composer: enumerated parameters, and length_view becomes reachable

length_view was in both families' defaults and in neither COMMON_GROUPS, so the
Full Length branch of model_length_mm could not be reached from the composer.
Every model was a 100 mm preview whatever member_length_ft said. Found while
specifying STL export: the sweep must equal model_length_mm(p), the value
already published as LENGTH_MM, or the record's VOLUME_MM3 and MASS_G describe a
different object than the file beside them.

Adding it to the list would have been one line and the wrong fix. The type
coercion falls through to the raw string for anything that is not a bool, int
or float, and model_length_mm compares against the literal "Full Length" and
silently falls back to preview for anything else -- the reference behaved the
same way, so the port is right to keep it. A free-text box would have replaced
an unreachable control with a silent one: "Full Length " with a trailing space
builds a 100 mm model and says nothing.

So ENUM_PARAMS declares which parameters take one of a fixed set of values.
The browser renders a select instead of an input, and a value outside the set
is dropped before the build rather than passed through -- the family default
stands, which is a valid model. The guard sits before the coercion, not after,
where it would be dead code. The next enumerated parameter needs no new
machinery.

Deliberately not mirrored as a check in the families. geometry_checks is shared
with the oracle and the reference accepted any string here; adding a check
there would put the 123 frozen cases at risk to fix a user-interface problem.
The guard belongs in the composer, which is not oracle-bearing.

19 assertions, mutation-proven: removing the guard fails 8 of them. Six are the
parametrised bad values. The other two assert the mechanism rather than its
effect -- that the string never enters values at all, and that the fallback
reaches input_id. Without them, a pass-through that happened to be ignored
downstream would look identical to a working fallback, and a later change to
model_length_mm would turn a passing test into a defect somewhere else.

length_view is a build parameter and stays in both hashes, unlike the author.
Two members of different lengths are different designs, and a test asserts it.

