;
 
 /// Planner abstraction for graph-to-physical planning

crates/lance-csr/src/datafusion_planner/scan_ops.rs

@@ -118,7 +118,7 @@ impl DataFusionPlanner {
         // No catalog attached - create empty source fallback for flexibility
         // This allows planners created with DataFusionPlanner::new() to work
         // without requiring a catalog, though they won't have actual data sources
-        let empty_source = Arc::new(lance_csr_catalog::SimpleTableSource::empty());
+        let empty_source = Arc::new(crate::SimpleTableSource::empty());
         let normalized_label = label.to_lowercase();
         let builder =
             LogicalPlanBuilder::scan(&normalized_label, empty_source, None).map_err(|e| {

crates/lance-csr/src/datafusion_planner/test_fixtures.rs

@@ -5,8 +5,8 @@
 
 use crate::config::GraphConfig;
 use crate::logical_plan::LogicalOperator;
+use crate::{InMemoryCatalog, SimpleTableSource};
 use arrow_schema::{DataType, Field, Schema};
-use lance_csr_catalog::{InMemoryCatalog, SimpleTableSource};
 use std::sync::Arc;
 
 pub fn person_schema() -> Arc<Schema> {
@@ -17,7 +17,7 @@ pub fn person_schema() -> Arc<Schema> {
     ]))
 }
 
-pub fn make_catalog() -> Arc<dyn lance_csr_catalog::GraphSourceCatalog> {
+pub fn make_catalog() -> Arc<dyn crate::GraphSourceCatalog> {
     let person_src = Arc::new(SimpleTableSource::new(person_schema()));
     let knows_schema = Arc::new(Schema::new(vec![
         Field::new("src_person_id", DataType::Int64, false),

crates/lance-csr/src/engine.rs

@@ -511,8 +511,9 @@ fn config_error(message: String) -> GraphError {
     }
 }
 
-/// Open a Lance dataset. `version = None` means the latest version.
-async fn open_lance_dataset(
+/// Open the Lance dataset at `uri`, pinned to `version`, or at its latest version when
+/// `version` is `None`. Storage options go to the object store, as in Lance itself.
+pub async fn open_lance_dataset(
     uri: &str,
     version: Option<u64>,
     storage_options: &HashMap<String, String>,