apache-arrow

What is apache-arrow?

The apache-arrow npm package provides a cross-language development platform for in-memory data. It is designed to improve the performance and efficiency of data processing and analytics by using a columnar memory format. This package is particularly useful for handling large datasets and performing complex data manipulations.

What are apache-arrow's main functionalities?

Reading and Writing Arrow Files

This feature allows you to read and write Arrow files, which are efficient for storing and transferring large datasets. The code sample demonstrates how to read an Arrow file into a table and how to write a new table to an Arrow file.

const arrow = require('apache-arrow');
const fs = require('fs');

// Reading an Arrow file
const arrowFile = fs.readFileSync('data.arrow');
const table = arrow.Table.from([arrowFile]);
console.log(table.toString());

// Writing an Arrow file
const newTable = arrow.Table.new([{ name: 'Alice', age: 30 }, { name: 'Bob', age: 25 }]);
const arrowBuffer = newTable.serialize();
fs.writeFileSync('newData.arrow', arrowBuffer);

DataFrame Operations

This feature provides DataFrame-like operations, such as creating tables, selecting columns, and filtering rows. The code sample shows how to create a DataFrame, select a column, and filter rows based on a condition.

const arrow = require('apache-arrow');

// Creating a DataFrame
const df = new arrow.Table({
  name: arrow.Utf8Vector.from(['Alice', 'Bob']),
  age: arrow.Int32Vector.from([30, 25])
});

// Selecting a column
const names = df.getColumn('name');
console.log(names.toArray());

// Filtering rows
const filtered = df.filter(row => row.get('age') > 25);
console.log(filtered.toString());

Interoperability with Other Languages

Apache Arrow supports interoperability with other languages like Python, R, and Java. The code sample demonstrates how to create a table in JavaScript and pass it to Python using the pyarrow library.

const arrow = require('apache-arrow');
const pyarrow = require('pyarrow'); // Assuming you have a Python environment set up

// Create a table in JavaScript
const table = arrow.Table.new([{ name: 'Alice', age: 30 }, { name: 'Bob', age: 25 }]);
const arrowBuffer = table.serialize();

// Pass the buffer to Python
const pyTable = pyarrow.Table.from_buffer(arrowBuffer);
print(pyTable)

Other packages similar to apache-arrow

pandas

Pandas is a powerful data manipulation and analysis library for Python. It provides DataFrame and Series data structures for handling tabular and time series data. While pandas is highly efficient for in-memory data manipulation, it does not use a columnar memory format like Apache Arrow, which can be less efficient for certain types of operations.

dask

Dask is a parallel computing library in Python that scales the existing Python ecosystem, including pandas and NumPy. It provides advanced parallelism for analytics, enabling the processing of large datasets that do not fit into memory. Unlike Apache Arrow, Dask focuses on parallel computing and distributed data processing.

README

Apache Arrow in JS

Arrow is a set of technologies that enable big data systems to process and transfer data quickly.

Install apache-arrow from NPM

npm install apache-arrow or yarn add apache-arrow

(read about how we package apache-arrow below)

Powering Columnar In-Memory Analytics

Apache Arrow is a columnar memory layout specification for encoding vectors and table-like containers of flat and nested data. The Arrow spec aligns columnar data in memory to minimize cache misses and take advantage of the latest SIMD (Single input multiple data) and GPU operations on modern processors.

Apache Arrow is the emerging standard for large in-memory columnar data (Spark, Pandas, Drill, Graphistry, ...). By standardizing on a common binary interchange format, big data systems can reduce the costs and friction associated with cross-system communication.

Get Started

Check out our API documentation to learn more about how to use Apache Arrow's JS implementation. You can also learn by example by checking out some of the following resources:

• /js/test/unit - Unit tests for Table and Vector

Cookbook

Get a table from an Arrow file on disk (in IPC format)

import { readFileSync } from 'fs';
import { tableFromIPC } from 'apache-arrow';

const arrow = readFileSync('simple.arrow');
const table = tableFromIPC(arrow);

console.table(table.toArray());

/*
 foo,  bar,  baz
   1,    1,   aa
null, null, null
   3, null, null
   4,    4,  bbb
   5,    5, cccc
*/

Create a Table when the Arrow file is split across buffers

import { readFileSync } from 'fs';
import { tableFromIPC } from 'apache-arrow';

const table = tableFromIPC([
    'latlong/schema.arrow',
    'latlong/records.arrow'
].map((file) => readFileSync(file)));

console.table([...table]);

/*
        origin_lat,         origin_lon
35.393089294433594,  -97.6007308959961
35.393089294433594,  -97.6007308959961
35.393089294433594,  -97.6007308959961
29.533695220947266, -98.46977996826172
29.533695220947266, -98.46977996826172
*/

Create a Table from JavaScript arrays

import { tableFromArrays } from 'apache-arrow';

const LENGTH = 2000;

const rainAmounts = Float32Array.from(
    { length: LENGTH },
    () => Number((Math.random() * 20).toFixed(1)));

const rainDates = Array.from(
    { length: LENGTH },
    (_, i) => new Date(Date.now() - 1000 * 60 * 60 * 24 * i));

const rainfall = tableFromArrays({
    precipitation: rainAmounts,
    date: rainDates
});

console.table([...rainfall]);

