A query / expression language for performing computations on JSON-like structures. Tuned for clientside ML feature extraction.

evinism evinism Last update: Apr 01, 2024

Screen Shot 2022-05-28 at 11 46 17 AM

GitHub license Python Node.js npm version npm version

MistQL is a query language for JSON-like structures, built for embedding across multiple domains. It supports logic for querying and manipulating JSON-like data in a simple, readable manner. MistQL serves as a powerful common expression language across multiple implementations.

For more detailed usage information, please visit MistQL's docs site.

Join the Discord!!

Links

At A Glance

MistQL is an embedded query language.

import mistql from 'mistql';

const animals = [
  {name: 'charlie', species: 'dog'},
  {name: 'mulberry', species: 'cat'},
]

const cats = mistql.query('@ | filter species == "dog" | count')

The primary power of MistQL comes from its strong cross-platform behavior semantics. For example, the following have the same behavior:

# Python
import mistql
import json

query = 'events | groupby type | keys'
print(mistql.query(query, json.loads(data)))
// JavaScript
import mistql from 'mistql'

const query = 'events | groupby type | keys';
console.log(mistql.query(query, JSON.parse(data)))

Developing MistQL

Contributions to MistQL are very welcome!

As MistQL is a small project, there are no formatting requirements for either issues or pull requests.

If you're planning on making a new implementation, ping the discord and we'll coordinate!

For an example PR that adds a function to MistQL, refer to this PR as an example of a new function with moderate complexity.

Code workflow

Code contributions to MistQL should roughly follow standard open source workflows:

  1. Fork the project
  2. Make code changes on your fork of the project.
  3. (if necessary) Pull upstream to bring in new changes
  4. Submit a pull request to MistQL's main branch.
  5. (if necessary) Implement changes requested by maintainers.
  6. Wait for the branch to be accepted and merged by maintainers!

MistQL standard

No MistQL standard yet exists, but we're aiming for the 0.5.0 release of mistql as a standardizable language. After the 0.5.0 release, we will create a language specification, separate from any implementation.

In the meantime, we're actually pretty close.

We have a Lark grammar which defines the language's syntax. This is likely the final grammar that will be formalized into ABNF, although it is possible that we may need to fix minor issues before 0.5.0. I expect this to barely change, if at all.

Additionally, our language-independent test suite is rather extensive and forms the de-facto standard of behaviors, as shared by both Python and JavaScript. While not strictly formalized, the tests and the docs together form a cohesive body of behaviors, that, except for a few minuitae, is of sufficient detail to be standardized.

Directory Structure

MistQL's directory structure is a monorepo, currently consisting of these main directories:

  1. /docs: Documentation Site (hosted at mistql.com)
  2. /js: MistQL's browser implementation (e.g. mistql on npm).
  3. /py: MistQL's python implementation (e.g. mistql on pypi).
  4. /shared: Shared assets between all implementation. Contains the language-independent test suite.

Developing for the docs site

Docs are built via a fairly standard Docusaurus 2 implementation. Please follow Docusaurus's docs for developing for the Docs site.

Developing for mistql on npm

mistql is written exclusively using typescript. Additionally, mistql uses yarn for dependency management, versioning, and uploading. JS-specific tests are stored alongside their implementation, using the suffix .spec.ts. Tests that describe the language itself are written in a language agnostic JSON format in the /shared directory. Writing tests for all feature additions and bug fixes is strongly encouraged.

For all major improvements, it is strongly encouraged to run yarn bundlesize to estimate gzipped impact of MistQL on a browser. MistQL for the browser should, in general, remain relatively close to 5kb.

The directory structure is relatively flat, except for the single src/builtins folder, which contains the implementation of all of MistQL's internal functions.

Developing for mistql on pypi

mistql is a fairly standard python package managed with poetry.

Tests can be run using pytest, e.g. poetry run pytest from within the /py directory.

Subscribe to our newsletter