feat(core): add replace partition
#512
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,330 @@ | ||
| // Licensed to the Apache Software Foundation (ASF) under one | ||
| // or more contributor license agreements. See the NOTICE file | ||
| // distributed with this work for additional information | ||
| // regarding copyright ownership. The ASF licenses this file | ||
| // to you under the Apache License, Version 2.0 (the | ||
| // "License"); you may not use this file except in compliance | ||
| // with the License. You may obtain a copy of the License at | ||
| // | ||
| // http://www.apache.org/licenses/LICENSE-2.0 | ||
| // | ||
| // Unless required by applicable law or agreed to in writing, | ||
| // software distributed under the License is distributed on an | ||
| // "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY | ||
| // KIND, either express or implied. See the License for the | ||
| // specific language governing permissions and limitations | ||
| // under the License. | ||
|
|
||
| package table | ||
|
|
||
| import ( | ||
| "context" | ||
| "errors" | ||
| "fmt" | ||
|
|
||
| "github.com/apache/iceberg-go" | ||
| "github.com/apache/iceberg-go/io" | ||
| "github.com/google/uuid" | ||
| ) | ||
|
|
||
| // ReplacePartitions is used to replace data in a table, partition by partition. | ||
| // | ||
| // This interface is similar to OverwriteFiles but only replaces specific partitions | ||
| // rather than files. Only data files that belong to partitions that match added files | ||
| // will be replaced. | ||
| // | ||
| // As an example, if you have a table partitioned by columns "year" and "month" and this | ||
| // action adds two data files with partition data (2021, 1) and (2021, 2), then all existing | ||
| // data files that contain records where year=2021 and month=1 OR year=2021 and month=2 | ||
| // will be replaced. | ||
| type ReplacePartitions interface { | ||
| // AddFile adds a data file to the table. | ||
| AddFile(dataFile iceberg.DataFile) ReplacePartitions | ||
|
|
||
| // AddFiles adds multiple data files to the table. | ||
| AddFiles(dataFiles ...iceberg.DataFile) ReplacePartitions | ||
|
|
||
| // ValidateAddedFilesMatchOverwriteFilter ensures that files added concurrently | ||
| // do not conflict with the replace operation. | ||
| ValidateAddedFilesMatchOverwriteFilter() ReplacePartitions | ||
|
|
||
| // ValidateFromSnapshot ensures that no delete files matching this operation | ||
| // are added between the snapshot used to plan the operation and the current snapshot. | ||
| ValidateFromSnapshot(snapshotID int64) ReplacePartitions | ||
|
|
||
| // Set sets a summary property in the snapshot produced by this update. | ||
| Set(property, value string) ReplacePartitions | ||
|
|
||
| // StageOnly stages the snapshot but does not update the current snapshot ID. | ||
| StageOnly() ReplacePartitions | ||
|
|
||
| // Commit executes this update and commits the result. | ||
| Commit(ctx context.Context) (*Table, error) | ||
| } | ||
|
|
||
| // BaseReplacePartitions implements the ReplacePartitions interface. | ||
| type BaseReplacePartitions struct { | ||
| txn *Transaction | ||
| addedFiles []iceberg.DataFile | ||
| snapshotProps iceberg.Properties | ||
| validateAddedFiles bool | ||
| fromSnapshotID *int64 | ||
| stageOnly bool | ||
| validatePartitionsFilter bool | ||
| } | ||
|
|
||
| // NewReplacePartitions creates a new BaseReplacePartitions instance. | ||
| func NewReplacePartitions(txn *Transaction) ReplacePartitions { | ||
| return &BaseReplacePartitions{ | ||
| txn: txn, | ||
| addedFiles: make([]iceberg.DataFile, 0), | ||
| snapshotProps: make(iceberg.Properties), | ||
| validateAddedFiles: false, | ||
| stageOnly: false, | ||
| validatePartitionsFilter: false, | ||
| } | ||
| } | ||
|
|
||
| // AddFile adds a data file to the table. | ||
| func (rp *BaseReplacePartitions) AddFile(dataFile iceberg.DataFile) ReplacePartitions { | ||
| rp.addedFiles = append(rp.addedFiles, dataFile) | ||
| return rp | ||
|
Check failure on line 91 in table/replace_partitions.go
|
||
| } | ||
|
|
||
| // AddFiles adds multiple data files to the table. | ||
| func (rp *BaseReplacePartitions) AddFiles(dataFiles ...iceberg.DataFile) ReplacePartitions { | ||
| rp.addedFiles = append(rp.addedFiles, dataFiles...) | ||
| return rp | ||
|
Check failure on line 97 in table/replace_partitions.go
|
||
| } | ||
|
|
||
