ChristopherWirt · Nov 13, 2026 · Nov 11, 2026 · Nov 11, 2026 · Nov 11, 2026 · Nov 11, 2026
diff --git a/Cargo.lock b/Cargo.lock
diff --git a/benchmarks/compress-bench/README.md b/benchmarks/compress-bench/README.md
@@ -11,7 +11,8 @@ ratios. Parquet provides the established reference for a compressed columnar rep
 
 The suite covers NYC taxi data and several
 [Public BI](https://github.com/cwida/public_bi_benchmark) tables (Arade, Bimbo,
-CMSprovider, Euro2016, Food, HashTags), TPC-H `l_comment` variants, and synthetic nested
+CMSprovider, Euro2016, Food, HashTags), TPC-H `l_comment` variants, the TPC-H `lineitem`,
+`orders`, and `partsupp` tables at scale factor 1 (`TPC-H SF1 <table>`), and synthetic nested
 data. This is the workload behind the `Compression` PR comment.
 
 See [`src/main.rs`](./src/main.rs) for the dataset list and CLI flags (`--formats`,

diff --git a/benchmarks/compress-bench/src/main.rs b/benchmarks/compress-bench/src/main.rs
@@ -41,6 +41,7 @@ use vortex_bench::datasets::struct_list_of_ints::StructListOfInts;
 use vortex_bench::datasets::taxi_data::TaxiData;
 use vortex_bench::datasets::tpch_l_comment::TPCHLCommentCanonical;
 use vortex_bench::datasets::tpch_l_comment::TPCHLCommentChunked;
+use vortex_bench::datasets::tpch_sf1::TpchSf1Table;
 use vortex_bench::display::DisplayFormat;
 use vortex_bench::display::print_measurements_json;
 use vortex_bench::display::render_table;
@@ -325,6 +326,7 @@ async fn run_compress(
         &DownloadableDataset::AirQuality,
     ]
     .into_iter()
+    .chain(TpchSf1Table::ALL.iter().map(|d| d as &dyn Dataset))
     .chain(structlistofints.iter().map(|d| d as &dyn Dataset))
     .chain(sparse_wide.iter().map(|d| d as &dyn Dataset))
     .collect();

diff --git a/benchmarks/string-bench/README.md b/benchmarks/string-bench/README.md
@@ -78,11 +78,14 @@ The current input catalog is:
 | `news/headlines` | "A Million News Headlines" (ABC News), `headline_text` | 1.23M |
 | `twitter/sentiment140` | Sentiment140 tweets, train split, `text` | 1.36M |
 | `fineweb/text` | FineWeb `sample/10BT/001_00000.parquet`, `text` | first 10k |
+| `tinystories/text` | TinyStories, train split, `text` | first 33,121 |
 
 The Amazon, news, and tweet inputs are four of the five datasets the
 [OnPair paper](https://arxiv.org/abs/2508.02280) evaluates on. The fifth,
 OpenWebText URLs, has no public parquet mirror; ClickBench `URL` covers that
-shape. FineWeb adds long web documents.
+shape. FineWeb adds long web documents. TinyStories adds short
+stories with a small vocabulary; 33,121 rows is one file of the NeMo Curator text IO
+benchmark.
 
 On first use, the selected ClickBench shard is downloaded through
 `vortex-bench`'s shared idempotent downloader. TPC-H SF1 is generated

diff --git a/benchmarks/string-bench/src/lib.rs b/benchmarks/string-bench/src/lib.rs
@@ -359,7 +359,8 @@ pub struct HfColumn {
 
 /// The datasets the OnPair paper (Gargiulo & Venturini, "OnPair: Short Strings
 /// Compression for Fast Random Access", arXiv:2508.02280) evaluates on, from
-/// public parquet mirrors, plus a slice of FineWeb as long-document web text.
+/// public parquet mirrors, plus a slice of FineWeb as long-document web text and
+/// a slice of TinyStories as short-story text.
 ///
 /// The paper's fifth dataset, OpenWebText URLs, has no parquet mirror; the
 /// ClickBench `URL` column covers that shape.
@@ -401,6 +402,14 @@ pub const HF_COLUMNS: &[HfColumn] = &[
         column: "text",
         max_rows: Some(10_000),
     },
+    // TinyStories story text: the first 33,121 stories (about 28 MiB), the
+    // size of one file that the NeMo Curator text IO benchmark writes.
+    HfColumn {
+        name: "tinystories/text",
+        url: "hf://datasets/roneneldan/TinyStories@37346026e81060382814d5a773f3c34799e118e7/default/train/0000.parquet",
+        column: "text",
+        max_rows: Some(33_121),
+    },
 ];
 