Suite 566 passed, of which 19 are new. None of the 547 moved.
This commit is contained in:
2026-09-12 10:19:47 -05:00
parent 56e0570f10
commit ecac644204
2 changed files with 214 additions and 7 deletions
+53 -5
View File
@@ -41,15 +41,44 @@ from mechcomp import svg
# as root:mechcomp 0640, and the service user can read it.
ENV_FILE = "/etc/mechcomp/mechcomp.env"
# Parameters whose value is one of a fixed set rather than a number.
#
# WHY THIS EXISTS RATHER THAN A SPECIAL CASE FOR length_view
# A free-text box for a parameter with two valid values is a trap. The
# type coercion below falls through to the raw string for anything that is
# not a bool, int or float, and ``_common.model_length_mm`` compares
# against the literal "Full Length" and silently falls back to the preview
# length for anything else. So "Full Length " with a trailing space would
# produce a 100 mm model with no rejection and no error -- a file that
# looks right and is the wrong object.
#
# Declaring the choices here does three things at once: the browser renders
# a select instead of an input, a value outside the set never reaches the
# build, and the next enumerated parameter needs no new machinery.
#
# NOT MIRRORED AS A CHECK IN THE FAMILIES, DELIBERATELY
# ``geometry_checks`` is shared with the oracle, and the reference accepted
# any string here and fell through to preview. Adding a check there would
# put the 123 frozen cases at risk to fix a user-interface problem. The
# guard belongs in the composer, which is not oracle-bearing.
ENUM_PARAMS: Dict[str, List[str]] = {
"length_view": ["Preview", "Full Length"],
}
# Parameters that apply to every profile, in the order they make sense to a
# person: what the stock is, how it fits, then how thick the printed walls are.
#
# ``length_view`` leads the Model group because it decides which of the two
# lengths below it actually applies. Without it the Full Length branch was
# unreachable from the composer and every model was a 100 mm preview whether
# or not ``member_length_ft`` had been set.
COMMON_GROUPS: List[Tuple[str, List[str]]] = [
("Stock", ["strap_width_mm", "strap_thickness_mm", "bundle_count"]),
("Fit", ["fit_clearance_mm"]),
("Walls", ["inside_wall_thickness_mm", "outside_wall_thickness_mm",
"edge_wall_thickness_mm", "min_wall_mm"]),
("Model", ["preview_length_mm", "member_length_ft", "material_density_g_cm3",
"facets"]),
("Model", ["length_view", "preview_length_mm", "member_length_ft",
"material_density_g_cm3", "facets"]),
]
# Which parameter prefixes belong to which profile, so a person is not shown
@@ -110,6 +139,16 @@ def build_payload(family_name: str, profile: str,
for key, raw in overrides.items():
if key not in family.defaults:
continue
# An enumerated parameter takes its value or nothing. Dropping an
# unrecognised value leaves the family default in place, which is a
# valid model, rather than passing a string the geometry will ignore
# without saying so.
if key in ENUM_PARAMS:
if raw in ENUM_PARAMS[key]:
typed[key] = raw
continue
default = family.defaults[key]
try:
if isinstance(default, bool):
@@ -127,7 +166,7 @@ def build_payload(family_name: str, profile: str,
"family": family_name,
"profile": profile,
"groups": [[title, keys] for title, keys in profile_params(family, profile)],
"defaults": {k: family.defaults[k] for k in family.defaults},
"choices": {k: v for k, v in ENUM_PARAMS.items() if k in family.defaults},
"values": {**family.defaults, **typed},
"profiles": sorted(family.catalogue),
}
@@ -255,6 +294,7 @@ async function refresh(sendValues) {
prof.onchange = () => { state.profile = prof.value; refresh(true); };
state.values = {};
const choices = data.choices || {};
const box = document.getElementById("controls");
box.innerHTML = "";
for (const [title, keys] of data.groups) {
@@ -267,9 +307,16 @@ async function refresh(sendValues) {
const name = document.createElement("span");
name.textContent = k.replace(/_mm$|_deg$|_ft$/, "").replace(/_/g, " ");
name.title = k;
const inp = document.createElement("input");
let inp;
if (choices[k]) {
inp = document.createElement("select");
for (const c of choices[k]) inp.add(new Option(c, c));
inp.value = v;
} else {
inp = document.createElement("input");
inp.value = v;
if (typeof v === "number") { inp.type = "number"; inp.step = "any"; }
}
inp.onchange = () => { state.values[k] = inp.value; refresh(true); };
lab.append(name, inp);
fs.append(lab);
@@ -341,7 +388,8 @@ class Handler(BaseHTTPRequestHandler):
except Exception as exc: # noqa: BLE001
payload = {"ok": False,
"message": "%s: %s" % (type(exc).__name__, exc),
"profiles": [], "groups": [], "values": {}}
"profiles": [], "groups": [], "values": {},
"choices": {}}
self._send(200, json.dumps(payload).encode("utf-8"),
"application/json; charset=utf-8")
return
+159
View File
@@ -0,0 +1,159 @@
"""
Enumerated parameters, and the length control that motivated them.
WHAT THIS IS ACTUALLY GUARDING
``model_length_mm`` compares ``length_view`` against the literal
"Full Length" and falls through to the preview length for anything else --
no rejection, no error. The reference behaved the same way, so the port is
right to keep it.
That makes the composer the only place the mistake can be caught. Before
this, ``length_view`` was not in COMMON_GROUPS at all, so the Full Length
