Datasets:
Mark the figures the release cannot reproduce
Browse files- tests/test_conformance.py +17 -0
tests/test_conformance.py
CHANGED
|
@@ -213,3 +213,20 @@ def test_index_matches_the_manifest_exactly():
|
|
| 213 |
subprocess.run([_sys.executable, str(here / "build_index.py")],
|
| 214 |
capture_output=True, check=True)
|
| 215 |
assert idx.read_bytes() == before, "index.jsonl is stale vs the manifest"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 213 |
subprocess.run([_sys.executable, str(here / "build_index.py")],
|
| 214 |
capture_output=True, check=True)
|
| 215 |
assert idx.read_bytes() == before, "index.jsonl is stale vs the manifest"
|
| 216 |
+
|
| 217 |
+
|
| 218 |
+
def test_readme_flags_the_one_figure_it_cannot_reproduce():
|
| 219 |
+
"""1.2441 is a pre-fix value from a superseded generator revision.
|
| 220 |
+
|
| 221 |
+
Every other number on the card re-derives from the committed corpus. This
|
| 222 |
+
one does not, so the card has to say so -- and the resonator it describes
|
| 223 |
+
must actually be passive today, or the anecdote is worse than the bug.
|
| 224 |
+
"""
|
| 225 |
+
readme = (HERE / "README.md").read_text()
|
| 226 |
+
assert "1.2441" in readme, "drop this guard along with the figure"
|
| 227 |
+
assert "cannot reproduce from the" in readme
|
| 228 |
+
assert "superseded generator revision" in readme
|
| 229 |
+
|
| 230 |
+
net = read_touchstone(str(HERE / "data" / "passive_resonator.s2p"))
|
| 231 |
+
sigma = float(np.max(np.linalg.svd(net.s, compute_uv=False)))
|
| 232 |
+
assert sigma <= 1.0 + 1e-9, f"shipped resonator is not passive: {sigma}"
|