Add TPC-DS SLT tests with plans and Parquet parity

vortex-sqllogictest/README.md

@@ -10,9 +10,9 @@ sequentially for one engine. Each test is named `slt::<engine>::<relative-path>`
 
 ## Running tests
 
-Some tests use TPC-H data at scale factor 0.1 and one shard of the partitioned ClickBench
-dataset (about one million rows of `hits`). Generate both first, then run the suite with
-`cargo nextest`:
+Some tests use TPC-H and TPC-DS data at scale factor 0.1 and one shard of the partitioned
+ClickBench dataset (about one million rows of `hits`). Generate them first, then run the suite
+with `cargo nextest`:
 
 ```shell
 ./vortex-sqllogictest/slt/generate_data.sh
@@ -22,35 +22,39 @@ cargo test -p vortex-sqllogictest --test sqllogictests
 ```
 
 `generate_data.sh` accepts dataset names to generate only some of the fixtures, for example
-`./vortex-sqllogictest/slt/generate_data.sh tpch` or `... clickbench`. Run it with `--help` to list
-the datasets. It runs each dataset's own script, `slt/tpch/generate_data.sh` or
-`slt/clickbench/generate_data.sh`, which can also be run directly.
-
-The generated Vortex and Parquet data lives under `slt/tpch/data/` and `slt/clickbench/data/`
-(git-ignored). Both formats are required; regenerate older fixtures if they only contain Vortex
-files. If either format is missing, that suite's tests are reported as **ignored**, so the rest of
-the suite still runs. These `.slt` files load their tables through paths relative to the crate
-root, so run the tests via `cargo nextest`/`cargo test`, which set the working directory
-accordingly.
-
-TPC-H scripts live under `slt/tpch/datafusion/` and `slt/tpch/duckdb/`, ClickBench scripts under
+`./vortex-sqllogictest/slt/generate_data.sh tpch`, `... tpcds` or `... clickbench`. Run it with
+`--help` to list the datasets. It runs each dataset's own script, `slt/tpch/generate_data.sh`,
+`slt/tpcds/generate_data.sh` or `slt/clickbench/generate_data.sh`, which can also be run directly.
+TPC-DS is generated by DuckDB's `tpcds` extension through `uvx`.
+
+The generated Vortex and Parquet data lives under `slt/tpch/data/`, `slt/tpcds/data/` and
+`slt/clickbench/data/` (git-ignored). Both formats are required; regenerate older fixtures if they
+only contain Vortex files. If either format is missing, that suite's tests are reported as
+**ignored**, so the rest of the suite still runs. These `.slt` files load their tables through
+paths relative to the crate root, so run the tests via `cargo nextest`/`cargo test`, which set the
+working directory accordingly.
+
+TPC-H scripts live under `slt/tpch/datafusion/` and `slt/tpch/duckdb/`, TPC-DS scripts under
+`slt/tpcds/datafusion/` and `slt/tpcds/duckdb/`, ClickBench scripts under
 `slt/clickbench/datafusion/` and `slt/clickbench/duckdb/`. Each engine has its own
-`create.slt.no`, `results/q*.slt.no` (`q1` to `q22` for TPC-H, `q0` to `q42` for ClickBench,
-matching the upstream numbering), and `drop.slt.no`. Its `tpch.slt`/`clickbench.slt` runs these
-against Vortex and asserts EXPLAIN output from the matching `plans/q*.slt.no`. Its `parquet.slt`
-runs the same queries against the original Parquet fixtures and checks the same expected results.
+`create.slt.no`, `results/q*.slt.no` (`q1` to `q22` for TPC-H, `q1` to `q99` for TPC-DS, `q0` to
+`q42` for ClickBench, matching the upstream numbering), and `drop.slt.no`. Its
+`tpch.slt`/`tpcds.slt`/`clickbench.slt` runs these against Vortex and asserts EXPLAIN output from
+the matching `plans/q*.slt.no`. Its `parquet.slt` runs the same queries against the original
+Parquet fixtures and checks the same expected results.
 DataFusion uses external tables; DuckDB uses views over files. The `FILE_FORMAT` substitution
 variable selects the format in each engine's table setup.
 
