Documentation
¶
Index ¶
- Variables
- func AllObjectsGranted(ctx context.Context, db storage.DBTX, subject uuid.UUID, objects []uuid.UUID, ...) (bool, error)
- func BatchDelete(ctx context.Context, tx storage.Transaction, keys []Key) error
- type Cursor
- type Grant
- type Key
- type Kind
- type ReadQuery
- type Relationship
- type ScanQuery
- type ScannedRelationship
Constants ¶
This section is empty.
Variables ¶
var ErrDuplicateGrant = errors.New("duplicate (subject, object, kind) in the batch")
ErrDuplicateGrant means a batch held two grants with the same (Subject, Object, Kind). The partial unique index allows one DIRECT role per grant, so the upsert touches a row twice and the DB rejects it; the handler maps this to InvalidArgument.
var ErrInvalidCursor = errors.New("invalid cursor")
ErrInvalidCursor means a page token failed to decode, or was minted for a different query than the one it's being used to page through.
var ErrNoObjects = errors.New("store: read requires at least one object")
ErrNoObjects means a read was requested without any object to scope it to.
Functions ¶
func AllObjectsGranted ¶ added in v1.18.0
func BatchDelete ¶ added in v1.33.0
BatchDelete removes the rows matching keys, in statements of at most maxBatchSize keys. Deletion is idempotent: a key with no matching row is not an error. BatchDelete takes the transaction it runs in so the delete stays atomic across chunks.
Types ¶
type Cursor ¶ added in v1.27.0
Cursor is the keyset position of the last row returned by a scan.
func DecodeCursor ¶ added in v1.30.0
DecodeCursor reverses Encode. An empty token is the first page, so it returns a nil Cursor and no error. fingerprint must match the one the token was minted with, or DecodeCursor returns ErrInvalidCursor.
type Key ¶ added in v1.33.0
Key identifies a relationship row to delete by (subject, object, kind); no role, since that key is already unique per DIRECT row.
type Kind ¶
type Kind int32
Kind mirrors relationshipspb.Kind but keeps this package free of a proto dependency; the handler maps the proto enum onto it.
type Relationship ¶
type Relationship struct {
Subject uuid.UUID
Object uuid.UUID
Kind Kind
// EdgeType is the stored edge class (1=DIRECT). This API only writes DIRECT
// rows; it is mapped through so the struct mirrors the row for future readers.
EdgeType int32
Role uuid.UUID
CreatedAt time.Time
}
func BatchUpsert ¶ added in v1.14.0
func BatchUpsert(ctx context.Context, tx storage.Transaction, grants []Grant) ([]Relationship, error)
BatchUpsert writes the grants in statements of at most maxBatchSize rows and returns the stored rows in the same order as grants. The write stays atomic across chunks: BatchUpsert takes the transaction it runs in, and on any error the caller rolls back, so the rows written by earlier chunks never commit. A batch must not contain two grants with the same (Subject, Object, Kind); BatchUpsert returns ErrDuplicateGrant.
type ScanQuery ¶ added in v1.27.0
type ScanQuery struct {
OrgID string
Objects []uuid.UUID
Kinds []Kind
Roles []uuid.UUID
After *Cursor // nil reads the first page
Limit int
}
ScanQuery is the input to LookupByOrg. After is nil for the first page. Limit must be between 1 and maxPageSize.
type ScannedRelationship ¶ added in v1.27.0
type ScannedRelationship struct {
Subject uuid.UUID
Object uuid.UUID
Kind Kind
Role uuid.UUID
CreatedAt time.Time
LocalID string
}
ScannedRelationship is one row of an org scan, plus the LocalID of the subject's identity in that org.
func LookupByOrg ¶ added in v1.27.0
func LookupByOrg(ctx context.Context, db storage.DBTX, query ScanQuery) ([]ScannedRelationship, error)
LookupByOrg returns one page of an org's relationships, keyset-paginated. It is read-only, so it takes a DBTX.
Source Files
Directories