 /// Reader for a remote parquet file, used only to fill the local input cache.

diff --git a/docs/_ext/vortex_site_files.py b/docs/_ext/vortex_site_files.py
@@ -95,6 +95,8 @@ def _llms(app: Sphinx) -> str:
         "",
         "- [Benchmark summaries](https://bench.vortex.dev/llms.txt): every benchmark group's rankings and methods",
         "- [Vortex homepage](https://vortex.dev/llms.txt): what Vortex is, how to install it, and its results",
+        "- [Questions and answers](https://vortex.dev/faq): how Vortex compares to Parquet and Lance, "
+        "who uses it, and how it is governed",
         "",
         "## Optional",
         "",

diff --git a/docs/_theme/vortex/layout.html b/docs/_theme/vortex/layout.html
@@ -132,6 +132,7 @@
   <img class="vx-footer-mark" src="{{ pathto('_static/lf_white.svg', 1) }}" width="1000" height="60" alt="The Linux Foundation" loading="lazy" />
   <p class="vx-footer-text">Copyright © Vortex a Series of LF Projects, LLC. For web site terms of use, trademark policy and other project policies please see <a class="vx-link" href="https://lfprojects.org/policies">lfprojects.org/policies</a>. <span class="vx-footer-credit">Donated by <a class="vx-link" href="https://spiraldb.com">Spiral</a>.</span></p>
   <nav class="vx-footer-nav" aria-label="{{ _('Project') }}">
+    <a class="vx-label-sm vx-nav-link" href="{{ site }}/faq">{{ _('FAQ') }}</a>
     <a class="vx-label-sm vx-nav-link" href="{{ site }}/charter.pdf">{{ _('Charter') }}</a>
     <a class="vx-label-sm vx-nav-link" href="https://lfprojects.org/policies/code-of-conduct/">{{ _('Code of conduct') }}</a>
     <a class="vx-label-sm vx-nav-link" href="{{ github }}/blob/develop/SECURITY.md">{{ _('Security') }}</a>

diff --git a/docs/api/java/index.rst b/docs/api/java/index.rst
@@ -6,7 +6,7 @@ The Vortex Java API provides bindings for the Vortex library, enabling Java appl
 The API is split into two main components:
 
 * **Vortex JNI**: Core JNI bindings for Vortex functionality
-* **Vortex Spark**: Apache Spark integration for reading Vortex files
+* **Vortex Spark**: Apache Spark integration for reading and writing Vortex files
 
 .. raw:: html
 
@@ -21,7 +21,10 @@ The API is split into two main components:
 Installation
 ------------
 
-The Java API can be included in your project using Gradle or Maven. Please refer to the main documentation for detailed installation instructions.
+The Java API can be included in your project using Gradle or Maven. See the
+:doc:`Java quickstart </getting-started/java>` for the dependency setup and for examples that
+write, read, filter, and project Vortex files, and the :doc:`Spark user guide </user-guide/spark>`
+for the Spark connector.
 
 
 Compatibility
@@ -36,7 +39,7 @@ The Java bindings are supported on the following architectures:
 
 They support any Linux distribution with a GLIBC version >= 2.31. This includes
 
-* Amazon Linux 2022 or newer
+* Amazon Linux 2023 or newer
 * Ubuntu 20.04 or newer
 
 
@@ -61,7 +64,7 @@ Here's a basic example of using the Vortex Java API to read a Vortex file:
     Session session = Session.create();
     DataSource source = DataSource.open(session, "path/to/file.vortex");
 
-    // A scan yields one partition per chunk of the file.
+    // A scan yields partitions; each partition is read as a stream of Arrow batches.
     Scan scan = source.scan(ScanOptions.of());
     while (scan.hasNext()) {
         Partition partition = scan.next();

diff --git a/docs/developer-guide/integrations/spark.md b/docs/developer-guide/integrations/spark.md
@@ -2,7 +2,11 @@
 
