stackchain/troupe-standup
2
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

beam: Q85 — lane CHECK grades module coverage; zeller_dow cell pins calendar branch

Browse Source
This commit is contained in:
Sentinel authored and ledger committed 2026-11-02 07:26:14 +00:00
1 parent e3f4875709
commit c05f05a87f
5 files changed
+646
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

wizard-work/beam/CHECK
+22
View File
@@ -35,6 +35,16 @@
# (tools/import_resolve_check.py) in front of THIS lane's test files, scoped to
# the lane so another lane's fault does not redden this gate;
# `tools/verify.py` still grades the whole tree.
#
# Runner note (Q85, 2026-11-02). Every cell above can be green while a lane
# module is read by none of them: `083981e` added zeller_dow.py, 38 lines, and
# this gate printed `Ran 214` before and after. `check_coverage.py` states the
# rule once -- a module must be imported by a collected test, named by a path
# literal in one, or run by this gate -- and fails with the names of the
# modules that are not. Six files in this lane are one-off artifacts rather
# than primitives, and each carries a written reason in that file's EXEMPT
# table; a waiver with no reason, or one left standing over a module that is
# now covered, is itself a fault, so the table cannot rot into a list of names.
set -e
set -o pipefail
cd "$(dirname "$(readlink -f "$0")")"
@@ -63,4 +73,16 @@ case $arc in
*) echo "CHECK: check_bindings.py exited $arc -- internal error." >&2; exit 1 ;;
esac
set +e
python3 check_coverage.py
crc=$?
set -e
case $crc in
0) ;;
1) echo "CHECK: a lane module is read by no collected test (Q85) -- see the names above." >&2; exit 1 ;;
2) echo "CHECK: coverage could not be graded -- see the message above." >&2; exit 2 ;;
127) echo "CHECK: python3 not found -- cannot grade." >&2; exit 2 ;;
*) echo "CHECK: check_coverage.py exited $crc -- internal error." >&2; exit 1 ;;
esac
python3 toposort/test_toposort.py
wizard-work/beam/check_coverage.py
+461
View File
@@ -0,0 +1,461 @@
#!/usr/bin/env python3
"""check_coverage.py -- a lane module no collected cell reads is invisible (Q85).
THE DEFECT THIS GRADES
`beam/CHECK` runs every `test_*.py` in this lane through `run_lane_tests.py`.
A primitive module that no collected test imports, and that no collected test
runs as a script, is never executed by the gate: it can be deleted, or land
broken, and the gate's count does not move. Q85 measured the shape three
times over (`083981e` added `zeller_dow.py`, 38 lines, and `CHECK` read
`Ran 214` before and after). Per-file items do not converge -- every new
module buys another item -- so this leg states the rule once, for the lane.
THE RULE (stated, not implied)
A lane module is READ BY THE GATE when any of these holds:
test-import a collected test file imports it -- directly, or through
another lane module's import chain (importing A executes
A's imports, so B is read too);
test-literal a collected test file names it in a string literal
(the `SCRIPT = os.path.join(HERE, "shadow_lint.py")` and
`from csv_rfc4180 import ...` cases: a module used as a
subprocess, or under a different test file's name);
gate-invoked `CHECK` itself runs it (`python3 <stem>.py`), i.e. the
gate executes it on every tick.
"No file named `test_<module>.py`" is NOT the fault -- `csv_rfc4180` has no
such file and is covered. "Read by nothing the gate runs" is.
Anything else is a fault, unless it is waived in EXEMPT below WITH A REASON.
A waiver is a sentence a reader can check against the file it waives, not a
silence: an empty reason, a waiver naming a file that is gone, and a waiver
naming a module that is now covered are each their own fault, so this table
cannot rot into a list of names nobody re-reads.
EXIT CODES (the repo's tools/ contract)
0 every lane module is covered, or waived with a reason
1 a graded fault: module(s) read by nothing (named), or a bad waiver
2 could not grade: no lane directory, no collected tests, a test file that
will not parse (coverage cannot be read out of a file that will not parse)
USAGE
python3 check_coverage.py [--lane DIR] [--json OUT] [--list] [--selftest]
"""
import argparse
import ast
import glob
import json
import os
import re
import sys
DEFAULT_LANE = os.path.dirname(os.path.abspath(__file__))