 ClickBench plans explain the upstream queries unchanged. Where an upstream query leaves the order
 of tied rows unspecified, its result record adds tie-breaking `ORDER BY` columns so both formats
-and repeated runs produce the same rows. The ClickBench generator also runs
-`slt/clickbench/duckdb/parity.slt` right after converting the shard; it reads both files through
-DuckDB and fails if the Parquet and Vortex data differ. After completing ClickBench DataFusion
-plans, replace the byte ranges in `file_groups` with `<slt:ignore>`, as the TPC-H plans do:
+and repeated runs produce the same rows. The ClickBench and TPC-DS generators also run
+`slt/clickbench/duckdb/parity.slt` and `slt/tpcds/duckdb/parity.slt` right after converting their
+fixtures; they read both formats through DuckDB and fail if the Parquet and Vortex data differ.
+After completing ClickBench or TPC-DS DataFusion plans, replace the byte ranges in `file_groups`
+with `<slt:ignore>`, as the TPC-H plans do:
 
 ```shell
-sed -i -E 's/hits\.vortex:[0-9]+\.\.[0-9]+/hits.vortex:<slt:ignore>/g' vortex-sqllogictest/slt/clickbench/datafusion/plans/*.slt.no
+sed -i -E 's/\.vortex:[0-9]+\.\.[0-9]+/.vortex:<slt:ignore>/g' vortex-sqllogictest/slt/{clickbench,tpcds}/datafusion/plans/*.slt.no
 ```
 
