100 file variant

.github/workflows/bench-pr.yml

@@ -41,13 +41,6 @@ jobs:
     steps:
       - uses: actions/checkout@v4
       - uses: ./.github/actions/cleanup
-      - name: Configure AWS Credentials for benchmark data
-        uses: aws-actions/configure-aws-credentials@v4
-        if: matrix.benchmark.id == 'clickbench'
-        with:
-          audience: sts.amazonaws.com
-          aws-region: us-west-2
-          role-to-assume: arn:aws:iam::375504701696:role/vortex-benchmark
       - uses: ./.github/actions/setup-rust
       - uses: spiraldb/actions/.github/actions/[email protected]
 
@@ -63,11 +56,6 @@ jobs:
         run: |
           echo "TMPDIR=/work" >> $GITHUB_ENV
 
-      - name: Download Clickbench data
-        if: matrix.benchmark.id == 'clickbench'
-        run:
-          aws s3 cp s3://vortex-bench-dev/clickbench/processed.parquet bench-vortex/data/clickbench/
-
       - name: Run benchmark
         shell: bash
         env:
@@ -94,10 +82,6 @@ jobs:
               | jq --slurp --compact-output '.' >${{ matrix.benchmark.id }}.json
 
           cat ${{ matrix.benchmark.id }}.json
-      - name: Cleanup Raw Data
-        if: always()
-        run:
-          rm -rf bench-vortex/data/
       - name: Store benchmark result
         if: '!cancelled()'
         uses: benchmark-action/github-action-benchmark@v1

bench-vortex/benches/clickbench.rs

@@ -25,7 +25,7 @@ fn benchmark(c: &mut Criterion) {
                     epoch_ms(ClientEventTime * 1000) AS ClientEventTime, \
                     epoch_ms(LocalEventTime * 1000) AS LocalEventTime, \
                         DATE '1970-01-01' + INTERVAL (EventDate) DAYS AS EventDate) \
-                FROM read_parquet('https://datasets.clickhouse.com/hits_compatible/athena_partitioned/hits_{idx}.parquet')) TO '{}' (FORMAT 'parquet');",
+                FROM read_parquet('https://datasets.clickhouse.com/hits_compatible/athena_partitioned/hits_{idx}.parquet', binary_as_string=True)) TO '{}' (FORMAT 'parquet');",
                 output_path.to_str().unwrap()
             );
             Command::new("duckdb")

bench-vortex/src/clickbench.rs

@@ -147,7 +147,7 @@ pub async fn register_vortex_file(
     let vortex_dir = input_path.parent().unwrap().join("vortex_compressed");
     create_dir_all(&vortex_dir).await?;
 
-    for idx in 0..1 {
+    for idx in 0..100 {
         let parquet_file_path = input_path.join(format!("hits_{idx}.parquet"));
         let output_path = vortex_dir.join(format!("hits_{idx}.{VORTEX_FILE_EXTENSION}"));
         idempotent_async(&output_path, |vtx_file| async move {

bench-vortex/src/tpch/mod.rs

@@ -26,7 +26,6 @@ use vortex_datafusion::memory::VortexMemTableOptions;
 use vortex_datafusion::persistent::format::VortexFormat;
 use vortex_datafusion::SessionContextExt;
 
-use crate::clickbench::HITS_SCHEMA;
 use crate::{idempotent_async, CTX, TARGET_BLOCK_BYTESIZE, TARGET_BLOCK_SIZE};
 
 pub mod dbgen;
@@ -337,9 +336,8 @@ async fn register_vortex_file(
     let table_url = ListingTableUrl::parse(vtx_file.to_str().unwrap())?;
     let config = ListingTableConfig::new(table_url)
         .with_listing_options(ListingOptions::new(format as _))
-        .with_schema(HITS_SCHEMA.clone().into());
-    // .infer_schema(&session.state())
-    // .await?;
+        .infer_schema(&session.state())
+        .await?;
 
     let listing_table = Arc::new(ListingTable::try_new(config)?);