Testing

Tests are written in the SQLLogicTest format under test/sql/:

make debug && make test    # make test does not rebuild; run make debug after editing Rust

They come in two kinds, and the split is deliberate:

File	Covers	Extra dependency
test/sql/quantstats/html_reports.test	Return-path behaviour: the registration surface (two names, one signature each), the result shape and its ordering (symbols ascending, benchmarks in list order inside one symbol), benchmark NULL with none configured, display names falling back to the symbol, NULL rows skipped and instruments dropped with them, empty input → NULL, multi-threaded combine consistency (single vs 4 threads, md5-equal), output_dir auto-naming and path filling, two calls never overwriting each other	none
test/sql/quantstats/html_reports_by_prices.test	Price-path behaviour: byte-identical to lag()-derived returns, each of several benchmarks byte-identical to its own differenced reference, the benchmark side differenced too, points with a zero predecessor skipped, single-point instruments dropped, benchmark symbol missing / too few points	none
test/sql/quantstats/html_reports_errors.test	Error paths: the four ways a benchmark list can be wrong (symbol missing / empty string / NULL element / duplicate), the list disagreeing across symbols (order included), extra benchmark_title entries being ignored (no error), periods_per_year = 0, output_dir = '', a NUL path, open_in_browser with a non-local path, an uncast options literal, the old API being gone	none
test/sql/quantstats/html_reports_values.test	Output content: parses the generated HTML with webbed's XPath and asserts the title, the date range, the rf echo, per-row metric numbers, the chart/table counts, the extra benchmark column, each report carrying its own benchmark column when there are several, benchmark display names paired by index (missing / NULL / empty falling back to the symbol), and each symbol's own title and file name	the webbed community extension

webbed is installed from inside the test file (INSTALL webbed FROM community;), which needs network on the first run and then goes through the local DuckDB extension cache. Delete that file if you do not want the dependency; the other files are unaffected.

The fast loop

DuckDB's test runner can drive the artifact directly, which skips make entirely. The recipe needs a Python environment with duckdb_sqllogictest installed — make configure creates one under configure/venv, or you can use any Python 3 environment that has it:

# Linux / macOS
./configure/venv/bin/python -m duckdb_sqllogictest \
    --test-dir test/sql \
    --external-extension target/debug/duckfn_quantstats.duckdb_extension

# Windows
.\configure\venv\Scripts\python.exe -m duckdb_sqllogictest `
    --test-dir test/sql `
    --external-extension target/debug/duckfn_quantstats.duckdb_extension

--test-dir is required: it is also the value of __TEST_DIR__, the directory a test writing files is given. To run a single file, add --file-path test/sql/quantstats/html_reports.test.

Asserting on the HTML

Asserting on long HTML via XPath beats length(...) > N and localises failures far better than an md5 comparison:

-- Report title and date range
SELECT html_extract_text(html, '//h1')[1] FROM report;
-- -> My Fund 2 Jan, 2024 - 12 Jan, 2024

-- The strategy column of one row of the metrics table (benchmark column comes first)
SELECT html_extract_text(html, '//div[@id="right"]/table[1]//tr[td[1]="Sharpe"]/td[2]')[1] FROM report;
-- -> 4.93

Note that webbed's html_extract_text(html, xpath) returns VARCHAR[] (all matches): index with [1] for a single value, or use array_to_string(..., ' | ') / array_length(...) to assert a whole group.

Conventions

• Every file starts from a clean database, and a file that uses the extension starts with require duckfn_quantstats.
• Expected errors are matched as substrings. Under statement error, a distinctive fragment of the message is enough — there is no need to reproduce DuckDB's whole error string, and doing so ties the test to a message that may well change.
• Watch how values print. A DOUBLE renders as 7.0; when a test is about a number rather than a type, cast it so the expectation stays stable if the type changes.
• Divide the files by concern, not by function count: behaviour in one file, error paths in another, and a .test that reaches for a community extension kept separate, because it needs the network the first time.

When adding a function, cover at least: normal values, NULL, boundary values and error paths (statement error). Before committing: just lint (cargo clippy --all-targets -- -D warnings).