# Modules read by nothing the gate runs, each with the reason. Every reason
# names what the file is and why hand-running it, rather than a collected
# cell, is the right reading. Keep this list short: it is a record of
# exceptions, not a coverage plan.
EXEMPT = {
"verify_toposort_none.py":
"one-off artifact, not a standing check: its own header (lines 1-4) records "
"that no committed gate invokes it, and it asserts toposort's None-dep and "
"one-shot cases by hand with SystemExit on failure. Its subject, toposort.py, "
"is covered by test_toposort.py and the toposort/ script; this file is the "
"hand-run verification transcript for Q61/Q63, kept as evidence.",
"verify_toposort_gen.py":
"one-off artifact for the same Q61 defect (one-shot iterators materialised "
"exactly once, with a consumption counter). Hand-run against toposort.py, "
"which the collected suite covers; this file is the verification record.",
"verify_wordcount_test.py":
"one-off artifact for wordcount.py, which is not a module of this lane "
"(it writes /root/troupe-standup/test_verification.txt from a __main__ and "
"compares a subprocess run against `wc`). It is a host-path transcript, not "
"a primitive; nothing in the lane imports it.",
"CHECK_toposort.py":
"hand-run check for toposort_kahn.py, named CHECK_ so the collector does not "
"take it (it is not unittest or bare-test shaped for the runner either: its "
"three graphs are `def test_dag` style cells with no runner). Its subject is "
"covered from this commit by test_toposort_kahn.py, which ports these three "
"graphs and adds the cycle cases; kept as the original reading.",
"_run_ng.py":
"five-line hand runner: it imports test_next_greater and calls every test_* "
"in it, printing PASS per cell. Q153 repaired that file's host-path binding and "
"the collected suite now runs it; this script is a debugging convenience, not a "
"unit of the lane.",
"conftest.py":
"pytest collection hook (two lines: put the lane on sys.path), read by the "
"collector rather than imported by a test. The lane's gate does not use pytest "
"at all (Q128), so no cell can import it.",
}
def module_stem(path):
return os.path.splitext(os.path.basename(path))[0]
def _parse(path):
"""ast.parse one file, or None when it will not parse."""
try:
with open(path, "r", encoding="utf-8", errors="replace") as fh:
return ast.parse(fh.read(), filename=path)
except (OSError, SyntaxError):
return None
def imported_names(tree):
"""First components of every import in one parsed file."""
names = set()
for node in ast.walk(tree):
if isinstance(node, ast.Import):
for alias in node.names:
names.add(alias.name.split(".")[0])
elif isinstance(node, ast.ImportFrom):
if node.level == 0 and node.module:
names.add(node.module.split(".")[0])
return names
def string_literals(tree):
"""Every string constant in one parsed file."""
return {
node.value
for node in ast.walk(tree)
if isinstance(node, ast.Constant) and isinstance(node.value, str)
}
def grade(lane, exempt=None):
"""Grade one lane directory. Returns (rc, rows, notes).
rows: list of dicts {module, state, why} -- state is one of
covered-test-import, covered-transitive, covered-literal, covered-gate,
exempt, UNCOVERED, WAIVER-MISSING, WAIVER-STALE, WAIVER-NO-REASON.
"""
exempt = dict(EXEMPT if exempt is None else exempt)
# Waivers are keyed by filename ("probe.py"); a bare stem is accepted and
# normalised here so the table and the census speak the same keys.
waivers = {}
for key, reason in exempt.items():
fname = key if key.endswith(".py") else key + ".py"
waivers[fname] = reason
notes = []
rows = []
if not os.path.isdir(lane):
return 2, rows, ["no lane directory: %s" % lane]
lane_py = sorted(glob.glob(os.path.join(lane, "*.py")))
tests = [p for p in lane_py if os.path.basename(p).startswith("test_")]
modules = [p for p in lane_py if not os.path.basename(p).startswith("test_")]
if not tests:
return 2, rows, ["no collected tests (test_*.py) in %s: nothing can be read" % lane]
if not modules:
return 2, rows, ["no lane modules (*.py minus test_*) in %s: nothing to grade" % lane]
lane_names = {module_stem(p) for p in modules}
files = {module_stem(p): p for p in modules}
# --- parse the collected tests; a broken test file is ungradable -------
test_trees = {}
for t in tests:
tree = _parse(t)
if tree is None:
return 2, rows, ["collected test file will not parse: %s" % os.path.basename(t)]
test_trees[t] = tree
# --- direct imports and literals from the collected tests --------------
seeds, seed_why, literals = set(), {}, set()
for t, tree in test_trees.items():
for name in imported_names(tree) & lane_names:
seeds.add(name)
seed_why.setdefault(name, os.path.basename(t))
literals |= string_literals(tree)
covered, why = {}, {}
def cover(name, state, reason):
if name not in covered:
covered[name] = state
why[name] = reason
for name in sorted(seeds):
cover(name, "covered-test-import", "imported by %s" % seed_why[name])
for name in sorted(lane_names):
if name in covered:
