Skip to main content

internal/infra/pg/citationstore.go

internal/infra/pg · 236 lines · 10 declarations · source

Declarations​

const MaxCitationEvidence, DefaultCitationEvidence, MaxCitationTextBytes and 1 more​

const (
	MaxCitationEvidence       = 32
	DefaultCitationEvidence   = 8
	MaxCitationTextBytes      = 256 << 10
	MaxCitationStatementBytes = 64 << 10
)

source

var ErrCitationNotFound, ErrCitationLimit, ErrCitationIntegrity and 1 more​

var (
	ErrCitationNotFound  = errors.New("citation not found")
	ErrCitationLimit     = errors.New("citation source exceeds supported text bounds; use the subject export")
	ErrCitationIntegrity = errors.New("citation evidence does not match its source")
	ErrInvalidCitation   = errors.New("invalid citation identifier or page")
)

source

type CitationCursor​

type CitationCursor struct {
	ObservationID string `json:"source_observation_id"`
	Ordinal       int    `json:"source_ordinal"`
	ByteStart     int    `json:"byte_start"`
}

source

type CitationInterval​

type CitationInterval struct {
	From           *time.Time `json:"from"`
	Until          *time.Time `json:"until"`
	FromInclusive  bool       `json:"from_inclusive"`
	UntilInclusive bool       `json:"until_inclusive"`
}

source

type CitationSource​

type CitationSource struct {
	AuthoredBy       *string   `json:"authored_by,omitempty"`
	ObservationID    string    `json:"source_observation_id"`
	Ordinal          int       `json:"source_ordinal"`
	LogOffset        int64     `json:"log_offset"`
	Role             string    `json:"role"`
	OccurredAt       time.Time `json:"occurred_at"`
	Quote            string    `json:"quote"`
	ByteStart        int       `json:"byte_start"`
	ByteEnd          int       `json:"byte_end"`
	ExtractorVersion string    `json:"extractor_version"`
	Context          string    `json:"context"`
	ContextStart     int       `json:"context_byte_start"`
	ContextEnd       int       `json:"context_byte_end"`
	ContextComplete  bool      `json:"context_complete"`
}

source

type Citation​

type Citation struct {
	Generation         *GenerationLink    `json:"generation,omitempty"`
	ReinterpretedBy    *GenerationLink    `json:"reinterpreted_by,omitempty"`
	Version            string             `json:"version"`
	Retraction         *RetractionDetails `json:"retraction,omitempty"`
	ID                 string             `json:"id"`
	SupersededBy       *string            `json:"superseded_by,omitempty"`
	SupersessionSource *string            `json:"supersession_source_observation_id,omitempty"`
	Scope              string             `json:"scope"`
	SubjectID          *string            `json:"subject_entity_id"`
	ObjectID           *string            `json:"object_entity_id"`
	Predicate          string             `json:"predicate"`
	Statement          string             `json:"statement"`
	Confidence         float32            `json:"confidence"`
	SourceRole         string             `json:"source_role"`
	RecordedAt         time.Time          `json:"recorded_at"`
	Valid              CitationInterval   `json:"valid"`
	Known              CitationInterval   `json:"known"`
	Status             string             `json:"status"`
	Evidence           []CitationSource   `json:"evidence"`
	Next               *CitationCursor    `json:"next,omitempty"`
}

Citation is the currently retained record and one page of exact supporting sources. The stored intervals describe the retained state; historical recall selects earlier versions.

source

type CitationStore​

type CitationStore struct {
	pool   *pgxpool.Pool
	schema Schema
}

source

func NewCitationStore​

func NewCitationStore(pool *pgxpool.Pool, schema Schema) *CitationStore

source

method CitationStore.Resolve​

func (s *CitationStore) Resolve(ctx context.Context, scope, id string, after *CitationCursor, limit int) (Citation, error)

source

const citationEvidenceSQL​

const citationEvidenceSQL = `SELECT e.source_observation_id::text,e.source_ordinal,e.byte_start,e.byte_end,
 octet_length(e.quote),octet_length(m.content),octet_length(e.extractor_version),
 CASE WHEN octet_length(e.quote)<=$4 THEN e.quote END,
 CASE WHEN octet_length(e.extractor_version)<=$6 THEN e.extractor_version END,
 m.role,m.occurred_at,o.log_offset,
 CASE WHEN octet_length(m.content)<=$5 AND e.byte_start>=0 AND e.byte_end<=octet_length(m.content)
 THEN substring(convert_to(m.content,'UTF8') from e.byte_start+1 for e.byte_end-e.byte_start) END,
 CASE WHEN octet_length(m.content)<=$5 AND e.byte_start>=0 AND e.byte_end<=octet_length(m.content)
 THEN substring(convert_to(m.content,'UTF8') from greatest(0,e.byte_start-$7)+1 for e.byte_end+$7-greatest(0,e.byte_start-$7)) END,
 greatest(0,e.byte_start-$7),c.principal_id::text
 FROM {schema}.fact_evidence e
 LEFT JOIN {schema}.observation o ON o.observation_id=e.source_observation_id AND o.scope=e.scope
 LEFT JOIN {schema}.turn_message m ON m.observation_id=o.observation_id AND m.ordinal=e.source_ordinal
 LEFT JOIN {schema}.curated_claim c ON c.scope=e.scope AND c.source_observation_id=e.source_observation_id
 WHERE e.scope=$1 AND e.fact_id=$2 {cursor}
 ORDER BY e.source_observation_id,e.source_ordinal,e.byte_start LIMIT $3`

source