 Because the harness is `libtest-mimic`-based, the standard test flags work, including
@@ -123,7 +127,7 @@ engine currently produces, instead of comparing against it. This is useful after
 change to query results or plan formatting.
 
 ```shell
-# Complete every file (generate TPC-H and ClickBench data first if you want their result files updated):
+# Complete every file (generate the TPC-H, TPC-DS and ClickBench data first if you want their result files updated):
 cargo test -p vortex-sqllogictest --test sqllogictests -- --complete
 # Complete only the files whose name matches a substring:
 cargo test -p vortex-sqllogictest --test sqllogictests -- --complete strings

vortex-sqllogictest/bin/sqllogictests-runner.rs

@@ -170,6 +170,7 @@ fn engines_for(path: &Path) -> (bool, bool) {
 /// Vortex and Parquet versions both have to exist for the suite to run.
 const GENERATED_DATASETS: &[(&str, &str)] = &[
     ("tpch", "tpch/data/lineitem"),
+    ("tpcds", "tpcds/data/store_sales"),
     ("clickbench", "clickbench/data/hits"),
 ];
 
@@ -242,9 +243,9 @@ fn main() -> anyhow::Result<ExitCode> {
     let mut trials = Vec::new();
     for path in files {
         let (run_datafusion, run_duckdb) = engines_for(&path);
-        // TPC-H and ClickBench trials are ignored (rather than removed) when the
-        // generated data is absent, so `--list` and the run summary still
-        // account for them.
+        // Generated-data trials (TPC-H, TPC-DS, ClickBench) are ignored (rather
+        // than removed) when the data is absent, so `--list` and the run summary
+        // still account for them.
         let ignored = missing_generated_data(&path);
         let name = path
             .strip_prefix(SLT_ROOT.as_path())

vortex-sqllogictest/slt/generate_data.sh

@@ -10,14 +10,15 @@
 #
 # Datasets:
 #   tpch        TPC-H at scale factor 0.1.
+#   tpcds       TPC-DS at scale factor 0.1 from DuckDB's tpcds extension.
 #   clickbench  One shard (~1M rows) of the partitioned ClickBench `hits` table.
 #
 # With no arguments every dataset is generated.
 
 set -e -o pipefail
 
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-ALL_DATASETS=(tpch clickbench)
+ALL_DATASETS=(tpch tpcds clickbench)
 
 usage() {
   echo "Usage: $(basename "${BASH_SOURCE[0]}") [DATASET...]"
@@ -32,7 +33,7 @@ for arg in "$@"; do
       usage
       exit 0
       ;;
-    tpch|clickbench)
+    tpch|tpcds|clickbench)
       datasets+=("${arg}")
       ;;
     *)

vortex-sqllogictest/slt/tpcds/.gitignore

@@ -0,0 +1 @@
+/data

vortex-sqllogictest/slt/tpcds/datafusion/create.slt.no

@@ -0,0 +1,127 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright the Vortex contributors
+
+# Keep fixture paths under WORK_DIR so EXPLAIN output can normalize them.
+# Plans ignore byte ranges because they depend on the generated fixture sizes.
+system ok
+cp slt/tpcds/data/*.$FILE_FORMAT "$WORK_DIR/"
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS call_center
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/call_center.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS catalog_page
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/catalog_page.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS catalog_returns
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/catalog_returns.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS catalog_sales
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/catalog_sales.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS customer
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/customer.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS customer_address
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/customer_address.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS customer_demographics
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/customer_demographics.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS date_dim
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/date_dim.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS household_demographics
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/household_demographics.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS income_band
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/income_band.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS inventory
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/inventory.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS item
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/item.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS promotion
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/promotion.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS reason
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/reason.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS ship_mode
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/ship_mode.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS store
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/store.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS store_returns
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/store_returns.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS store_sales
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/store_sales.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS time_dim
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/time_dim.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS warehouse
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/warehouse.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS web_page
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/web_page.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS web_returns
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/web_returns.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS web_sales
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/web_sales.${FILE_FORMAT}';
+
+statement ok
+CREATE EXTERNAL TABLE IF NOT EXISTS web_site
+STORED AS ${FILE_FORMAT}
+LOCATION '${WORK_DIR}/web_site.${FILE_FORMAT}';

vortex-sqllogictest/slt/tpcds/datafusion/drop.slt.no

@@ -0,0 +1,74 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright the Vortex contributors
+
+statement ok
+DROP TABLE web_site;
+
+statement ok
+DROP TABLE web_sales;
+
+statement ok
+DROP TABLE web_returns;
+
+statement ok
+DROP TABLE web_page;
+
+statement ok
+DROP TABLE warehouse;
+
+statement ok
+DROP TABLE time_dim;
+
+statement ok
+DROP TABLE store_sales;
+
+statement ok
+DROP TABLE store_returns;
+
+statement ok
+DROP TABLE store;
+
+statement ok
+DROP TABLE ship_mode;
+
+statement ok
+DROP TABLE reason;
+
+statement ok
+DROP TABLE promotion;
+
+statement ok
+DROP TABLE item;
+
+statement ok
+DROP TABLE inventory;
+
+statement ok
+DROP TABLE income_band;
+
+statement ok
+DROP TABLE household_demographics;
+
+statement ok
+DROP TABLE date_dim;
+
+statement ok
+DROP TABLE customer_demographics;
+
+statement ok
+DROP TABLE customer_address;
+
+statement ok
+DROP TABLE customer;
+
+statement ok
+DROP TABLE catalog_sales;
+
+statement ok
+DROP TABLE catalog_returns;
+
+statement ok
+DROP TABLE catalog_page;
+
+statement ok
+DROP TABLE call_center;

vortex-sqllogictest/slt/tpcds/datafusion/parquet.slt

@@ -0,0 +1,11 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright the Vortex contributors
+
+include ../../setup.slt.no
+
+let FILE_FORMAT
+SELECT 'parquet';
+
+include ./create.slt.no
+include ./results/*.slt.no
+include ./drop.slt.no