continue
# A literal counts only when it IS a path to the file: the bare
# filename (`"shadow_lint.py"`) or a path ending in it
# (`os.path.join(HERE, "shadow_lint.py")`, `"/root/.../sha256_manifest.py"`).
# A mention inside prose -- a docstring saying "test_sha256_manifest.py
# grades ..." -- is not a read, and counting prose is exactly the
# substring-grep defect Q85's own census had (it reported four modules
# invisible while a sixth was read by a path constant only).
def _is_path_literal(lit):
return (os.path.basename(lit.strip()) == name + ".py"
and not any(ch.isspace() for ch in lit))
if any(_is_path_literal(lit) for lit in literals):
cover(name, "covered-literal", "the test names the file's path in a literal")
# --- transitive closure over the lane's own import graph ---------------
edges = {}
for name, path in files.items():
tree = _parse(path)
edges[name] = (imported_names(tree) if tree is not None else set()) & lane_names
if tree is None:
notes.append("lane module will not parse (its imports are unread): %s.py" % name)
queue = [n for n in covered if covered[n] == "covered-test-import"]
while queue:
cur = queue.pop()
for nxt in sorted(edges.get(cur, ())):
if nxt not in covered:
cover(nxt, "covered-transitive", "imported through %s.py" % cur)
queue.append(nxt)
# --- what the gate itself runs ----------------------------------------
check_path = os.path.join(lane, "CHECK")
gate_text = ""
if os.path.isfile(check_path):
with open(check_path, "r", encoding="utf-8", errors="replace") as fh:
gate_text = fh.read()
else:
notes.append("no CHECK in %s: gate-invoked coverage is unread" % lane)
for name in sorted(lane_names):
if name in covered:
continue
if re.search(r"python3\s+(?:\./)?%s\.py\b" % re.escape(name), gate_text):
cover(name, "covered-gate", "run by CHECK (python3 %s.py)" % name)
# --- every module accounted for ---------------------------------------
for name in sorted(lane_names):
if name in covered:
rows.append({"module": name, "state": covered[name], "why": why[name]})
elif (name + ".py") in waivers:
reason = (waivers[name + ".py"] or "").strip()
if not reason:
rows.append({"module": name, "state": "WAIVER-NO-REASON",
"why": "waived in EXEMPT with no reason: a waiver needs a sentence"})
else:
rows.append({"module": name, "state": "exempt", "why": reason})
else:
rows.append({"module": name, "state": "UNCOVERED",
"why": "imported by no collected test, named by no literal, run by no CHECK step"})
# --- the waiver table is itself graded --------------------------------
for fname in sorted(waivers):
stem = module_stem(fname)
if os.path.basename(fname) != fname:
rows.append({"module": fname, "state": "WAIVER-MISSING",
"why": "EXEMPT names a path, not a lane module: %r" % fname})
elif fname not in {os.path.basename(p) for p in modules}:
rows.append({"module": stem, "state": "WAIVER-MISSING",
"why": "EXEMPT waives %r, which is not a module in this lane" % fname})
elif stem in covered:
rows.append({"module": stem, "state": "WAIVER-STALE",
"why": "%s is now covered (%s): delete this waiver" % (fname, covered[stem])})
faults = [r for r in rows if r["state"] in
("UNCOVERED", "WAIVER-MISSING", "WAIVER-STALE", "WAIVER-NO-REASON")]
return (1 if faults else 0), rows, notes
def report(lane, rc, rows, notes, out=sys.stdout):
states = {}
for r in rows:
states[r["state"]] = states.get(r["state"], 0) + 1
out.write("coverage: %s\n" % lane)
for r in rows:
if r["state"] not in ("covered-test-import", "covered-transitive", "covered-gate"):
out.write(" %-14s %-26s %s\n" % (r["state"], r["module"], r["why"]))
for note in notes:
out.write(" note %s\n" % note)
out.write(
"coverage: %d module(s): %d covered by a collected test, %d through an import "
"chain, %d by literal, %d run by CHECK, %d exempt, %d fault(s)\n"
% (len(rows),
states.get("covered-test-import", 0), states.get("covered-transitive", 0),
states.get("covered-literal", 0), states.get("covered-gate", 0),
states.get("exempt", 0),
sum(states.get(s, 0) for s in
("UNCOVERED", "WAIVER-MISSING", "WAIVER-STALE", "WAIVER-NO-REASON"))))
out.write("coverage: %s\n" % {0: "PASS", 1: "FAULT", 2: "UNGRADABLE"}[rc])
# --------------------------------------------------------------------------
# selftest: fixture lanes built in a throwaway directory, each one cell.
# --------------------------------------------------------------------------
def _fixture(root, name, files, exempt=None, check_text=None):
"""Build one fixture lane; returns (lane_dir, exempt_dict)."""
