Classify source relatedness as data-quality advisories

This commit is contained in:
2026-08-27 15:49:53 +00:00
parent 1f1967c8d2
commit ccba2ce3f9
25 changed files with 232 additions and 189 deletions

View File

@@ -1,4 +1,4 @@
// Package sourcerelatedness warns when cited source text does not mention a location occurrence.
// Package sourcerelatedness reports advisory findings when cited source text does not mention a location occurrence.
package sourcerelatedness
import (
@@ -14,10 +14,9 @@ import (
)
const (
Key = "extract/dnd/location-occurrences/source_relatedness"
WarningReasonCode = "location_occurrence_not_near_source"
OmittedReasonCode = "location_occurrence_relatedness_warnings_omitted"
policy = "dnd.location_occurrences.validator.source_relatedness.v1"
Key = "extract/dnd/location-occurrences/source_relatedness"
ReasonCode = "location_occurrence_not_near_source"
policy = "dnd.location_occurrences.validator.source_relatedness.v1"
)
type Options struct{}
@@ -56,9 +55,9 @@ func (v *Validator) Validate(_ context.Context, req contracts.TypedValidationReq
if shared.ContainsTokenSequence(citedTexts[index], occurrence.Name) {
continue
}
warnings = append(warnings, contracts.Warning{Scope: fmt.Sprintf("occurrences[%d]", index), ReasonCode: WarningReasonCode, Message: fmt.Sprintf("Location occurrence name %s was not found in cited source text", diagnostics.Quote(occurrence.Name))})
warnings = append(warnings, contracts.Warning{Scope: fmt.Sprintf("occurrences[%d]", index), ReasonCode: ReasonCode, Message: fmt.Sprintf("Location occurrence name %s was not found in cited source text", diagnostics.Quote(occurrence.Name))})
}
return contracts.ValidationResult{Approved: true, Warnings: diagnostics.LimitWarnings(warnings, "location_occurrences", OmittedReasonCode)}, nil
return diagnostics.DataQualityResult(warnings)
}
func Spec() pipeline.ValidatorSpec {
return pipeline.ValidatorSpec{Key: Key, ExecutionClass: contracts.ExecutionClassDeterministic}

View File

@@ -10,21 +10,20 @@ import (
"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
"gitea.maximumdirect.net/eric/notarius/internal/framework/pipeline"
"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd"
"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/shared/diagnostics"
)
func TestValidatorUsesOnlyCitedTranscriptEvidence(t *testing.T) {
value := occurrenceList("O'Rin Gate")
doc := &source.SourceDocument{ID: "session", Units: []source.SourceUnit{{ID: 1, Text: "The party enters orin gate."}, {ID: 2, Text: "No location here."}}}
result, err := New(Options{}).Validate(context.Background(), request(doc, value, contracts.ReferenceSet{}))
if err != nil || !result.Approved || len(result.Warnings) != 0 {
if err != nil || !result.Approved || len(result.Diagnostics) != 0 {
t.Fatalf("cited match = %#v, %v", result, err)
}
value = occurrenceList("Glossary Keep")
before := cloneList(value)
references := contracts.ReferenceSet{Slots: map[string]contracts.ResolvedReferenceSlot{"glossary": {Items: []contracts.ReferenceItem{{Content: []byte("Glossary Keep")}}}}}
result, err = New(Options{}).Validate(context.Background(), request(doc, value, references))
if err != nil || !result.Approved || len(result.Warnings) != 1 || result.Warnings[0].ReasonCode != WarningReasonCode || !strings.Contains(result.Warnings[0].Message, "Glossary Keep") || !reflect.DeepEqual(value, before) {
if err != nil || !result.Approved || len(result.Warnings) != 0 || len(result.Diagnostics) != 1 || result.Diagnostics[0].ReasonCode != ReasonCode || !strings.Contains(result.Diagnostics[0].Samples[0].Message, "Glossary Keep") || !reflect.DeepEqual(value, before) {
t.Fatalf("reference-only match = %#v, %v", result, err)
}
}
@@ -34,16 +33,16 @@ func TestValidatorDefersUnreadableEvidenceAndBoundsWarnings(t *testing.T) {
invalid.Occurrences[0].SourceRefs[0].StartUnitID = 99
invalid.Occurrences[0].SourceRefs[0].EndUnitID = 99
result, err := New(Options{}).Validate(context.Background(), request(document(), invalid, contracts.ReferenceSet{}))
if err != nil || !result.Approved || len(result.Warnings) != 0 {
if err != nil || !result.Approved || len(result.Diagnostics) != 0 {
t.Fatalf("unreadable evidence = %#v, %v", result, err)
}
occurrences := make([]dnd.LocationOccurrence, diagnostics.MaxWarnings+1)
occurrences := make([]dnd.LocationOccurrence, contracts.MaxDiagnosticSamples+2)
for index := range occurrences {
occurrences[index] = occurrenceList("Missing").Occurrences[0]
}
result, err = New(Options{}).Validate(context.Background(), request(document(), dnd.LocationOccurrenceList{Occurrences: occurrences}, contracts.ReferenceSet{}))
if err != nil || !result.Approved || len(result.Warnings) != diagnostics.MaxWarnings || result.Warnings[len(result.Warnings)-1].ReasonCode != OmittedReasonCode {
t.Fatalf("bounded warnings = %#v, %v", result, err)
if err != nil || !result.Approved || len(result.Warnings) != 0 || len(result.Diagnostics) != 1 || result.Diagnostics[0].OccurrenceCount != len(occurrences) || len(result.Diagnostics[0].Samples) != contracts.MaxDiagnosticSamples {
t.Fatalf("bounded advisories = %#v, %v", result, err)
}
}