Documentation
¶
Index ¶
- Variables
- func Drop(driver string, config map[string]any, logger *zap.Logger) error
- func NewPermissionDeniedError(msg string) error
- func RecordDownloadMetrics(ctx context.Context, m *DownloadMetrics)
- func Register(name string, driver Driver)
- func RegisterAsConnector(name string, driver Driver)
- type CatalogStore
- type Dialect
- type DownloadMetrics
- type Driver
- type FileIterator
- type FileStore
- type Handle
- type InformationSchema
- type IngestionSummary
- type Instance
- type NoOpProgress
- type OLAPStore
- type ObjectStore
- type ObjectType
- type PermissionDeniedError
- type Progress
- type ProgressUnit
- type PropertySchema
- type PropertySchemaType
- type QueryOption
- type RegistryStore
- type RepoObjectStat
- type RepoStore
- type Resource
- type Result
- type RowIterator
- type SQLStore
- type Spec
- type Statement
- type Table
- type TransferOptions
- type Transporter
- type WatchCallback
- type WatchEvent
- type WithConnectionFunc
Constants ¶
This section is empty.
Variables ¶
var Connectors = make(map[string]Driver)
Connectors tracks all registered connector drivers.
var Drivers = make(map[string]Driver)
Drivers is a registry of drivers.
var ErrDropNotSupported = errors.New("driver: drop not supported")
ErrDropNotSupported indicates the driver doesn't support dropping its underlying store.
var ErrFileAlreadyExists = errors.New("file already exists")
var ErrInconsistentControllerVersion = errors.New("controller: inconsistent version")
ErrInconsistentControllerVersion is returned from Controller when an unexpected controller version is observed in the DB. An unexpected controller version will only be observed if multiple controllers are running simultanesouly (split brain).
var ErrIteratorDone = errors.New("empty iterator")
var ErrNotFound = errors.New("driver: not found")
ErrNotFound indicates the resource wasn't found.
var ErrNotImplemented = errors.New("driver: not implemented")
var ErrResourceAlreadyExists = errors.New("controller: resource already exists")
ErrResourceAlreadyExists is returned from catalog functions when attempting to create a resource that already exists.
var ErrResourceNotFound = errors.New("controller: resource not found")
ErrResourceNotFound is returned from catalog functions when a referenced resource does not exist.
var ErrStorageLimitExceeded = fmt.Errorf("connectors: exceeds storage limit")
var ErrUnsupportedConnector = errors.New("drivers: connector not supported")
ErrUnsupportedConnector is returned from Ingest for unsupported connectors.
Functions ¶
func Drop ¶ added in v0.27.0
Drop tears down a store. Drivers that do not support it return ErrDropNotSupported.
func NewPermissionDeniedError ¶ added in v0.30.0
func RecordDownloadMetrics ¶ added in v0.30.0
func RecordDownloadMetrics(ctx context.Context, m *DownloadMetrics)
func RegisterAsConnector ¶ added in v0.30.0
RegisterAsConnector tracks a connector driver.
Types ¶
type CatalogStore ¶
type CatalogStore interface {
NextControllerVersion(ctx context.Context) (int64, error)
CheckControllerVersion(ctx context.Context, v int64) error
FindResources(ctx context.Context) ([]Resource, error)
CreateResource(ctx context.Context, v int64, r Resource) error
UpdateResource(ctx context.Context, v int64, r Resource) error
DeleteResource(ctx context.Context, v int64, k, n string) error
DeleteResources(ctx context.Context) error
}
CatalogStore is implemented by drivers capable of storing catalog info for a specific instance. Implementations should treat resource kinds as case sensitive, but resource names as case insensitive.
type DownloadMetrics ¶ added in v0.30.0
type Driver ¶
type Driver interface {
Spec() Spec
// Open opens a new connection to an underlying store.
Open(config map[string]any, shared bool, client activity.Client, logger *zap.Logger) (Handle, error)
// Drop tears down a store. Drivers that do not support it return ErrDropNotSupported.
Drop(config map[string]any, logger *zap.Logger) error
// HasAnonymousSourceAccess returns true if external system can be accessed without credentials
HasAnonymousSourceAccess(ctx context.Context, src map[string]any, logger *zap.Logger) (bool, error)
// TertiarySourceConnectors returns a list of connectors required to resolve a source, excluding the source and sink connectors.
TertiarySourceConnectors(ctx context.Context, src map[string]any, logger *zap.Logger) ([]string, error)
}
Driver represents an underlying DB.
type FileIterator ¶ added in v0.30.0
type FileIterator interface {
// Close do cleanup and release resources
Close() error
// Next returns a list of file downloaded from external sources
// and cleanups file created in previous batch
Next() ([]string, error)
// Size returns size of data downloaded in unit.
