(performance): speedup for `zarr`-based sparse indexing #2641

	name: Benchmark

	on:
	push:
	branches: [main, "[0-9]+.[0-9]+.x"]
	pull_request:
	branches: [main]

	jobs:
	benchmark:
	runs-on: ${{ matrix.os }}
	defaults:
	run:
	shell: bash -el {0} # -e to fail on error, -l for mamba

	strategy:
	fail-fast: false
	matrix:
	python: ["3.12"]
	os: [ubuntu-latest]

	env:
	OS: ${{ matrix.os }}
	PYTHON: ${{ matrix.python }}
	ASV_DIR: "./benchmarks"

	steps:
	- uses: actions/checkout@v4
	with:
	fetch-depth: 0

	- run: git fetch origin main:main
	if: ${{ github.ref_name != 'main' }}
	# Errors on main branch

	- uses: mamba-org/setup-micromamba@v2
	with:
	environment-name: asv
	cache-environment: true
	# Deps documented in https://asv.readthedocs.io/en/latest/installing.html
	# libmambapy upper bound: https://github.com/airspeed-velocity/asv/issues/1438
	create-args: >-
	python=${{ matrix.python }}
	asv
	libmambapy<2
	conda-build

	- name: Cache datasets
	uses: actions/cache@v4
	with:
	path: \|
	~/.cache
	key: benchmark-state-${{ hashFiles('benchmarks/**') }}

	- name: Quick benchmark run
	working-directory: ${{ env.ASV_DIR }}
	run: \|
	asv machine --yes
	asv run --quick --show-stderr --verbose

Provide feedback