test: distinguish clean and unsupported inputs

This commit is contained in:
Dandaedre 2026-09-29 15:28:32 +02:00
parent 6fe5f90c17
commit c62318304a
Signed by: Dandaedre
GPG Key ID: 440C6592E8A26D99
4 changed files with 34 additions and 3 deletions

View File

@ -24,8 +24,10 @@ cluster of closely related phrases from being treated as independent proof.
If the resulting evidence sum is `E`, the displayed score is
`round(100 × E / (E + 50))`, bounded to `0..=100`. The configured `low`,
`medium`, and `high` thresholds name the resulting band. Evidence below `low`
is explicitly reported as below threshold. No findings produce score zero and
no confidence verdict.
is explicitly reported as below threshold. Empty input and text with no
findings produce score zero and no confidence verdict. Binary, invalid UTF-8,
oversized, and unsupported inputs remain typed input errors; they are never
reported as clean analyses.
Weights and thresholds are review policy, not calibrated probabilities. A
score of 75 does not mean a 75% chance that AI wrote the input.

View File

@ -751,4 +751,16 @@ mod tests {
assert_eq!(analysis.findings.len(), 3);
assert_ne!(analysis.confidence_band, Some(ConfidenceBand::High));
}
#[test]
fn empty_and_clean_text_have_zero_confidence_and_no_verdict() {
let detector = Detector::new(&Config::embedded()).unwrap();
for source in ["", "A specific, carefully sourced statement."] {
let analysis = detector.analyze(source);
assert!(analysis.findings.is_empty());
assert_eq!(analysis.confidence.score(), 0);
assert_eq!(analysis.confidence_band, None);
}
}
}

View File

@ -94,6 +94,23 @@ fn invalid_utf8_is_a_per_document_error() {
assert!(documents.next().is_none());
}
#[test]
fn binary_content_is_an_input_error_not_a_clean_document() {
let directory = tempdir().expect("temp dir");
let path = directory.path().join("binary.txt");
fs::write(&path, b"clean prefix\0binary suffix").expect("write fixture");
let mut documents = read(&path, &limits()).expect("open input");
let error = documents.next().expect("one result").unwrap_err();
assert!(matches!(
error,
InputError::BinaryContent {
source_id: SourceId::File(error_path)
} if error_path == path
));
}
#[test]
fn missing_path_is_rejected() {
let directory = tempdir().expect("temp dir");

View File

@ -120,7 +120,7 @@ configuration is explainable, and all precedence/validation tests pass.
- [x] Design the confidence formula: bounded weighted evidence, per-rule caps,
correlated-family damping, normalization to `0..=100`, and configurable
low/medium/high thresholds.
- [ ] Return zero confidence and no verdict for empty/no-evidence input. Keep
- [x] Return zero confidence and no verdict for empty/no-evidence input. Keep
unsupported/binary input distinct from a clean analysis.
- [ ] Sort findings deterministically and deduplicate identical rule/span pairs.
- [ ] Add table-driven unit tests for every shipped rule with positive examples