Apache DataFusion SQL Query Engine
ci: reintroduce code coverage reporting with cargo-llvm-cov (#23336)
Burak Şen80bc64f3d225
diff --git a/.github/workflows/rust.yml b/.github/workflows/rust.ymlindex 94a583f09e..d69f6a69c1 100644--- a/.github/workflows/rust.yml+++ b/.github/workflows/rust.yml@@ -306,16 +306,22 @@ jobs:uses: ./.github/actions/setup-builderwith:rust-version: stable+ - name: Install llvm-tools-preview+ run: rustup component add llvm-tools-preview+ - name: Install cargo-llvm-cov+ uses: taiki-e/install-action@9bcaee1dcae34154180f412e2fa69355a7cda9f6 # v2.82.6+ with:+ tool: cargo-llvm-cov- name: Rust Dependency Cacheuses: Swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 # v2.9.1with:- save-if: ${{ github.ref_name == 'main' }}- shared-key: "amd-ci"+ save-if: ${{ github.ref_name == 'main' }}+ shared-key: "amd-ci"- name: Run tests (excluding doctests and datafusion-cli)env:RUST_BACKTRACE: 1run: |- cargo test \+ cargo llvm-cov \--profile ci \--exclude datafusion-examples \--exclude ffi_example_table_provider \@@ -324,7 +330,9 @@ jobs:--lib \--tests \--bins \- --features serde,avro,json,backtrace,integration-tests,parquet_encryption,substrait+ --features serde,avro,json,backtrace,integration-tests,parquet_encryption,substrait \+ --codecov \+ --output-path target/codecov.json- parallel:- name: Verify Working Directory Cleanrun: git diff --exit-code@@ -337,6 +345,12 @@ jobs:echo "$STATUS"exit 1fi+ - name: Upload coverage to codecov.io+ uses: codecov/codecov-action@0fb7174895f61a3b6b78fc075e0cd60383518dac # v5+ with:+ files: target/codecov.json+ fail_ci_if_error: false+ token: ${{ secrets.CODECOV_TOKEN }}# datafusion-cli testslinux-test-datafusion-cli:@@ -639,50 +653,6 @@ jobs:- name: Check workflow tool installsrun: ci/scripts/check_no_cargo_install_in_workflows.sh- # Coverage job disabled due to- # https://github.com/apache/datafusion/issues/3678-- # coverage:- # name: coverage- # runs-on: ubuntu-latest- # steps:- # - uses: actions/checkout@v4- # with:- # submodules: true- # - name: Install protobuf compiler- # shell: bash- # run: |- # mkdir -p $HOME/d/protoc- # cd $HOME/d/protoc- # export PROTO_ZIP="protoc-21.4-linux-x86_64.zip"- # curl -LO https://github.com/protocolbuffers/protobuf/releases/download/v21.4/$PROTO_ZIP- # unzip $PROTO_ZIP- # export PATH=$PATH:$HOME/d/protoc/bin- # protoc --version- # - name: Setup Rust toolchain- # run: |- # rustup toolchain install stable- # rustup default stable- # rustup component add rustfmt clippy- # - name: Cache Cargo- # uses: actions/cache@v4- # with:- # path: /home/runner/.cargo- # # this key is not equal because the user is different than on a container (runner vs github)- # key: cargo-coverage-cache3-- # - name: Install cargo-tarpaulin- # uses: taiki-e/install-action@43aecc8d72668fbcfe75c31400bc4f890f1c5853 # v2.83.2- # with:- # tool: [email protected]- # - name: Run coverage- # run: |- # export PATH=$PATH:$HOME/d/protoc/bin- # rustup toolchain install stable- # rustup default stable- # cargo tarpaulin --all --out Xml- # - name: Report coverage- # continue-on-error: true- # run: bash <(curl -s https://codecov.io/bash)clippy:name: clippydiff --git a/Cargo.toml b/Cargo.tomlindex 72676c8263..fef23162ff 100644--- a/Cargo.toml+++ b/Cargo.toml@@ -226,8 +226,8 @@ assigning_clones = "warn"[workspace.lints.rust]unexpected_cfgs = { level = "warn", check-cfg = ['cfg(datafusion_coop, values("tokio", "tokio_fallback", "per_stream"))',- "cfg(tarpaulin)",- "cfg(tarpaulin_include)",+ "cfg(coverage)",+ "cfg(coverage_nightly)",] }unused_qualifications = "deny"diff --git a/datafusion/core/tests/fuzz_cases/sort_fuzz.rs b/datafusion/core/tests/fuzz_cases/sort_fuzz.rsindex 0d8a066d43..675854ddb5 100644--- a/datafusion/core/tests/fuzz_cases/sort_fuzz.rs+++ b/datafusion/core/tests/fuzz_cases/sort_fuzz.rs@@ -40,7 +40,6 @@ use test_utils::{batches_to_vec, partitions_to_sorted_vec};const KB: usize = 1 << 10;#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]async fn test_sort_10k_mem() {for (batch_size, should_spill) in [(5, false), (20000, true), (500000, true)] {let (input, collected) = SortTest::new()@@ -58,7 +57,6 @@ async fn test_sort_10k_mem() {}#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]async fn test_sort_100k_mem() {for (batch_size, should_spill) in[(5, false), (10000, false), (20000, true), (1000000, true)]@@ -78,7 +76,6 @@ async fn test_sort_100k_mem() {}#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]async fn test_sort_strings_100k_mem() {for (batch_size, should_spill) in[(5, false), (1000, false), (10000, true), (20000, true)]@@ -116,7 +113,6 @@ async fn test_sort_strings_100k_mem() {}#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]async fn test_sort_multi_columns_100k_mem() {for (batch_size, should_spill) in[(5, false), (1000, false), (10000, true), (20000, true)]diff --git a/datafusion/core/tests/sql/explain_analyze.rs b/datafusion/core/tests/sql/explain_analyze.rsindex 2293098bb8..4c8b8f9c01 100644--- a/datafusion/core/tests/sql/explain_analyze.rs+++ b/datafusion/core/tests/sql/explain_analyze.rs@@ -827,7 +827,7 @@ async fn test_physical_plan_display_indent_multi_children() {}#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]+#[cfg_attr(coverage, ignore)]async fn csv_explain_analyze() {// This test uses the execute function to run an actual plan under EXPLAIN ANALYZElet ctx = SessionContext::new();@@ -849,7 +849,7 @@ async fn csv_explain_analyze() {}#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]+#[cfg_attr(coverage, ignore)]async fn csv_explain_analyze_order_by() {let ctx = SessionContext::new();register_aggregate_csv_by_sql(&ctx).await;@@ -866,7 +866,7 @@ async fn csv_explain_analyze_order_by() {}#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]+#[cfg_attr(coverage, ignore)]async fn parquet_explain_analyze() {let ctx = SessionContext::new();register_alltypes_parquet(&ctx).await;@@ -913,7 +913,7 @@ async fn parquet_explain_analyze() {// (e.g. nested/recursive expansion causing full schema to be scanned).// Keeping this test ensures we don't regress that behavior.#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]+#[cfg_attr(coverage, ignore)]async fn parquet_recursive_projection_pushdown() -> Result<()> {use parquet::arrow::arrow_writer::ArrowWriter;use parquet::file::properties::WriterProperties;@@ -1030,7 +1030,7 @@ async fn parquet_recursive_projection_pushdown() -> Result<()> {}#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]+#[cfg_attr(coverage, ignore)]async fn parquet_explain_analyze_verbose() {let ctx = SessionContext::new();register_alltypes_parquet(&ctx).await;@@ -1047,7 +1047,7 @@ async fn parquet_explain_analyze_verbose() {}#[tokio::test]-#[cfg_attr(tarpaulin, ignore)]+#[cfg_attr(coverage, ignore)]async