ioutils

package
v1.0.1 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Apr 27, 2022 License: Apache-2.0 Imports: 15 Imported by: 11

Documentation

Overview

Package ioutils contains utilities for interacting with the IO Layer of FlytePropeller Metastore For example, utilities like reading inputs, writing outputs, computing output paths, prefixes. These helpers are intended to be used by FlytePropeller and aim to reduce the burden of implementing simple io functions

Index

Constants

View Source 
const (
	// InputsSuffix specifies the name of the file that contains the task inputs in the form core.LiteralMap
	InputsSuffix = "inputs.pb"
	// TaskTemplateSuffix In case a task requests for a task template, it is passed into the task using this filename.
	// The format is of type core.TaskTemplate
	TaskTemplateSuffix = "task.pb"
	// FuturesSuffix specifies that for dynamic workflows, the futures files is written with this filename/suffix.
	// The format is core.DynamicJobSpec
	FuturesSuffix = "futures.pb"
	// OutputsSuffix specifies that outputs are assumed to be written to this "file"/"suffix" under the given prefix
	// The outputs file has a format of core.LiteralMap
	OutputsSuffix = "outputs.pb"
	// ErrorsSuffix specifies that the errors are written to this prefix/file under the given prefix. The Error File
	// has a format of core.ErrorDocument
	ErrorsSuffix      = "error.pb"
	IndexLookupSuffix = "indexlookup.pb"
	// CheckpointPrefix specifies the common prefix that can be used as a directory where all the checkpoint information
	// will be stored. This directory is under the raw output-prefix path
	CheckpointPrefix = "_flytecheckpoints"
)
View Source 
const (
	ErrFailedRead string = "READ_FAILED"
)

Variables

This section is empty.

Functions

func ConstructCheckpointPath added in v0.8.0 

func ConstructCheckpointPath(store storage.ReferenceConstructor, rawOutputPrefix storage.DataReference) storage.DataReference

ConstructCheckpointPath returns a checkpoint path under the given `base` / rawOutputPrefix, following the conventions of the store

func GenerateAlphabet added in v0.5.29 

func GenerateAlphabet(b []rune) []rune

Generates the entire latin alphabet and appends it to the passed in array and returns the new array

func GenerateArabicNumerals added in v0.5.29 

func GenerateArabicNumerals(b []rune) []rune

Generates all arabic numerals and appends to the passed in array and returns the new array/slice

func GetIndexLookupPath 

func GetIndexLookupPath(ctx context.Context, store storage.ReferenceConstructor, prefix storage.DataReference) (res storage.DataReference, err error)

GetIndexLookupPath returns the indexpath suffixed to IndexLookupSuffix

func GetTaskTemplatePath added in v0.5.41 

func GetTaskTemplatePath(ctx context.Context, store storage.ReferenceConstructor, base storage.DataReference) (storage.DataReference, error)

GetTaskTemplatePath returns a protobuf file path where TaskTemplate is stored

func NewCachedInputReader 

func NewCachedInputReader(ctx context.Context, in io.InputReader) io.InputReader

Creates a new Read-through cached Input Reader. the returned reader is not thread-safe It caches the inputs on a successful read from the underlying input reader

func NewDeterministicUniqueRawOutputPath added in v0.5.29 

func NewDeterministicUniqueRawOutputPath(ctx context.Context, rawOutputPrefix, outputMetadataPrefix storage.DataReference, store storage.ReferenceConstructor) (io.RawOutputPaths, error)

Constructs an output path that is deterministic and unique within the given outputPrefix. No sharding is performed

func NewLazyUploadingTaskReader added in v0.5.41 

func NewLazyUploadingTaskReader(baseTaskReader SimpleTaskReader, remotePath storage.DataReference, store storage.ProtobufStore) pluginsCore.TaskReader

NewLazyUploadingTaskReader decorates an existing TaskReader and adds a functionality to allow lazily uploading the task template to a remote location, only when the location information is accessed

func NewRawOutputPaths added in v0.5.29 

func NewRawOutputPaths(_ context.Context, rawOutputPrefix storage.DataReference) io.RawOutputPaths

A simple Output sandbox at a given path

func NewShardedDeterministicRawOutputPath added in v0.5.29 

func NewShardedDeterministicRawOutputPath(ctx context.Context, sharder ShardSelector, basePrefix, outputMetadataPrefix storage.DataReference, store storage.ReferenceConstructor) (io.RawOutputPaths, error)

Creates a deterministic RawOutputPath whose path is distributed based on the ShardSelector passed in. Determinism depends on the outputMetadataPath Potential performance problem, as creating a new RawPath creation may be expensive as it hashes the outputMetadataPath the final RawOutputPath is created in the shard selected by the sharder at the basePath and then appended by a hashed value of the outputMetadata

func NewShardedRawOutputPath added in v0.5.29 

func NewShardedRawOutputPath(ctx context.Context, sharder ShardSelector, basePath storage.DataReference, uniqueID string, store storage.ReferenceConstructor) (io.RawOutputPaths, error)