| // ValidateAddedFilesMatchOverwriteFilter ensures that files added concurrently | ||
| // do not conflict with the replace operation. | ||
| func (rp *BaseReplacePartitions) ValidateAddedFilesMatchOverwriteFilter() ReplacePartitions { | ||
| rp.validateAddedFiles = true | ||
| return rp | ||
|
Check failure on line 104 in table/replace_partitions.go
|
||
| } | ||
|
|
||
| // ValidateFromSnapshot ensures that no delete files matching this operation | ||
| // are added between the snapshot used to plan the operation and the current snapshot. | ||
| func (rp *BaseReplacePartitions) ValidateFromSnapshot(snapshotID int64) ReplacePartitions { | ||
| rp.fromSnapshotID = &snapshotID | ||
| return rp | ||
| } | ||
|
|
||
| // Set sets a summary property in the snapshot produced by this update. | ||
| func (rp *BaseReplacePartitions) Set(property, value string) ReplacePartitions { | ||
| rp.snapshotProps[property] = value | ||
| return rp | ||
| } | ||
|
|
||
| // StageOnly stages the snapshot but does not update the current snapshot ID. | ||
| func (rp *BaseReplacePartitions) StageOnly() ReplacePartitions { | ||
| rp.stageOnly = true | ||
| return rp | ||
| } | ||
|
|
||
| // Commit executes this update and commits the result. | ||
| func (rp *BaseReplacePartitions) Commit(ctx context.Context) (*Table, error) { | ||
| if len(rp.addedFiles) == 0 { | ||
| return nil, errors.New("cannot commit replace partitions operation with no files added") | ||
| } | ||
|
|
||
| fs, err := rp.txn.tbl.fsF(ctx) | ||
| if err != nil { | ||
| return nil, err | ||
| } | ||
|
|
||
| // Create a replace partitions snapshot producer | ||
| producer := rp.createSnapshotProducer(fs.(io.WriteFileIO)) | ||
|
|
||
| // Add all the new files | ||
| for _, file := range rp.addedFiles { | ||
| producer.appendDataFile(file) | ||
| } | ||
|
|
||
| // Find and mark files to be deleted based on partition matching | ||
| if err := rp.markFilesForDeletion(ctx, producer); err != nil { | ||
| return nil, err | ||
| } | ||
|
|
||
| // Perform validation if requested | ||
| if err := rp.performValidation(ctx); err != nil { | ||
| return nil, err | ||
| } | ||
|
|
||
| // Commit the changes | ||
| updates, reqs, err := producer.commit() | ||
| if err != nil { | ||
| return nil, err | ||
| } | ||
|
|
||
| if err := rp.txn.apply(updates, reqs); err != nil { | ||
| return nil, err | ||
| } | ||
|
|
||
| if rp.stageOnly { | ||
| // If stageOnly is true, return the staged table without committing the transaction | ||
| stagedTable, err := rp.txn.StagedTable() | ||
| if err != nil { | ||
| return nil, err | ||
| } | ||
| return stagedTable.Table, nil | ||
| } | ||
|
|
||
| // Commit the transaction | ||
| return rp.txn.Commit(ctx) | ||
| } | ||
|
|
||
| func (rp *BaseReplacePartitions) createSnapshotProducer(fs io.WriteFileIO) *snapshotProducer { | ||
| commitUUID := uuid.New() | ||
|
|
||
|
Check failure on line 180 in table/replace_partitions.go
|
||
| // For replace partitions, we use the REPLACE operation if there are existing files, | ||
| // otherwise APPEND for the first data | ||
| operation := OpReplace | ||
| if rp.txn.meta.currentSnapshot() == nil { | ||
| operation = OpAppend | ||
| } | ||
|
|
||
| return createSnapshotProducer(operation, rp.txn, fs, &commitUUID, rp.snapshotProps) | ||
| } | ||
|
|
||
| func (rp *BaseReplacePartitions) markFilesForDeletion(ctx context.Context, producer *snapshotProducer) error { | ||
| currentSnapshot := rp.txn.meta.currentSnapshot() | ||
| if currentSnapshot == nil { | ||
| // No existing data to replace | ||
| return nil | ||
| } | ||
|
|
||
| fs, err := rp.txn.tbl.fsF(ctx) | ||
| if err != nil { | ||
| return err | ||
| } | ||
|
|
||
| // Get the partition types for all specs | ||
| allSpecs := rp.txn.meta.specs | ||
| specs := make(map[int]iceberg.PartitionSpec) | ||
| for _, spec := range allSpecs { | ||
| specs[spec.ID()] = spec | ||
| } | ||
| schema := rp.txn.meta.CurrentSchema() | ||
|
|
||
| // Create partition evaluators for matching | ||
| addedPartitions := make(map[int]map[string]bool) // specID -> partition key -> exists | ||
|
|
||
| // Extract partition keys from added files | ||
| for _, addedFile := range rp.addedFiles { | ||
| specID := int(addedFile.SpecID()) | ||
| spec, exists := specs[specID] | ||
| if !exists { | ||
| return fmt.Errorf("partition spec with ID %d not found", specID) | ||
| } | ||
|
|
||