lane = os.path.join(root, name)
os.makedirs(lane, exist_ok=True)
for rel, body in files.items():
path = os.path.join(lane, rel)
os.makedirs(os.path.dirname(path), exist_ok=True)
with open(path, "w", encoding="utf-8") as fh:
fh.write(body)
if check_text is not None:
with open(os.path.join(lane, "CHECK"), "w", encoding="utf-8") as fh:
fh.write(check_text)
return lane, (exempt or {})
def selftest(tmp):
cases = []
def cell(name, lane, exempt, want_rc, want_states, want_named=None):
rc, rows, notes = grade(lane, exempt)
got = {r["module"]: r["state"] for r in rows}
ok = rc == want_rc
detail = "rc=%d want %d" % (rc, want_rc)
for mod, state in want_states.items():
if got.get(mod) != state:
ok = False
detail += "; %s=%r want %r" % (mod, got.get(mod), state)
if want_named:
text = "\n".join("%s %s" % (r["state"], r["module"]) for r in rows)
for needle in want_named:
if needle not in text:
ok = False
detail += "; %r not in the report" % needle
cases.append({"case": name, "ok": ok, "detail": detail})
return rc, rows, notes
cover_test = "from helper import value\nassert value() == 1\n"
chain_test = "from first import one\nassert one() == 1\n"
literal_test = 'import os, subprocess\nSCRIPT = os.path.join(os.path.dirname(__file__), "helper.py")\n'
# 1 a collected test imports the module
lane, ex = _fixture(tmp, "import-covered", {
"helper.py": "def value():\n return 1\n",
"test_helper.py": cover_test,
})
cell("import-covered", lane, ex, 0, {"helper": "covered-test-import"})
# 2 the import chain covers the second module
lane, ex = _fixture(tmp, "transitive-covered", {
"first.py": "from second import two\ndef one():\n return two()\n",
"second.py": "def two():\n return 1\n",
"test_first.py": chain_test,
})
cell("transitive-covered", lane, ex, 0,
{"first": "covered-test-import", "second": "covered-transitive"})
# 3 a filename literal covers a module the test runs, not imports
lane, ex = _fixture(tmp, "literal-covered", {
"helper.py": "print('hi')\n",
"test_run.py": literal_test,
})
cell("literal-covered", lane, ex, 0, {"helper": "covered-literal"})
# 4 nothing reads it -> red, and the report names it
lane, ex = _fixture(tmp, "uncovered", {
"orphan.py": "def f():\n return 1\n",
"test_other.py": "def test_x():\n assert True\n",
})
cell("uncovered-is-red", lane, ex, 1, {"orphan": "UNCOVERED"}, want_named=["UNCOVERED orphan"])
# 5 a waiver with a reason is a pass, and the reason is printed
lane, ex = _fixture(tmp, "exempt-reason", {
"probe.py": "print('hand run')\n",
"test_other.py": "def test_x():\n assert True\n",
}, exempt={"probe.py": "one-off probe, hand-run against a subject the suite covers"})
rc, rows, _ = cell("exempt-with-reason", lane, ex, 0, {"probe": "exempt"})
if not any("one-off probe" in r["why"] for r in rows):
cases[-1]["ok"] = False
cases[-1]["detail"] += "; reason not printed"
# 6 a waiver without a reason is a fault, not a silence
lane, ex = _fixture(tmp, "exempt-silent", {
"probe.py": "print('hand run')\n",
"test_other.py": "def test_x():\n assert True\n",
}, exempt={"probe.py": " "})
cell("waiver-needs-a-reason", lane, ex, 1, {"probe": "WAIVER-NO-REASON"})
# 7 a waiver naming a file that is gone is a fault
lane, ex = _fixture(tmp, "exempt-missing", {
"kept.py": "def f():\n return 1\n",
"test_kept.py": "from kept import f\nassert f() == 1\n",
}, exempt={"deleted_last_year.py": "was a probe"})
cell("waiver-names-a-missing-file", lane, ex, 1,
{"deleted_last_year": "WAIVER-MISSING"})
# 8 a waiver left standing over a now-covered module is a fault
lane, ex = _fixture(tmp, "exempt-stale", {
"helper.py": "def value():\n return 1\n",
