huaxingao commented on code in PR #16961: URL: https://github.com/apache/iceberg/pull/16961#discussion_r3522276414
########## format/index.md: ########## @@ -0,0 +1,310 @@ +--- +title: "Index Spec" +--- +<!-- + - 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. + --> + +# Iceberg Index Specification + +## Background and Motivation + +Indexes enable query engines to locate relevant rows without scanning entire datasets. +They can accelerate point lookups, range predicates, and other retrieval patterns +while preserving Iceberg's table format, snapshot isolation, and interoperability. + +Indexes are optional. Engines may choose to create, maintain, consume, or ignore them. + +## Goals + +- Define a portable metadata format for indexes +- Provide a common storage architecture for index data +- Allow indexes to evolve independently from table metadata as catalog-managed objects +- Enable index sharing across engines +- Provide a framework for defining new index types and transform functions + +## Overview + +Indexes are stored as a collection of files with some Iceberg table like semantics. At a high level they consist of a tracking file (similar to a root manifest file) which contains listings for a defined set of leaf files (similar to data files.) Leaf files store an ordered set of rows containing at least a key and the path of a Iceberg Table data file and the position within that file where the row where that key is stored. The organization of leaf files is defined by an Indexing Transform which varies based on the type of index. This structure is recorded in an Index metadata.json file which contains a set of snapshots, each of which points to a single tracking file mapping to the complete state of an Iceberg table at a given Iceberg table snapshot. + +Like Iceberg tables, views, and functions: + +- Metadata files (index metadata and tracking files) and data files (leaf files) are immutable +- Updates create new metadata files +- Catalogs perform atomic metadata swaps + +Each index snapshot references a tracking file which describes the leaf files belonging to the snapshot. + +```text +Index Metadata + | + +-- Index Snapshot + | + +-- Tracking File + | + +-- Leaf Data Files +``` + +Transform functions derive a transform value from the key columns and determine how index entries are organized within +the leaf files. +- The transform value space is divided into non-overlapping ranges. +- Each leaf file stores entries for a single range. +- The tracking file stores range bounds for each leaf file. + +This structure enables efficient planning while keeping the data layout flexible for different index implementations. + +## Definitions + +### Index Type + +The index type defines the logical category of an index and the class of queries it is designed to accelerate. + +The following index type is defined in this specification: + +| Type | +|--------| +| SCALAR | + +The following index types are reserved for future specifications: + +| Type | +|--------| +| VECTOR | +| TERM | + +The index type communicates the capabilities of an index to query engines and helps determine whether an index is +applicable to a particular query. + +### Index Transform Function + +The index transform function defines how the index organization key is derived from source table columns when rows are +stored in the index. + +The transform function determines the physical organization of the indexed data and therefore influences which query +patterns can efficiently leverage the index. + +The following transform functions are defined in this specification: + +| Transform | +|-----------| +| IDENTITY | +| HASH | +| HILBERT | + +The following transform function is reserved for future specifications: + +| Transform | +|-----------| +| IVF | + +Index Instances may share the same index type while using different transform functions. + +### Index Instance + +An index instance is a concrete realization of an index type and function applied to a specific table. + +Users create index instances by specifying: + +- The source table +- The index type +- The transform function +- The indexed columns +- The included columns +- Index properties + +Multiple instances of the same index type may exist for a table. + +### Index Snapshot + +An index snapshot is an immutable version of the index data generated from a specific table snapshot. + +Each index snapshot references a complete set of index files and contains all data from the referenced table snapshot. + +## Index Metadata + +The index metadata file stores the index definition and snapshot history. + +### Index Metadata File + +| Requirement | Field | Type | Description | +|-------------|---------------------|--------------------------|-------------------------------------------------| +| required | format-version | int | Index specification version | +| required | uuid | string | Stable UUID assigned at creation | +| required | table-uuid | string | UUID of the indexed table | +| required | location | string | Index root location | +| required | type | string | Logical index type | +| required | transform-function | string | Physical organization transform | +| required | key-column-ids | list<int> | Indexed columns | +| optional | included-column-ids | list<int> | Included columns | +| optional | properties | map<string,string> | Index properties applicable for every snapshot | +| optional | current-snapshot-id | long | ID of the current index snapshot | +| required | snapshots | list<index-snapshot> | Index snapshots | + +## Index Snapshot + +Each index snapshot corresponds to one version of the index data. + +| Requirement | Field | Type | Description | +|-------------|--------------------------|--------------------|---------------------------------------------------------------------| +| required | snapshot-id | long | Index snapshot identifier | +| required | source-table-snapshot-id | long | Source table snapshot | +| required | timestamp-ms | long | Snapshot creation timestamp | +| required | tracking-file | string | Tracking file location | +| optional | properties | map<string,string> | Snapshot properties specific to this snapshot | +| optional | key-metadata | binary | Implementation-specific key metadata, for tracking file encryption. | + +## Tracking File + +Each index snapshot references exactly one tracking file. + +It contains summary metadata about all leaf files belonging to the index snapshot and enables efficient planning +without scanning every leaf file. + +The tracking file may be stored using any supported metadata file format. + +### Tracking File Entry + +Each tracking file contains a collection of tracking file entries. A tracking file entry describes a single leaf file Review Comment: When re-reading the spec, I feel that "Leaf File Entry" is confusing here, since it lives under the Tracking File section. I'd like to keep this as Tracking File Entry to match Iceberg's manifest entry convention — it lives in the tracking file and describes a leaf file, just as a manifest entry lives in the manifest and describes a data file. I've kept the body text explicit ("describes a single leaf file") to avoid the ambiguity you noted. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