fn csv_explain_analyze_verbose() {// This test uses the execute function to run an actual plan under EXPLAIN VERBOSE ANALYZElet ctx = SessionContext::new();diff --git a/datafusion/ffi/Cargo.toml b/datafusion/ffi/Cargo.tomlindex e50530c868..affcff3dbd 100644--- a/datafusion/ffi/Cargo.toml+++ b/datafusion/ffi/Cargo.toml@@ -91,4 +91,3 @@ integration-tests = ["datafusion-functions-window",]parquet = ["datafusion-proto/parquet"]-tarpaulin_include = [] # Exists only to prevent warnings on stable and still have accurate coveragediff --git a/datafusion/ffi/src/arrow_wrappers.rs b/datafusion/ffi/src/arrow_wrappers.rsindex 1c921b0f83..62fb36f836 100644--- a/datafusion/ffi/src/arrow_wrappers.rs+++ b/datafusion/ffi/src/arrow_wrappers.rs@@ -49,7 +49,6 @@ impl From<SchemaRef> for WrappedSchema {/// Since going through the FFI always has the potential to fail, we need to catch these errors,/// give the user a warning, and return some kind of result. In this case we default to an/// empty schema.-#[cfg(not(tarpaulin_include))]fn catch_df_schema_error(e: &ArrowError) -> Schema {error!("Unable to convert from FFI_ArrowSchema to DataFusion Schema in FFI_PlanProperties. {e}"diff --git a/datafusion/ffi/src/tests/utils.rs b/datafusion/ffi/src/tests/utils.rsindex e1374c7862..b6b50cbce8 100644--- a/datafusion/ffi/src/tests/utils.rs+++ b/datafusion/ffi/src/tests/utils.rs@@ -21,29 +21,6 @@ use datafusion_common::{DataFusionError, Result};use crate::tests::ForeignLibraryModule;-/// Compute the path to the built cdylib. Checks debug, release, and ci profile dirs.-fn compute_library_dir(target_path: &Path) -> PathBuf {- let debug_dir = target_path.join("debug");- let release_dir = target_path.join("release");- let ci_dir = target_path.join("ci");-- let all_dirs = vec![debug_dir.clone(), release_dir, ci_dir];-- all_dirs- .into_iter()- .filter(|dir| dir.join("deps").exists())- .filter_map(|dir| {- dir.join("deps")- .metadata()- .and_then(|m| m.modified())- .ok()- .map(|date| (dir, date))- })- .max_by_key(|(_, date)| *date)- .map(|(dir, _)| dir)- .unwrap_or(debug_dir)-}-/// Find the cdylib file for datafusion_ffi in the given directory.fn find_cdylib(deps_dir: &Path) -> Result<PathBuf> {let lib_prefix = if cfg!(target_os = "windows") {@@ -71,19 +48,24 @@ fn find_cdylib(deps_dir: &Path) -> Result<PathBuf> {))}+/// Locate the built `datafusion_ffi` cdylib.+///+/// The cdylib sits next to the running test binary, so this follows Cargo's+/// actual output directory and is robust to the active profile and a custom+/// `--target-dir` (e.g. `cargo llvm-cov`).+fn find_library() -> Result<PathBuf> {+ let exe =+ std::env::current_exe().map_err(|e| DataFusionError::External(Box::new(e)))?;+ let deps_dir = exe.parent().ok_or_else(|| {+ DataFusionError::External("Failed to find test binary directory".into())+ })?;+ find_cdylib(deps_dir)+}+pub fn get_module() -> Result<ForeignLibraryModule> {let expected_version = crate::version();- let crate_root = Path::new(env!("CARGO_MANIFEST_DIR"));- let target_dir = crate_root- .parent()- .expect("Failed to find crate parent")- .parent()- .expect("Failed to find workspace root")- .join("target");-- let library_dir = compute_library_dir(target_dir.as_path());- let lib_path = find_cdylib(&library_dir.join("deps"))?;+ let lib_path = find_library()?;// Load the library using libloadinglet lib = unsafe {