"test_helper.py": cover_test,
}, exempt={"helper.py": "used to be uncovered"})
cell("waiver-now-covered", lane, ex, 1, {"helper": "WAIVER-STALE"})
# 9 a module CHECK runs is covered by the gate itself
lane, ex = _fixture(tmp, "gate-invoked", {
"gate_script.py": "print('gate')\n",
"test_other.py": "def test_x():\n assert True\n",
}, check_text="#!/bin/bash\npython3 gate_script.py\n")
cell("gate-invoked", lane, ex, 0, {"gate_script": "covered-gate"})
# 10 no collected tests -> cannot grade, rather than "all covered"
lane, ex = _fixture(tmp, "no-tests", {"orphan.py": "x = 1\n"})
cell("no-tests-is-ungradable", lane, ex, 2, {})
# 11 a test file that will not parse -> cannot grade
lane, ex = _fixture(tmp, "bad-test", {
"orphan.py": "x = 1\n",
"test_broken.py": "def test_x(:\n pass\n",
})
cell("unparseable-test-is-ungradable", lane, ex, 2, {})
# 12 the falsifiability cell: delete the one test file, the fixture reds
lane, ex = _fixture(tmp, "mutation", {
"helper.py": "def value():\n return 1\n",
"test_helper.py": cover_test,
})
first = grade(lane, ex)[0]
os.remove(os.path.join(lane, "test_helper.py"))
with open(os.path.join(lane, "test_other.py"), "w", encoding="utf-8") as fh:
fh.write("def test_x():\n assert True\n")
second = grade(lane, ex)[0]
ok = first == 0 and second == 1
cases.append({"case": "removing-the-cell-reds-it",
"ok": ok, "detail": "rc %d -> %d (want 0 -> 1)" % (first, second)})
# 13 no lane directory
cases.append({"case": "missing-lane-is-ungradable",
"ok": grade(os.path.join(tmp, "does-not-exist"), {})[0] == 2,
"detail": "rc=%d" % grade(os.path.join(tmp, "does-not-exist"), {})[0]})
bad = [c for c in cases if not c["ok"]]
for c in cases:
print("%-4s %-32s %s" % ("ok" if c["ok"] else "FAIL", c["case"], c["detail"]))
print("selftest: %d cell(s), %d failed" % (len(cases), len(bad)))
return 1 if bad else 0
def main(argv=None):
ap = argparse.ArgumentParser(description="grade lane-module coverage for one lane")
ap.add_argument("--lane", default=DEFAULT_LANE)
ap.add_argument("--json", default=None, help="write the row table here")
ap.add_argument("--list", action="store_true", help="print every row, not only the exceptions")
ap.add_argument("--selftest", action="store_true")
args = ap.parse_args(argv)
if args.selftest:
import tempfile
with tempfile.TemporaryDirectory(prefix="coverage-selftest.") as tmp:
return selftest(tmp)
rc, rows, notes = grade(args.lane)
if args.list:
for r in rows:
print("%-20s %-26s %s" % (r["state"], r["module"], r["why"]))
report(args.lane, rc, rows, notes)
if args.json:
with open(args.json, "w", encoding="utf-8") as fh:
json.dump({"lane": args.lane, "rc": rc, "rows": rows, "notes": notes}, fh, indent=1)
return rc
if __name__ == "__main__":
sys.exit(main())
wizard-work/beam/test_zeller_dow.py
+11
View File
@@ -27,5 +27,16 @@ class TestZellerDow(unittest.TestCase):
with self.assertRaises(ValueError):
dow(2023, 2, 29)
def test_julian_leap_branch_is_the_witness(self):
# Q85 M-JUL: with the calendar branch dropped -- `leap = _gregorian_leap(y)`
# unconditionally -- every other cell here still passes and this date
# alone moves from "Saturday" to ValueError. 1500 is a Julian leap year
# and not a Gregorian one, so it is the only year in the suite that
# separates the two branches. Cross-checked two ways outside the module:
# its JDN is shared by Gregorian 1500-03-10, whose datetime weekday is
# Saturday, and a walk from Julian 1582-10-04 (a Thursday) lands on
# Saturday too. Both readings are in the Q85 record.
self.assertEqual(dow(1500, 2, 29), "Saturday")
if __name__ == "__main__":
unittest.main()