DEVX-13: feat: add per-test timing quality gate to check_test_speed

src/devx/tools/check_test_speed.py

@@ -1,12 +1,21 @@
 #!/usr/bin/env python3
-"""Run unit tests and enforce a maximum execution-time budget.
+"""Run unit tests and enforce execution-time budgets.
+
+Checks two quality gates:
+1. **Total suite time** must not exceed ``--max-seconds``.
+2. **Per-test time** — no individual test may exceed ``--max-single-seconds``.
 
 Usage:
-  python3 -m devx.tools.check_test_speed [--max-seconds N]
+  python3 -m devx.tools.check_test_speed [--max-seconds N] [--max-single-seconds S]
+
+The module runs ``make test-unit`` with ``PYTEST_ADDOPTS=--durations=0`` so
+that pytest emits per-test timing lines alongside the summary.  Both the
+total wall-clock time and individual test durations are parsed and validated.
 """
 
 from __future__ import annotations
 
+import os
 import re
 import subprocess  # nosec B404
 
@@ -14,18 +23,32 @@ import click
 
 from devx.i18n import _
 
-DEFAULT_MAX_SECONDS = 2.0
+DEFAULT_MAX_SECONDS = 10.0
+DEFAULT_MAX_SINGLE_SECONDS = 0.5
 TEST_COMMAND = ["make", "test-unit"]
+
+# Matches pytest summary line: "234 passed in 0.70s"
 _TIMING_RE = re.compile(r"(\d+) passed.* in ([0-9.]+)s")
 
+# Matches per-test duration lines from --durations=0:
+#   0.51s call     tests/test_foo.py::test_bar
+_DURATION_LINE_RE = re.compile(r"^(\d+\.?\d*)s\s+(?:setup|call|teardown)\s+(.+)$")
+
 
 def run_tests() -> tuple[str, str]:
-    """Execute the unit-test suite and return (stdout, stderr)."""
+    """Execute the unit-test suite and return (stdout, stderr).
+
+    Sets ``PYTEST_ADDOPTS=--durations=0`` so pytest emits per-test timings.
+    """
+    env = os.environ.copy()
+    existing = env.get("PYTEST_ADDOPTS", "")
+    env["PYTEST_ADDOPTS"] = f"--durations=0 {existing}".strip()
     result = subprocess.run(  # nosec B603
         TEST_COMMAND,
         capture_output=True,
         text=True,
         check=False,
+        env=env,
     )
     return result.stdout, result.stderr
 
@@ -43,8 +66,23 @@ def parse_duration(output: str) -> float:
     raise click.ClickException(_("Could not parse test execution time from output."))
 
 
+def parse_per_test_durations(output: str) -> list[tuple[str, float]]:
+    """Extract per-test timings from ``--durations=0`` output.
+
+    Returns a list of ``(test_name, seconds)`` tuples sorted by duration
+    (slowest first).
+    """
+    durations: list[tuple[str, float]] = []
+    for line in output.splitlines():
+        match = _DURATION_LINE_RE.match(line.strip())
+        if match:
+            durations.append((match.group(2).strip(), float(match.group(1))))
+    durations.sort(key=lambda x: x[1], reverse=True)
+    return durations
+
+
 def check_speed(duration: float, max_seconds: float) -> None:
-    """Validate duration is within budget; raise on violation."""
+    """Validate total duration is within budget; raise on violation."""
     if duration > max_seconds:
         raise click.ClickException(
             _(
@@ -57,19 +95,58 @@ def check_speed(duration: float, max_seconds: float) -> None:
         )
 
 
-def main(max_seconds: float) -> None:
-    """Run tests, parse timing, and enforce the budget."""
+def check_per_test_speed(
+    durations: list[tuple[str, float]],
+    max_single_seconds: float,
+) -> list[str]:
+    """Return a list of violation messages for tests exceeding the per-test limit.
+
+    An empty list means all tests are within budget.
+    """
+    violations: list[str] = []
+    for name, elapsed in durations:
+        if elapsed > max_single_seconds:
+            violations.append(
+                _(
+                    "Test '{name}' took {elapsed:.2f}s (limit: {limit}s). "
+                    "Optimise: use lighter fixtures, reduce I/O, or mock external calls.",
+                    name=name,
+                    elapsed=elapsed,
+                    limit=max_single_seconds,
+                )
+            )
+    return violations
+
+
+def main(max_seconds: float, max_single_seconds: float) -> None:
+    """Run tests, parse timings, and enforce both budgets."""
     stdout, stderr = run_tests()
     combined = stdout + "\n" + stderr
     click.echo(combined, err=False)
 
     duration = parse_duration(combined)
     check_speed(duration, max_seconds)
+
+    if max_single_seconds > 0:
+        per_test = parse_per_test_durations(combined)
+        violations = check_per_test_speed(per_test, max_single_seconds)
+        if violations:
+            msg = _(
+                "Per-test speed check FAILED: {count} test(s) exceed {limit}s limit.",
+                count=len(violations),
+                limit=max_single_seconds,
+            )
+            click.echo(f"\n{msg}", err=True)
+            for v in violations:
+                click.echo(f"  - {v}", err=True)
+            raise click.ClickException(msg)
+
     click.echo(
         _(
-            "Unit tests passed in {duration:.2f}s (under {max}s limit).",
+            "Unit tests passed in {duration:.2f}s (under {max}s limit, all tests under {single}s per-test limit).",
             duration=duration,
             max=max_seconds,
+            single=max_single_seconds,
         )
     )
 
@@ -80,10 +157,17 @@ def main(max_seconds: float) -> None:
     type=float,
     default=DEFAULT_MAX_SECONDS,
     show_default=True,
-    help="Maximum allowed execution time in seconds.",
+    help="Maximum allowed total execution time in seconds.",
 )
-def cli(max_seconds: float) -> None:
-    main(max_seconds)
+@click.option(
+    "--max-single-seconds",
+    type=float,
+    default=DEFAULT_MAX_SINGLE_SECONDS,
+    show_default=True,
+    help="Maximum allowed per-test time in seconds (0 to disable).",
+)
+def cli(max_seconds: float, max_single_seconds: float) -> None:
+    main(max_seconds, max_single_seconds)
 
 
 if __name__ == "__main__":  # pragma: no cover