Creates an OutputSandbox in the basePath using the uniqueID and a sharder This implementation is faster than the Randomized strategy

func NewTaskIDRawOutputPath added in v0.5.29 

func NewTaskIDRawOutputPath(ctx context.Context, rawOutputPrefix storage.DataReference, taskID *core2.TaskExecutionIdentifier, store storage.ReferenceConstructor) (io.RawOutputPaths, error)

Generates a RawOutput Path that looks like the TaskExecutionID and can be easily cross referenced with Flyte generated TaskExecution ID

Types

type BufferedOutputWriter 

type BufferedOutputWriter struct {
	io.OutputFilePaths
	// contains filtered or unexported fields
}

A Buffered outputWriter just records the io.OutputReader and can be accessed using special methods.

func NewBufferedOutputWriter 

func NewBufferedOutputWriter(ctx context.Context, paths io.OutputFilePaths) *BufferedOutputWriter

Returns a new object of type BufferedOutputWriter

func (*BufferedOutputWriter) GetReader 

func (o *BufferedOutputWriter) GetReader() io.OutputReader

func (*BufferedOutputWriter) Put 

func (o *BufferedOutputWriter) Put(ctx context.Context, reader io.OutputReader) error

type InMemoryOutputReader 

type InMemoryOutputReader struct {
	// contains filtered or unexported fields
}

func NewInMemoryOutputReader 

func NewInMemoryOutputReader(literals *core.LiteralMap, err *io.ExecutionError) InMemoryOutputReader

func (InMemoryOutputReader) Exists 

func (r InMemoryOutputReader) Exists(ctx context.Context) (bool, error)

func (InMemoryOutputReader) IsError 

func (r InMemoryOutputReader) IsError(ctx context.Context) (bool, error)

func (InMemoryOutputReader) IsFile 

func (r InMemoryOutputReader) IsFile(ctx context.Context) bool

func (InMemoryOutputReader) Read 

func (r InMemoryOutputReader) Read(ctx context.Context) (*core.LiteralMap, *io.ExecutionError, error)

func (InMemoryOutputReader) ReadError 

func (r InMemoryOutputReader) ReadError(ctx context.Context) (io.ExecutionError, error)

type PrecomputedShardSelector added in v0.5.29 

type PrecomputedShardSelector struct {
	// contains filtered or unexported fields
}

this sharder distributes data into one of the precomputed buckets. The bucket is deterministically determined given the input s

func (*PrecomputedShardSelector) GetShardPrefix added in v0.5.29 

func (d *PrecomputedShardSelector) GetShardPrefix(_ context.Context, s []byte) (string, error)

Generates deterministic shard id for the given string s

type RemoteCheckpointPaths added in v0.8.0 

type RemoteCheckpointPaths struct {
	RemoteFileOutputPaths
	// contains filtered or unexported fields
}

RemoteCheckpointPaths implements the CheckpointPaths Interface and adds on top of the OutputFilePaths Interface

func NewCheckpointRemoteFilePaths added in v0.8.0 

func NewCheckpointRemoteFilePaths(ctx context.Context, store storage.ReferenceConstructor, outputPrefix storage.DataReference, sandbox io.RawOutputPaths, previousCheckpointPath storage.DataReference) RemoteCheckpointPaths

NewCheckpointRemoteFilePaths returns a new object constructed with an optional previousCheckpointPath and derives a new checkpointPath from the outputPrefix

func NewReadOnlyOutputFilePaths added in v0.8.0 

func NewReadOnlyOutputFilePaths(ctx context.Context, store storage.ReferenceConstructor, outputPrefix storage.DataReference) RemoteCheckpointPaths

NewReadOnlyOutputFilePaths can be used when data is only to be read from an existing remote location

func (RemoteCheckpointPaths) GetCheckpointPrefix added in v0.8.0 

func (r RemoteCheckpointPaths) GetCheckpointPrefix() storage.DataReference

GetCheckpointPrefix returns a new checkpoint path under the raw output prefix.

func (RemoteCheckpointPaths) GetPreviousCheckpointsPrefix added in v0.8.0 

func (r RemoteCheckpointPaths) GetPreviousCheckpointsPrefix() storage.DataReference

GetPreviousCheckpointsPrefix returns the Prefix path for checkpoints for the previous attempt, or "" if this is the first attempt

type RemoteFileInputReader 

type RemoteFileInputReader struct {
	io.InputFilePaths
	// contains filtered or unexported fields
}

func NewRemoteFileInputReader 

func NewRemoteFileInputReader(_ context.Context, store storage.ProtobufStore, inputPaths io.InputFilePaths) RemoteFileInputReader

func (RemoteFileInputReader) Get 

func (r RemoteFileInputReader) Get(ctx context.Context) (*core.LiteralMap, error)

type RemoteFileOutputPaths 

type RemoteFileOutputPaths struct {

	// Arbitrary supplied of the RawOutputPath
	io.RawOutputPaths
	// contains filtered or unexported fields
}

