deltalake-tools 0.1.21

A set of easy to use convenient tools for deltalake tables.

Deltalake Tools

Introduction

A set of easy to use commands for deltalake, with a command line interface.

You probably don't need this, especially if you're already using delta-rs (deltalake). Just like you don't need awscli when you already have a boto3 client. It's useful as a delta cli ie. as a cron job.

Also working on a Rust implementation, with python bindings using PYO3/maturin.

Delta Table Commands currently supported:

• compact
• vacuum
• create-checkpoint
• table-version
• delete-table
• create-test-table
• convert-to-table
• ...

Storage services currently supported:

• local storage
• AWS S3 (virtual and path addressing style)
• IBM COS (virtual and path addressing style)
• Azure blob storage
• Google cloud storage
• ...

Platforms supported

	arm64	amd64
linux	x	x
mac	x	x
windows

Minimal Python version

3.10

Getting started

Install

pip install deltalake-tools

check out astral's rye, uv and ruff projects

(uv is a blazingly fast drop-in replacement for pip.)

$ uv install deltalake-tools

If you prefer rye:

$ rye add deltalake-tools

Usage

help

$ deltalake-tools -h
Usage: deltalake-tools [OPTIONS] COMMAND [ARGS]...

Options:
  --version   Show the version and exit.
  -h, --help  Show this message and exit.

Commands:
  compact
  create-checkpoint
  table-version
  vacuum
$

Consider the following test table:

/tmp/test_delta_table
├── 0-ed28f60b-7569-47fc-90fa-9cbaad8ccd27-0.parquet
├── 1-dde22814-9070-4df6-be8c-5ec564c8cfd3-0.parquet
├── 10-eafcb45c-6fdd-467b-b35d-d8bf127ae243-0.parquet
├── 2-7bcb5e5d-d2e6-4975-a2d6-a399413b2883-0.parquet
├── 3-7412d097-1b4c-4d45-9ac1-7bc312f20f62-0.parquet
├── 4-9a5bd960-08fb-4b8f-8e94-8f81574799e5-0.parquet
├── 5-a2182ca3-c334-43be-8935-210fc839ff77-0.parquet
├── 6-f9f4597c-9709-4029-a0ee-a11d757072bf-0.parquet
├── 7-a3c8fbef-27eb-4ee1-ae2a-2a30c9964fbc-0.parquet
├── 8-8221beb7-9a14-4ce5-a205-c5b862cbce0d-0.parquet
├── 9-63ec9854-d58c-4d48-b3cf-9c18c43e2d17-0.parquet
└── _delta_log
    ├── 00000000000000000000.json
    ├── 00000000000000000001.json
    ├── 00000000000000000002.json
    ├── 00000000000000000003.json
    ├── 00000000000000000004.json
    ├── 00000000000000000005.json
    ├── 00000000000000000006.json
    ├── 00000000000000000007.json
    ├── 00000000000000000008.json
    ├── 00000000000000000009.json
    └── 00000000000000000010.json

table-version

$ deltalake-tools table-version /tmp/test_delta_table       
10

compact

• increments version
• rewrites the date into 1 file. Here, 11 files are replaced by 1. This has a considerable beneficial effect on read performance.

$ deltalake-tools table-version /tmp/test_delta_table
10
$ deltalake-tools compact /tmp/test_delta_table
{'numFilesAdded': 1, 'numFilesRemoved': 11, 'filesAdded': '{"avg":1034.0,"max":1034,"min":1034,"totalFiles":1,"totalSize":1034}', 'filesRemoved': '{"avg":909.3636363636364,"max":913,"min":873,"totalFiles":11,"totalSize":10003}', 'partitionsOptimized': 1, 'numBatches': 11, 'totalConsideredFiles': 11, 'totalFilesSkipped': 0, 'preserveInsertionOrder': True}
$ deltalake-tools table-version /tmp/test_delta_table
11

vacuum

• increments version
• by default, the minimal retention hours is 168. This can be overridden, but read the docs first.

arguments:

• --retention-hours: how long do you want to keep, for time travelling. Default: 168 (1 week)
• --disable-retention-duration: disable the safety check
• --force: by default, this is a dry-run operation. Use this to actually perform the vacuum command.

These safety features are implemented intentionally. Read the docs for more information.

$ deltalake-tools table-version /tmp/test_delta_table
11
$deltalake-tools vacuum /tmp/test_delta_table --retention-hours 0 --disable-retention-duration --force 
['3-7412d097-1b4c-4d45-9ac1-7bc312f20f62-0.parquet', '6-f9f4597c-9709-4029-a0ee-a11d757072bf-0.parquet', '5-a2182ca3-c334-43be-8935-210fc839ff77-0.parquet', '2-7bcb5e5d-d2e6-4975-a2d6-a399413b2883-0.parquet', '1-dde22814-9070-4df6-be8c-5ec564c8cfd3-0.parquet', '8-8221beb7-9a14-4ce5-a205-c5b862cbce0d-0.parquet', '9-63ec9854-d58c-4d48-b3cf-9c18c43e2d17-0.parquet', '7-a3c8fbef-27eb-4ee1-ae2a-2a30c9964fbc-0.parquet', '0-ed28f60b-7569-47fc-90fa-9cbaad8ccd27-0.parquet', '10-eafcb45c-6fdd-467b-b35d-d8bf127ae243-0.parquet', '4-9a5bd960-08fb-4b8f-8e94-8f81574799e5-0.parquet']
$ deltalake-tools table-version /tmp/test_delta_table                                                  
12

create-checkpoint

• does not increment version
• use this if your deltalake clients require this _last_checkpoint file, as it is not written by default. Not by spark, not by delta-rs.

$ deltalake-tools table-version /tmp/test_delta_table
13
$ deltalake-tools create-checkpoint /tmp/test_delta_table
Checkpoint created successfully.
$ deltalake-tools table-version /tmp/test_delta_table    
13

Contribute

Test

• pytest
• moto[server]: needs to be started before initializing any clients (boto3, delta)

Running tests:

$ rye run pytest tests

management

• rye
• make

changelog

• git-cliff