// Returns 0,false if not able to compute size in given unit
Size(unit ProgressUnit) (int64, bool)
// KeepFilesUntilClose marks the iterator to keep the files until close is called.
// This is used when the entire list of files is used at once in certain cases.
KeepFilesUntilClose(keepFilesUntilClose bool)
// Format returns general file format (json, csv, parquet, etc)
// Returns an empty string if there is no general format
Format() string
}
FileIterator provides ways to iteratively download files from external sources Clients should call close once they are done with iterator to release any resources
type Handle ¶ added in v0.32.0
type Handle interface {
// Driver type (like "duckdb")
Driver() string
// Config used to open the Connection
Config() map[string]any
// Migrate prepares the connection for use. It will be called before the connection is first used.
// (Not to be confused with migrating artifacts, which is handled by the runtime and tracked in the catalog.)
Migrate(ctx context.Context) error
// MigrationStatus returns the connection's current and desired migration version (if applicable)
MigrationStatus(ctx context.Context) (current int, desired int, err error)
// Close closes the connection
Close() error
// AsRegistry returns a AsRegistry if the driver can serve as such, otherwise returns false.
// The registry is responsible for tracking instances and repos.
AsRegistry() (RegistryStore, bool)
// AsCatalogStore returns a AsCatalogStore if the driver can serve as such, otherwise returns false.
// A catalog is used to store state about migrated/deployed objects (such as sources and metrics views).
AsCatalogStore(instanceID string) (CatalogStore, bool)
// AsRepoStore returns a AsRepoStore if the driver can serve as such, otherwise returns false.
// A repo stores file artifacts (either in a folder or virtualized in a database).
AsRepoStore(instanceID string) (RepoStore, bool)
// AsOLAP returns an AsOLAP if the driver can serve as such, otherwise returns false.
// OLAP stores are where we actually store, transform, and query users' data.
AsOLAP(instanceID string) (OLAPStore, bool)
// AsObjectStore returns an ObjectStore if the driver can serve as such, otherwise returns false.
AsObjectStore() (ObjectStore, bool)
// AsFileStore returns a Filetore if the driver can serve as such, otherwise returns false.
AsFileStore() (FileStore, bool)
// AsTransporter optionally returns an implementation for moving data between two connectors.
// One of the input connections may be the Connection itself.
// Examples:
// a) myDuckDB.AsTransporter(myGCS, myDuckDB)
// b) myBeam.AsTransporter(myGCS, myS3) // In the future
AsTransporter(from Handle, to Handle) (Transporter, bool)
// AsSQLStore returns a SQLStore if the driver can serve as such, otherwise returns false.
AsSQLStore() (SQLStore, bool)
}
Handle represents a connection to an underlying DB. It should implement one or more of RegistryStore, CatalogStore, RepoStore, and OLAPStore.
type InformationSchema ¶
type InformationSchema interface {
All(ctx context.Context) ([]*Table, error)
Lookup(ctx context.Context, name string) (*Table, error)
}
InformationSchema contains information about existing tables in an OLAP driver. Table lookups should be case insensitive.
type IngestionSummary ¶ added in v0.24.0
type IngestionSummary struct {
BytesIngested int64
}
IngestionSummary is details about ingestion
type Instance ¶
type Instance struct {
// Identifier
ID string
// Driver name to connect to for OLAP
OLAPConnector string
// Driver name for reading/editing code artifacts
RepoConnector string
// CreatedOn is when the instance was created
CreatedOn time.Time `db:"created_on"`
// UpdatedOn is when the instance was last updated in the registry
UpdatedOn time.Time `db:"updated_on"`
// Instance specific connectors
Connectors []*runtimev1.Connector `db:"connectors"`
// ProjectVariables contains default connectors from rill.yaml
ProjectConnectors []*runtimev1.Connector `db:"project_connectors"`
// Variables contains user-provided variables
Variables map[string]string `db:"variables"`
// ProjectVariables contains default variables from rill.yaml
// (NOTE: This can always be reproduced from rill.yaml, so it's really just a handy cache of the values.)
ProjectVariables map[string]string `db:"project_variables"`
// Annotations to enrich activity events (like usage tracking)
Annotations map[string]string
// EmbedCatalog tells the runtime to store the instance's catalog in its OLAP store instead
// of in the runtime's metadata store. Currently only supported for the duckdb driver.
EmbedCatalog bool `db:"embed_catalog"`
// WatchRepo indicates whether to watch the repo for file changes and reconcile them automatically.