RemoteFileOutputPaths records all metadata output paths / keys on a remote storage system, e.g. S3 / GCS or any other key-value store. Theoretically if the storage.DataReference can support BigTable, this will work with it.

func NewRemoteFileOutputPaths 

func NewRemoteFileOutputPaths(_ context.Context, store storage.ReferenceConstructor, outputPrefix storage.DataReference, sandbox io.RawOutputPaths) RemoteFileOutputPaths

NewRemoteFileOutputPaths returns a RemoteFileOutputPaths object, where all the paths are configured using the given outputPrefix and constructed using the storage.ReferenceConstructor

func (RemoteFileOutputPaths) GetErrorPath 

func (w RemoteFileOutputPaths) GetErrorPath() storage.DataReference

func (RemoteFileOutputPaths) GetFuturesPath 

func (w RemoteFileOutputPaths) GetFuturesPath() storage.DataReference

func (RemoteFileOutputPaths) GetOutputPath 

func (w RemoteFileOutputPaths) GetOutputPath() storage.DataReference

func (RemoteFileOutputPaths) GetOutputPrefixPath 

func (w RemoteFileOutputPaths) GetOutputPrefixPath() storage.DataReference

type RemoteFileOutputReader 

type RemoteFileOutputReader struct {
	// contains filtered or unexported fields
}

func NewRemoteFileOutputReader 

func NewRemoteFileOutputReader(_ context.Context, store storage.ComposedProtobufStore, outPaths io.OutputFilePaths, maxDatasetSize int64) RemoteFileOutputReader

func (RemoteFileOutputReader) Exists 

func (r RemoteFileOutputReader) Exists(ctx context.Context) (bool, error)

func (RemoteFileOutputReader) IsError 

func (r RemoteFileOutputReader) IsError(ctx context.Context) (bool, error)

func (RemoteFileOutputReader) IsFile 

func (r RemoteFileOutputReader) IsFile(ctx context.Context) bool

func (RemoteFileOutputReader) Read 

func (r RemoteFileOutputReader) Read(ctx context.Context) (*core.LiteralMap, *io.ExecutionError, error)

func (RemoteFileOutputReader) ReadError 

func (r RemoteFileOutputReader) ReadError(ctx context.Context) (io.ExecutionError, error)

type RemoteFileOutputWriter 

type RemoteFileOutputWriter struct {
	io.OutputFilePaths
	// contains filtered or unexported fields
}

RemoteFileOutputWriter adds storage Write APIs to output paths / keys. In retrospect, the `path` should be generally replaced with keys

func NewRemoteFileOutputWriter 

func NewRemoteFileOutputWriter(_ context.Context, store storage.ProtobufStore, outputFilePaths io.OutputFilePaths) RemoteFileOutputWriter

NewRemoteFileOutputWriter returns a writer that records all outputs to remote files / objects. Given outputs, it will automatically write it to the outputFile / key that is configured.

func (RemoteFileOutputWriter) Put 

func (w RemoteFileOutputWriter) Put(ctx context.Context, reader io.OutputReader) error

type ShardSelector added in v0.5.29 

type ShardSelector interface {
	GetShardPrefix(ctx context.Context, s []byte) (string, error)
}

This interface allows shard selection for OutputSandbox.

func NewBase36PrefixShardSelector added in v0.5.29 

func NewBase36PrefixShardSelector(ctx context.Context) (ShardSelector, error)

Creates a PrecomputedShardSelector with 36*36 unique shards. Each shard is of the format {[0-9a-z][0-9a-z]}, i.e. 2 character long.

func NewConstantShardSelector added in v0.5.29 

func NewConstantShardSelector(shards []string) ShardSelector

uses the given shards to select a shard

type SimpleInputFilePath 

type SimpleInputFilePath struct {
	// contains filtered or unexported fields
}

func NewInputFilePaths 

func NewInputFilePaths(_ context.Context, store storage.ReferenceConstructor, inputPathPrefix storage.DataReference) SimpleInputFilePath

func (SimpleInputFilePath) GetInputPath 

func (s SimpleInputFilePath) GetInputPath() storage.DataReference

func (SimpleInputFilePath) GetInputPrefixPath 

func (s SimpleInputFilePath) GetInputPrefixPath() storage.DataReference

type SimpleTaskReader added in v0.5.41 

type SimpleTaskReader interface {
	Read(ctx context.Context) (*core.TaskTemplate, error)
}

SimpleTaskReader provides only the TaskReader interface. This is created to conveniently use the uploading taskreader interface

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.