Load data with fetch

import { tableFromIPC } from "apache-arrow";

const table = await tableFromIPC(fetch("/simple.arrow"));

console.table([...table]);

Vectors look like JS Arrays

You can create vector from JavaScript typed arrays with makeVector and from JavaScript arrays with vectorFromArray. makeVector is a lot faster and does not require a copy.

import { makeVector } from "apache-arrow";

const LENGTH = 2000;

const rainAmounts = Float32Array.from(
    { length: LENGTH },
    () => Number((Math.random() * 20).toFixed(1)));

const vector = makeVector(rainAmounts);

const typed = vector.toArray()

assert(typed instanceof Float32Array);

for (let i = -1, n = vector.length; ++i < n;) {
    assert(vector.get(i) === typed[i]);
}

String vectors

Strings can be encoded as UTF-8 or dictionary encoded UTF-8. Dictionary encoding encodes repeated values more efficiently. You can create a dictionary encoded string conveniently with vectorFromArray or efficiently with makeVector.

import { makeVector, vectorFromArray, Dictionary, Uint8, Utf8 } from "apache-arrow";

const utf8Vector = vectorFromArray(['foo', 'bar', 'baz'], new Utf8);

const dictionaryVector1 = vectorFromArray(
    ['foo', 'bar', 'baz', 'foo', 'bar']
);

const dictionaryVector2 = makeVector({
    data: [0, 1, 2, 0, 1],  // indexes into the dictionary
    dictionary: utf8Vector,
    type: new Dictionary(new Utf8, new Uint8)
});

Getting involved

See DEVELOP.md

Even if you do not plan to contribute to Apache Arrow itself or Arrow integrations in other projects, we'd be happy to have you involved:

• Join the mailing list: send an email to [e-mail]. Share your ideas and use cases for the project
• Follow our activity on GitHub issues
• Learn the format
• Contribute code to one of the reference implementations

We prefer to receive contributions in the form of GitHub pull requests. Please send pull requests against the github.com/apache/arrow repository.

If you are looking for some ideas on what to contribute, check out the GitHub issues for the Apache Arrow project. Comment on the issue and/or contact [e-mail] with your questions and ideas.

If you’d like to report a bug but don’t have time to fix it, you can still post it on GitHub issues, or email the mailing list [e-mail]

Packaging

apache-arrow is written in TypeScript, but the project is compiled to multiple JS versions and common module formats.

The base apache-arrow package includes all the compilation targets for convenience, but if you're conscientious about your node_modules footprint, we got you.

The targets are also published under the @apache-arrow namespace:

npm install apache-arrow # <-- combined es2015/CommonJS/ESModules/UMD + esnext/UMD
npm install @apache-arrow/ts # standalone TypeScript package
npm install @apache-arrow/es5-cjs # standalone es5/CommonJS package
npm install @apache-arrow/es5-esm # standalone es5/ESModules package
npm install @apache-arrow/es5-umd # standalone es5/UMD package
npm install @apache-arrow/es2015-cjs # standalone es2015/CommonJS package
npm install @apache-arrow/es2015-esm # standalone es2015/ESModules package
npm install @apache-arrow/es2015-umd # standalone es2015/UMD package
npm install @apache-arrow/esnext-cjs # standalone esNext/CommonJS package
npm install @apache-arrow/esnext-esm # standalone esNext/ESModules package
npm install @apache-arrow/esnext-umd # standalone esNext/UMD package

Why we package like this

The JS community is a diverse group with a varied list of target environments and tool chains. Publishing multiple packages accommodates projects of all stripes.

If you think we missed a compilation target and it's a blocker for adoption, please open an issue.

Supported Browsers and Platforms

The bundles we compile support moderns browser released in the last 5 years. This includes supported versions of Firefox, Chrome, Edge, and Safari. We do not actively support Internet Explorer. Apache Arrow also works on maintained versions of Node.

People

Full list of broader Apache Arrow committers.

• Brian Hulette, committer
• Paul Taylor, committer
• Dominik Moritz, committer

Powered By Apache Arrow in JS

Full list of broader Apache Arrow projects & organizations.

Open Source Projects

• Apache Arrow -- Parent project for Powering Columnar In-Memory Analytics, including affiliated open source projects
• Perspective -- Perspective is an interactive analytics and data visualization component well-suited for large and/or streaming datasets. Perspective leverages Arrow C++ compiled to WebAssembly.
• Falcon is a visualization tool for linked interactions across multiple aggregate visualizations of millions or billions of records.
• Vega is an ecosystem of tools for interactive visualizations on the web. The Vega team implemented an Arrow loader.
• Arquero is a library for query processing and transformation of array-backed data tables.
• OmniSci is a GPU database. Its JavaScript connector returns Arrow dataframes.

License

Apache 2.0