Documentation
¶
Overview ¶
Package corpus defines the neutral record unit that Stroma indexes.
Index ¶
Constants ¶
const ( // DefaultKind is used when callers omit a more specific record kind. DefaultKind = "artifact" // FormatMarkdown stores Markdown bodies that can be heading-chunked. FormatMarkdown = "markdown" // FormatPlaintext stores plain text bodies as a single chunk. FormatPlaintext = "plaintext" )
Variables ¶
This section is empty.
Functions ¶
func Fingerprint ¶
Fingerprint summarizes a set of records deterministically and returns an error if any record fails Normalize. Prior versions silently skipped invalid records, which meant "corpus with an invalid record" produced the same digest as "same corpus without that record" — masking silent reuse of a snapshot that was missing data the caller thought they had. The loud failure forces callers to surface the problem instead.
func FingerprintFromPairs ¶
FingerprintFromPairs returns the same digest as Fingerprint([]Record) for already-normalized (Ref, ContentHash) inputs: non-empty after trimming, with ContentHash already computed. Pairs with an empty Ref or ContentHash return an error — unlike Fingerprint([]Record), this helper cannot apply Normalize() defaults or regenerate ContentHash via HashRecord from other fields, so its output only matches Fingerprint when the inputs already satisfy that invariant. Callers reading persisted rows must enforce the invariant at read time (as index.loadCurrentRefHashes does) or use Fingerprint instead.
func HashRecord ¶
HashRecord returns a deterministic content hash for a normalized record.
Each field contributes a `%q=%q` pair so the encoding is injective: no combination of key/value strings (including ones that contain `=` or newline characters) can produce the same serialized prefix as a different field. Serialized parts are then sorted and SHA-256-joined in fingerprint. Changing this encoding requires a schema_version bump and a migration that rewrites persisted content_hash values (see migrateV3ToV4).
Types ¶
type Record ¶
type Record struct {
Ref string
Kind string
Title string
SourceRef string
BodyFormat string
BodyText string
ContentHash string
Metadata map[string]string
}
Record is the neutral corpus unit Stroma indexes.
func NewRecord ¶
NewRecord constructs a Record for the common path: a reference, a human title, and a Markdown body. All other fields default through Normalize at the point the record enters a Rebuild / Update call. Callers with non-default kinds, custom metadata, or plaintext bodies should build the struct directly and pass it through Normalize.
func (Record) Normalize ¶
Normalize returns a trimmed, validated record with safe defaults applied: missing Kind defaults to DefaultKind, Title and SourceRef default to Ref, BodyFormat defaults to FormatMarkdown, and ContentHash is regenerated from HashRecord when empty. The returned record satisfies Validate by construction.
This is the entry point callers should use before invoking Validate. Calling Validate on a raw Record will reject records that would have been accepted after defaults — for example Record{Ref:"x"} fails Validate (no kind) but Normalize fills in DefaultKind and succeeds.
func (Record) Normalized
deprecated
Normalized is a deprecated alias for Normalize retained so downstream callers that upgrade without switching name immediately keep compiling. New code should use Normalize. Scheduled for removal one release after adoption settles.
Deprecated: use Normalize.
func (Record) Validate ¶
Validate reports whether the record is complete enough to persist. Callers constructing a Record by hand should prefer Normalize, which applies safe defaults before validation — calling Validate directly on a raw struct will reject records that Normalize would have accepted (e.g. a missing Kind).
Source Files