| partitionType := spec.PartitionType(schema) | ||
| partitionKey := partitionKeyString(addedFile.Partition(), partitionType) | ||
|
|
||
| if addedPartitions[specID] == nil { | ||
| addedPartitions[specID] = make(map[string]bool) | ||
| } | ||
| addedPartitions[specID][partitionKey] = true | ||
| } | ||
|
|
||
| // Iterate through existing data files and mark for deletion if partition matches | ||
| for dataFile, err := range currentSnapshot.dataFiles(fs, nil) { | ||
| if err != nil { | ||
| return err | ||
| } | ||
|
|
||
| specID := int(dataFile.SpecID()) | ||
| spec, exists := specs[specID] | ||
| if !exists { | ||
| continue // Skip files with unknown specs | ||
| } | ||
|
|
||
| partitionType := spec.PartitionType(schema) | ||
| partitionKey := partitionKeyString(dataFile.Partition(), partitionType) | ||
|
|
||
| // Check if this partition should be replaced | ||
| if partitions, exists := addedPartitions[specID]; exists && partitions[partitionKey] { | ||
| producer.deleteDataFile(dataFile) | ||
| } | ||
| } | ||
|
|
||
| return nil | ||
| } | ||
|
|
||
| func (rp *BaseReplacePartitions) performValidation(ctx context.Context) error { | ||
| if !rp.validateAddedFiles && rp.fromSnapshotID == nil { | ||
| return nil // No validation requested | ||
| } | ||
|
|
||
| currentSnapshot := rp.txn.meta.currentSnapshot() | ||
| if currentSnapshot == nil { | ||
| return nil // No validation needed for empty tables | ||
| } | ||
|
|
||
| if rp.fromSnapshotID != nil { | ||
| // Validate that no conflicting changes happened since the specified snapshot | ||
| baseSnapshot, err := rp.txn.meta.SnapshotByID(*rp.fromSnapshotID) | ||
| if err != nil || baseSnapshot == nil { | ||
| return fmt.Errorf("snapshot %d not found for validation", *rp.fromSnapshotID) | ||
| } | ||
|
|
||
| if currentSnapshot.SnapshotID != *rp.fromSnapshotID { | ||
| // Check if any conflicting changes were made | ||
| if err := rp.validateNoConflictingChanges(ctx, baseSnapshot, currentSnapshot); err != nil { | ||
| return err | ||
| } | ||
| } | ||
| } | ||
|
|
||
| if rp.validateAddedFiles { | ||
| // Validate that added files don't conflict with the overwrite filter | ||
| // This is a more complex validation that would check for concurrent modifications | ||
| // For now, we'll implement a basic check | ||
| return rp.validateAddedFilesMatchFilter(ctx) | ||
| } | ||
|
|
||
| return nil | ||
| } | ||
|
|
||
| func (rp *BaseReplacePartitions) validateNoConflictingChanges(ctx context.Context, baseSnapshot, currentSnapshot *Snapshot) error { | ||
| // This is a simplified validation - in a full implementation, you would | ||
| // check for conflicting delete files, schema changes, etc. | ||
| // For now, we'll just check if the partition spec or schema changed | ||
|
|
||
| if baseSnapshot.SchemaID != nil && currentSnapshot.SchemaID != nil && | ||
| *baseSnapshot.SchemaID != *currentSnapshot.SchemaID { | ||
| return errors.New("schema changed since snapshot used for planning replace partitions") | ||
| } | ||
|
|
||
| return nil | ||
| } | ||
|
|
||
| func (rp *BaseReplacePartitions) validateAddedFilesMatchFilter(ctx context.Context) error { | ||
| // This would perform more sophisticated validation to ensure that | ||
| // concurrently added files don't conflict with the replace operation | ||
| // For now, we'll implement a basic check | ||
| return nil | ||
| } | ||
|
|
||
| // partitionKeyString creates a string representation of a partition for comparison | ||
| func partitionKeyString(partition map[int]any, partitionType *iceberg.StructType) string { | ||
| if len(partition) == 0 { | ||
| return "__HIVE_DEFAULT_PARTITION__" | ||
| } | ||
|
|
||
| result := "" | ||
| for i, field := range partitionType.FieldList { | ||
| if i > 0 { | ||
| result += "/" | ||
| } | ||
| value := partition[field.ID] | ||
| if value == nil { | ||
| result += fmt.Sprintf("%s=__HIVE_DEFAULT_PARTITION__", field.Name) | ||
|
Check failure on line 323 in table/replace_partitions.go
|
||
| } else { | ||
| result += fmt.Sprintf("%s=%v", field.Name, value) | ||
| } | ||
| } | ||
|
|
||
| return result | ||
| } | ||
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
this and all other PR-s looks like a 1-1 adaptation for java iceberg interfaces
this approach is not really work with golang, it's not ideomatic.
Better to stick with what we have with for example snapshot
No need to define interface just for the sake of interface.