WatchRepo bool `db:"watch_repo"`
// StageChanges indicates whether to stage source and model changes before applying them.
StageChanges bool `db:"stage_changes"`
// ModelDefaultMaterialize indicates whether to materialize models by default.
ModelDefaultMaterialize bool `db:"model_default_materialize"`
// ModelMaterializeDelaySeconds adds a delay before materializing models.
ModelMaterializeDelaySeconds uint32 `db:"model_materialize_delay_seconds"`
// IgnoreInitialInvalidProjectError indicates whether to ignore an invalid project error when the instance is initially created.
IgnoreInitialInvalidProjectError bool `db:"-"`
}
Instance represents a single data project, meaning one OLAP connection, one repo connection, and one catalog connection.
func (*Instance) ResolveVariables ¶ added in v0.23.0
ResolveVariables returns the final resolved variables
type NoOpProgress ¶ added in v0.30.0
type NoOpProgress struct{}
func (NoOpProgress) Observe ¶ added in v0.30.0
func (n NoOpProgress) Observe(val int64, unit ProgressUnit)
func (NoOpProgress) Target ¶ added in v0.30.0
func (n NoOpProgress) Target(val int64, unit ProgressUnit)
type OLAPStore ¶
type OLAPStore interface {
Dialect() Dialect
WithConnection(ctx context.Context, priority int, longRunning, tx bool, fn WithConnectionFunc) error
Exec(ctx context.Context, stmt *Statement) error
Execute(ctx context.Context, stmt *Statement) (*Result, error)
InformationSchema() InformationSchema
EstimateSize() (int64, bool)
CreateTableAsSelect(ctx context.Context, name string, view bool, sql string) error
InsertTableAsSelect(ctx context.Context, name string, byName bool, sql string) error
DropTable(ctx context.Context, name string, view bool) error
// RenameTable is force rename
RenameTable(ctx context.Context, name, newName string, view bool) error
AddTableColumn(ctx context.Context, tableName, columnName string, typ string) error
AlterTableColumn(ctx context.Context, tableName, columnName string, newType string) error
}
OLAPStore is implemented by drivers that are capable of storing, transforming and serving analytical queries. NOTE crud APIs are not safe to be called with `WithConnection`
type ObjectStore ¶ added in v0.30.0
type ObjectType ¶ added in v0.16.0
type ObjectType int
Constants representing the kinds of catalog objects.
const ( ObjectTypeUnspecified ObjectType = 0 ObjectTypeTable ObjectType = 1 ObjectTypeSource ObjectType = 2 ObjectTypeModel ObjectType = 3 ObjectTypeMetricsView ObjectType = 4 )
type PermissionDeniedError ¶ added in v0.30.0
type PermissionDeniedError struct {
// contains filtered or unexported fields
}
func (*PermissionDeniedError) Error ¶ added in v0.30.0
func (e *PermissionDeniedError) Error() string
type Progress ¶ added in v0.30.0
type Progress interface {
Target(val int64, unit ProgressUnit)
// Observe is used by caller to provide incremental updates
Observe(val int64, unit ProgressUnit)
}
Progress is an interface for communicating progress info
type ProgressUnit ¶ added in v0.30.0
type ProgressUnit int
const ( ProgressUnitByte ProgressUnit = iota ProgressUnitFile ProgressUnitRecord )
type PropertySchema ¶ added in v0.30.0
type PropertySchema struct {
Key string
Type PropertySchemaType
Required bool
DisplayName string
Description string
Placeholder string
// Default can be different from placeholder in the sense that placeholder should not be used as default value.
// If a default is set then it should also be used as a placeholder.
Default string
Hint string
Href string
Secret bool
ValidateFunc func(any interface{}) error
TransformFunc func(any interface{}) interface{}
}
PropertySchema provides the schema for a property supported by a connector.
func (PropertySchema) ValidateType ¶ added in v0.30.0
func (ps PropertySchema) ValidateType(val any) bool
ValidateType checks that val has the correct type.
type PropertySchemaType ¶ added in v0.30.0
type PropertySchemaType int
PropertySchemaType is an enum of types supported for connector properties.
const ( UnspecifiedPropertyType PropertySchemaType = iota StringPropertyType NumberPropertyType BooleanPropertyType InformationalPropertyType )
type QueryOption ¶ added in v0.33.2
type QueryOption struct {
// TotalLimitInBytes rerpresent the max limit on the bytes that should be downloaded in a file
TotalLimitInBytes int64
}
type RegistryStore ¶
type RegistryStore interface {
FindInstances(ctx context.Context) ([]*Instance, error)
FindInstance(ctx context.Context, id string) (*Instance, error)
CreateInstance(ctx context.Context, instance *Instance) error
DeleteInstance(ctx context.Context, id string) error
EditInstance(ctx context.Context, instance *Instance) error
}
RegistryStore is implemented by drivers capable of storing and looking up instances and repos.
type RepoObjectStat ¶ added in v0.15.0
type RepoStore ¶
type RepoStore interface {
Driver() string
// Root returns directory where artifacts are stored.
