feat: add CodSpeed to the project

Author: adriencaccia

.github/workflows/ibis-benchmarks.yml

@@ -1,20 +1,12 @@
 name: Benchmarks
-
 on:
   push:
     branches:
-      - main
-      - "*.x.x"
-  merge_group:
-
-# since we're writing to cloud storage, we don't want to have multiple
-# instances of this job running at one time
-concurrency: benchmarks-${{ github.repository }}
-
-permissions:
-  # increase the rate limit for github operations, but limit token permissions
-  # to read-only
-  contents: read
+      - "main" # or "master"
+  pull_request:
+  # `workflow_dispatch` allows CodSpeed to trigger backtest
+  # performance analysis in order to generate initial data.
+  workflow_dispatch:
 
 jobs:
   benchmarks:
@@ -38,47 +30,8 @@ jobs:
       - name: install ibis
         run: poetry install --without dev --without docs --all-extras
 
-      - name: make benchmark output dir
-        run: mkdir .benchmarks
-
-      - name: benchmark
-        run: poetry run pytest --benchmark-enable --benchmark-json .benchmarks/output.json ibis/tests/benchmarks
-
-      - uses: google-github-actions/auth@v2
+      - name: Run benchmarks
+        uses: CodSpeedHQ/action@v2
         with:
-          credentials_json: ${{ secrets.GCP_CREDENTIALS }}
-
-      - uses: google-github-actions/setup-gcloud@v2
-
-      - name: show gcloud info
-        run: gcloud info
-
-      - name: download the latest duckdb release
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          set -euo pipefail
-
-          gh release download -R duckdb/duckdb --pattern 'duckdb_cli-linux-amd64.zip'
-          unzip duckdb_cli-linux-amd64.zip
-
-      - name: convert json data to parquet
-        run: |
-          set -euo pipefail
-
-          # sort json keys
-          jq --sort-keys -rcM < "$PWD/.benchmarks/output.json" > output.json
-
-          # connect to a file to allow spilling to disk
-          ./duckdb json2parquet.ddb <<EOF
-            COPY (
-              SELECT * FROM read_ndjson_auto('output.json', maximum_object_size=2**27)
-            ) TO 'output.parquet' (FORMAT PARQUET, COMPRESSION ZSTD)
-          EOF
-
-      - name: copy data to gcs
-        run: |
-          set -euo pipefail
-
-          timestamp="$(date --iso-8601=ns --utc | tr ','  '.')"
-          gsutil cp output.parquet "gs://ibis-benchmark-data/ci/${timestamp}.parquet"
+          token: ${{ secrets.CODSPEED_TOKEN }}
+          run: poetry run pytest --benchmark-enable ibis/tests/benchmarks