branch was unreachable and every model was a 100 mm preview whatever
``member_length_ft`` said. Exposing it as free text would have replaced an
unreachable control with a silent one: "Full Length " with a trailing space
builds a preview and says nothing.
So the claims under test are: the branch is reachable, it moves the length
it is supposed to move, and a value outside the declared set cannot reach
the build.
"""
from __future__ import annotations
import pytest
app = pytest.importorskip("mechcomp.web.app")
def payload(**overrides):
return app.build_payload("3x", "Y", dict(overrides))
def length_of(**overrides):
p = payload(**overrides)
assert p["ok"], p.get("message")
return p["report"]["LENGTH_MM"]
# ---------------------------------------------------------------------------
# The declaration
# ---------------------------------------------------------------------------
def test_length_view_is_declared_with_exactly_its_two_values():
assert app.ENUM_PARAMS["length_view"] == ["Preview", "Full Length"]
def test_every_enumerated_parameter_is_a_real_family_parameter():
"""
A declared choice for a parameter no family has would render a control
that does nothing -- the failure COMMON_GROUPS has a comment about.
"""
for family in app.families().values():
for key in app.ENUM_PARAMS:
assert key in family.defaults, key
def test_every_declared_default_is_one_of_its_own_choices():
"""
If a family's default were outside the list, the select would open showing
a value it cannot represent and the first interaction would silently move
the model.
"""
for family in app.families().values():
for key, allowed in app.ENUM_PARAMS.items():
assert family.defaults[key] in allowed
# ---------------------------------------------------------------------------
# Reachability -- the defect that started this
# ---------------------------------------------------------------------------
def test_length_view_is_reachable_from_the_composer():
for name, family in app.families().items():
exposed = {k for _, keys in app.profile_params(family, "Y" if name == "3x" else "Cross")
for k in keys}
assert "length_view" in exposed, name
def test_the_browser_is_told_the_choices():
assert payload()["choices"]["length_view"] == ["Preview", "Full Length"]
# ---------------------------------------------------------------------------
# It moves the length it claims to move
# ---------------------------------------------------------------------------
def test_preview_is_the_preview_length():
assert length_of(length_view="Preview", preview_length_mm=100) == 100
def test_full_length_is_feet_times_304_8():
assert length_of(length_view="Full Length", member_length_ft=10) == 3048
def test_full_length_follows_member_length_ft():
"""
The positive control for the pair above: a length that ignored its input
would satisfy either test alone by returning a constant.
"""
assert length_of(length_view="Full Length", member_length_ft=20) == 6096
def test_preview_ignores_member_length_ft():
assert length_of(length_view="Preview", member_length_ft=40) == 100
# ---------------------------------------------------------------------------
# A value outside the set never reaches the build
# ---------------------------------------------------------------------------
@pytest.mark.parametrize("bad", ["Full Length ", "full length", "FULL LENGTH",
"", "Fully Lengthed", "Preview\n"])
def test_an_unrecognised_value_falls_back_to_the_default(bad):
"""
Not a rejection -- a fallback. The family default is a valid model, and
the composer is a viewer rather than a validator. What must not happen is
the string reaching ``model_length_mm`` and being silently ignored there,
which is how a trailing space becomes a 100 mm export of a 10 ft member.
"""
p = payload(length_view=bad, member_length_ft=10)
assert p["ok"]
assert p["values"]["length_view"] == "Preview"
assert p["report"]["LENGTH_MM"] == 100
def test_the_unrecognised_value_is_not_merely_unused_but_absent():
"""
Asserts the mechanism rather than its effect. If the string were passed
through and happened to be ignored downstream, the test above would still
pass and a later change to ``model_length_mm`` would turn it into a defect.
"""
p = payload(length_view="Full Length ", member_length_ft=10)
assert "Full Length " not in p["values"].values()
def test_a_recognised_value_does_reach_the_build():
"""The positive control: the fallback is not simply dropping everything."""
p = payload(length_view="Full Length", member_length_ft=10)
assert p["values"]["length_view"] == "Full Length"
assert p["report"]["LENGTH_MM"] == 3048
# ---------------------------------------------------------------------------
# The length control does not disturb identity
# ---------------------------------------------------------------------------
def test_length_view_changes_the_design_id():
"""
It is a build parameter, so it belongs in the hash -- unlike the author.
Two members of different lengths are different designs.
"""
preview = payload(length_view="Preview")["input_id"]
full = payload(length_view="Full Length")["input_id"]
assert preview != full
def test_an_unrecognised_value_yields_the_default_design_id():
"""Follows from the fallback, and is the property someone would rely on."""
assert payload(length_view="Full Length ")["input_id"] == payload()["input_id"]