Root() string
CommitHash(ctx context.Context) (string, error)
ListRecursive(ctx context.Context, glob string) ([]string, error)
Get(ctx context.Context, path string) (string, error)
Stat(ctx context.Context, path string) (*RepoObjectStat, error)
Put(ctx context.Context, path string, reader io.Reader) error
Rename(ctx context.Context, fromPath string, toPath string) error
Delete(ctx context.Context, path string) error
Sync(ctx context.Context) error
Watch(ctx context.Context, cb WatchCallback) error
}
RepoStore is implemented by drivers capable of storing code artifacts. It mirrors a file system, but may be virtualized by a database for non-local deployments.
type Result ¶ added in v0.15.0
type Result struct {
*sqlx.Rows
Schema *runtimev1.StructType
// contains filtered or unexported fields
}
Result wraps the results of query.
func (*Result) Close ¶ added in v0.18.0
Close wraps rows.Close and calls the Result's cleanup function (if it is set). Close should be idempotent.
func (*Result) SetCleanupFunc ¶ added in v0.18.0
SetCleanupFunc sets a function, which will be called when the Result is closed.
type RowIterator ¶ added in v0.31.0
type RowIterator interface {
// Schema of the underlying data
Schema(ctx context.Context) (*runtimev1.StructType, error)
// Next fetches next row
Next(ctx context.Context) ([]driver.Value, error)
// Close closes the iterator and frees resources
Close() error
// Size returns total size of data downloaded in unit.
// Returns 0,false if not able to compute size in given unit
Size(unit ProgressUnit) (uint64, bool)
}
RowIterator returns an iterator to iterate over result of a sql query
type SQLStore ¶ added in v0.31.0
type SQLStore interface {
// Query returns driver.RowIterator to iterate over results row by row
Query(ctx context.Context, props map[string]any) (RowIterator, error)
// QueryAsFiles downloads results into files and returns an iterator to iterate over them
QueryAsFiles(ctx context.Context, props map[string]any, opt *QueryOption, p Progress) (FileIterator, error)
}
SQLStore is implemented by drivers capable of running sql queries and generating an iterator to consume results. In future the results can be produced in other formats like arrow as well. May be call it DataWarehouse to differentiate from OLAP or postgres?
type Spec ¶ added in v0.30.0
type Spec struct {
DisplayName string
Description string
ServiceAccountDocs string
SourceProperties []PropertySchema
ConfigProperties []PropertySchema
Help string
}
Spec provides metadata about a connector and the properties it supports.
type Statement ¶
type Statement struct {
Query string
Args []any
DryRun bool
Priority int
LongRunning bool
ExecutionTimeout time.Duration
}
Statement wraps a query to execute against an OLAP driver.
type Table ¶
type Table struct {
Database string
DatabaseSchema string
Name string
View bool
Schema *runtimev1.StructType
}
Table represents a table in an information schema.
type TransferOptions ¶ added in v0.34.1
type TransferOptions struct {
AllowHostAccess bool
RepoRoot string
Progress Progress
AcquireConnector func(string) (Handle, func(), error)
}
TransferOptions provide execution context for Transporter.Transfer
type Transporter ¶ added in v0.30.0
type Transporter interface {
Transfer(ctx context.Context, srcProps, sinkProps map[string]any, opts *TransferOptions) error
}
Transporter implements logic for moving data between two connectors (the actual connector objects are provided in AsTransporter)
type WatchCallback ¶ added in v0.30.0
type WatchCallback func(event []WatchEvent)
type WatchEvent ¶ added in v0.30.0
type WithConnectionFunc ¶ added in v0.18.0
type WithConnectionFunc func(wrappedCtx context.Context, ensuredCtx context.Context, conn *sql.Conn) error
WithConnectionFunc is a callback function that provides a context to be used in further OLAP store calls to enforce affinity to a single connection. It also provides pointers to the actual database/sql and database/sql/driver connections. It's called with two contexts: wrappedCtx wraps the input context (including cancellation), and ensuredCtx wraps a background context (ensuring it can never be cancelled).