feat(content): reject all-filler multi-token strings in reject_trivial

The single non-empty/non-placeholder gate caught a lone banned token
(wip) and below-floor strings, but multi-token soup made entirely of
placeholders (wip wip, tbd / na, todo todo todo) slipped through both
checks. Add an all-tokens-filler test that strips edge punctuation per
token and rejects when every meaningful token is banned — without
flagging real prose that merely contains a filler word (none of the
tests failed). Strengthens every content model + gateway anti-soup
guard that composes reject_trivial.
This commit is contained in:
Renn F
2026-06-21 04:42:26 +02:00
parent db74f546a3
commit b1f1a91066
2 changed files with 81 additions and 4 deletions
+22 -4
View File
@@ -14,6 +14,7 @@ Two jobs:
from __future__ import annotations
import string
from typing import Any
# Placeholder tokens that are never an acceptable whole-field value. Extends the
@@ -54,19 +55,36 @@ class ContentValidationError(Exception):
super().__init__(f"{field}: {reason}")
def _all_tokens_filler(text: str) -> bool:
"""True when every whitespace token (sans edge punctuation) is a placeholder.
Catches multi-token soup that no single check would — ``wip wip``,
``tbd / na``, ``todo todo todo`` — without flagging real prose that merely
*contains* a filler word (``none of the tests failed``). Pure-punctuation
tokens (``/``) strip to empty and are dropped before the all-filler test.
"""
meaningful = [
stripped
for tok in text.split()
if (stripped := tok.strip(string.punctuation).lower())
]
return bool(meaningful) and all(tok in BANNED_PHRASES for tok in meaningful)
def reject_trivial(value: str, *, field: str, min_chars: int = 1) -> str:
"""Return the trimmed value, or raise ``ValueError`` if it is trivial.
Trivial = empty, shorter than ``min_chars``, or a known placeholder token.
Raises ``ValueError`` (not ``ContentValidationError``) so it can be used
directly inside Pydantic field validators.
Trivial = empty, shorter than ``min_chars``, a known placeholder token, or a
string whose every token is a placeholder (``wip wip``). Raises
``ValueError`` (not ``ContentValidationError``) so it can be used directly
inside Pydantic field validators.
"""
text = (value or "").strip()
if not text:
raise ValueError(f"{field} must not be empty")
if len(text) < min_chars:
raise ValueError(f"{field} must be at least {min_chars} characters")
if text.lower() in BANNED_PHRASES:
if text.lower() in BANNED_PHRASES or _all_tokens_filler(text):
raise ValueError(f"{field} must not be placeholder text (got {value!r})")
return text
@@ -0,0 +1,59 @@
"""Tests for the shared content validators (``reject_trivial``).
``reject_trivial`` is the single non-empty / non-placeholder gate reused by
every structured-content field validator AND the gateway anti-soup guards. It
must reject: empty, too-short, a lone placeholder token, AND a string whose
every whitespace token is a placeholder (``wip wip``, ``tbd / na``).
"""
from __future__ import annotations
import pytest
from roboco.foundation.policy.content.validators import reject_trivial
def test_returns_trimmed_value_when_substantive() -> None:
assert reject_trivial(" real content here ", field="x") == "real content here"
@pytest.mark.parametrize("blank", ["", " ", "\t\n"])
def test_rejects_empty(blank: str) -> None:
with pytest.raises(ValueError, match="must not be empty"):
reject_trivial(blank, field="summary")
def test_rejects_too_short() -> None:
with pytest.raises(ValueError, match="at least 10 characters"):
reject_trivial("short", field="summary", min_chars=10)
@pytest.mark.parametrize(
"token", ["wip", "TBD", "asdf", "n/a", "none", "-", "...", "x"]
)
def test_rejects_lone_placeholder_token(token: str) -> None:
with pytest.raises(ValueError, match="placeholder"):
reject_trivial(token, field="summary")
@pytest.mark.parametrize(
"soup",
["wip wip", "tbd / na", "todo todo todo", "asdf asdf asdf asdf", "none none"],
)
def test_rejects_all_filler_token_string(soup: str) -> None:
# Every token is a placeholder — the whole string is soup even though it is
# neither a single banned token nor below the length floor.
with pytest.raises(ValueError, match="placeholder"):
reject_trivial(soup, field="summary")
@pytest.mark.parametrize(
"ok",
[
"none of the tests failed", # 'none' present but not all-filler
"fixed the x coordinate bug", # 'x' present but not all-filler
"rate_limited", # a real substitute reason
"LGTM, merging now",
],
)
def test_accepts_real_text_containing_a_filler_word(ok: str) -> None:
assert reject_trivial(ok, field="reason", min_chars=3) == ok.strip()