 The `vortex-spark` connector is built on Spark's file-source DataSource V2 framework. Shared
 sources compile into Spark 3.5/Scala 2.12, Spark 3.5/Scala 2.13, and Spark 4.0/Scala 2.13
-artifacts; the Spark 4 artifact is also tested on Spark 4.1.
+artifacts; the Spark 4 artifact is also tested on Spark 4.1, 4.2, and 4.3.
+
+The connector uses the public `vortex-jni` API: `Session`, `DataSource`, `Scan`, and
+`VortexWriter`, with the `NativeReadable` and `NativeWritable` I/O interfaces. The
+[Java quickstart](../../getting-started/java.md) describes that API.
 
 ## File-source integration
 
@@ -30,6 +34,18 @@ connector sees the same schemes and credentials as Spark's file index and commit
 Native arrays cross into Spark through the Arrow C Data Interface. Each partition reader owns its
 native scan, Arrow allocator, and exported batches and closes them at task completion.
 
+All readers and writers in a JVM share one native Vortex `Session`, held by
+`VortexSparkSession`. The session lives for the lifetime of the JVM. The `vortex.session.provider`
+option names a `VortexSessionProvider` class, so that the driver and the executors can use a
+custom session.
+
+## Writes
+
+`VortexFileFormat` and `VortexWrite` give Spark a `VortexOutputWriterFactory`, which creates a
+`VortexOutputWriter` for each file that Spark opens. Each writer converts its rows to Arrow batches and writes them to one Vortex file through
+a `VortexWriter` over a `HadoopWritable`. Spark's commit protocol names the files, opens one
+file for each partition directory that a task writes to, and commits or aborts the job.
+
 ## Pushdown and statistics
 
 Schema inference merges every footer, so the dataset schema is the union of the top-level fields its

diff --git a/docs/developer-guide/internals/architecture.md b/docs/developer-guide/internals/architecture.md
@@ -74,12 +74,11 @@ Language bindings expose Vortex to non-Rust environments.
 
 Query engine integrations allow Vortex files to be queried through existing analytics engines.
 
-| Crate / Directory                | Engine     | Notes                                        |
-|----------------------------------| ---------- |----------------------------------------------|
-| `vortex-datafusion/`             | DataFusion | `TableProvider` and `FileFormat` integration |
-| `vortex-duckdb/`                 | DuckDB     | Table function integration                   |
-| `java/vortex-spark/`             | Spark      | Versioned Spark file-source connector via JNI |
-| `java/vortex-trino/`             | Trino      | Trino connector (in development)             |
+| Crate / Directory    | Engine     | Notes                                         |
+|----------------------|------------|-----------------------------------------------|
+| `vortex-datafusion/` | DataFusion | `TableProvider` and `FileFormat` integration  |
+| `vortex-duckdb/`     | DuckDB     | Table function integration                    |
+| `java/vortex-spark/` | Spark      | Versioned Spark file-source connector via JNI |
 
 ## Other Crates
 

diff --git a/docs/developer-guide/language-bindings.md b/docs/developer-guide/language-bindings.md
@@ -67,13 +67,13 @@ layout plugins, and extension DTypes.
 
 ## Per-Language Status
 
-| Language      | Current Tier | Technology |
-|---------------|--------------|------------|
-| Rust          | 3            | Native     |
-| Python        | ~2           | PyO3       |
-| C             | ~1           | cbindgen   |
-| C++           | ~1           | C API      |
-| Java (JNI)    | ~1           | JNI        |
+| Language   | Current Tier | Technology |
+|------------|--------------|------------|
+| Rust       | 3            | Native     |
+| Python     | ~2           | PyO3       |
+| C          | ~1           | cbindgen   |
+| C++        | ~1           | C API      |
+| Java (JNI) | ~1           | JNI        |
 
 ### Rust
 
@@ -98,5 +98,5 @@ inspection, compute execution, and stabilized error handling and memory ownershi
 ### Java (JNI)
 
 Java JNI bindings provide Tier ~1 capabilities today (Arrow I/O and basic scan). JNI will remain
-at Tier 1 for broad JDK compatibility. This is the current integration point for Spark and Trino
-connectors.
+at Tier 1 for broad JDK compatibility. This is the current integration point for the Spark
+connector.