Skip to content

Latest commit

 

History

History
402 lines (286 loc) · 19.3 KB

CONTRIBUTING.md

File metadata and controls

402 lines (286 loc) · 19.3 KB

Contributing

Introduction

Welcome! Thank you for contributing to SchemaStore!

There are various ways you can contribute:

  • Add a new JSON Schema
    • Local schema
    • Remote schema
  • Enhance existing JSON schemas:
    • Fix typos
    • Fix bugs
    • Improve constraints
    • Add positive/negative tests
    • Refactor to pass under strict mode

Most people want to add a new schema. For steps on how to do this, read the How to add a JSON Schema that's hosted in this repository section below.

If you want to contribute, but not sure what needs fixing, see the help wanted and good first issue labels on GitHub.

Overview

Schema files are located in src/schemas/json. Each schema file has a corresponding entry in the Schema Catalog. Each catalog entry has a fileMatch field. IDEs use this field to know which files the schema should be used for (in autocompletion).

Some schema files have associated positive and negative tests, located at src/test and src/negative_test, respectively. This repository has Grunt tasks that automatically load these files and use a validator (either AJV or SchemaSafe) to ensure that they either pass or fail validation.

There are three types of schema validation modes:

  • AJV strict mode: The default validation mode that is most stringent
  • AJV non-strict mode: Some rules are relaxed for the sake of brevity. To validate under non-strict mode, add your schema to the ajvNotStrictMode field in schema-validation.json
  • SchemaSafe: Helps catch errors within schemas that would otherwise be missed. This is a WIP.

Recommended Extensions

We highly recommend installing the following extensions for your IDE:

  • EditorConfig to automatically configure editor settings
  • Prettier to automatically configure file formatting

If you are modifying JavaScript files, we also recommend:

  • ESLint to automatically show JavaScript issues

Schema Authoring

The goal of JSON Schemas in this repository is to correctly validate schemas that are used by the actual tools. That means, if a property is undocumented or deprecated, it should still be included in the schema.

Best practices

✔️ Use the most recent JSON Schema version (specified by $schema) that's widely supported by editors and IDEs. Currently, the best supported version is draft-07. Later versions of JSON Schema are not recommended for use in SchemaStore until editor/IDE support improves for those versions.

✔️ Use base.json schema for draft-07 and base-04.json for draft-04 with some common types for all schemas.

There is an unofficial draft-07 schema that uses JSON Schema to validate your JSON Schema. It checks that:

  • type, title, description properties are required
  • There are no empty arrays. For instance, it's impossible to write less than 2 sub-schemas for allOf
  • type can't be an array, which is intentional, anyOf/oneOf should be used in this case
  • It links to understanding-json-schema for each hint/check

Don't forget add test files.

  • Always be consistent across your schema: order properties and describe in the one style.
  • Always use title when property type is an object to enhance editor experience which use this property to show errors (like VS Code). Why?
  • Always use description, type, additionalProperties.
    • Always set additionalProperties to false until documentation permits additional properties explicitly. That tool the JSON schema is created for can be changed in the future to allow wrong extra properties.
  • Always use minLength/maxLength/pattern/etc for property values.
  • Don't end title/description values with colon.
  • Always omit leading articles for title-s and trailing punctuation to make expected object values look more like types in programming languages. Also start title-s with a lowercase letter and try use nouns for titles instead of sentences.
  • Always explicitly state whether some setting is global for some tool or local for a project created with this tool. For instance if some settings is local then add for the current <project-type> at the end of the description like Whether to ignore a theme configuration for the current site for Jekyll.
  • Always add documentation url to descriptions when available in the following format: <description>\n<url> like "Whether to ignore a theme configuration for the current site\nhttps://jekyllrb.com/docs/configuration/options/#global-configuration".

Avoiding overconstraint

Sometimes, constraints do more harm than good. For example, cron strings validation regexes. False positives are likely as due to their complexity and abundance of implementations; and, when there is an error, the error message isn't helpful. Such cases can include:

  • cron regexes
  • string-embedded DSLs
  • SSH URLs, HTTPS URLs, and other complex URIs

Undocumented Features

The use of undocumented features in schemas is permitted and encouraged. However they must be labeled as such.

It is preferred to add UNDOCUMENTED. to the beginning of description.

{
  "type": "object",
  "properties": {
    "experimental_useBranchPrediction": {
      "type": "string",
      "description": "UNDOCUMENTED. Enables branch prediction in the build."
    }
  }
}

However, that is not always possible or correct. Alternatively, use $comment:

{
  "type": "object",
  "tsBuildInfoFile": {
    "$comment": "The value of 'null' is UNDOCUMENTED.",
    "description": "Specify the folder for .tsbuildinfo incremental compilation files.",
    "default": ".tsbuildinfo",
    "type": ["string", "null"],
    "description": "Specify the folder for .tsbuildinfo incremental compilation files."
  }
}

In this case, { "tsBuildInfoFile": null } is not documented. Using a string value is, however.

Note that JSON Schema draft 2019-09 adds support for a deprecated field. While this would be the best option, most schemas in this repository are draft-07 - and as a result, Editors and IDEs may not use it.

API Compatibility

Care must be taken to reduce breaking changes; some include:

1. Preserving schema names

When renaming a schema name, the old version must continue to exist. Its content will look something like:

{
  "$ref": "https://json.schemastore.org/NEWNAME.json"
}

2. Preserving schema paths

Many tools, such as validate-pyproject, accept passing in subpaths for validation like so:

validate-pyproject --tool cibuildwheel=https://json.schemastore.org/cibuildwheel.toml#/properties/tool/properties

This means that renames in subschema paths aren't zero-cost. If a rename is necessary, keep the old path and $ref where necessary.

Troubleshooting

  • There may be git merge conflicts in catalog.json because you added the item to the end of the list instead of alphabetically
  • The pre-commit build server failed because the PR was created/push from an organization and not from your own account

How-to

How to add a JSON Schema that's hosted in this repository

Follow these instructions if you want to add the JSON schema file directly to this repository. If you want to keep the JSON schema hosted elsewhere, see How to add a JSON Schema that's self-hosted/remote/external.

When uploading a new schema file, make sure it targets a file that is commonly used or has potential for broad uptake.

First, clone the repository:

git clone https://github.com/SchemaStore/schemastore
cd schemastore

Be sure that NodeJS is installed. The minimum required NodeJS version is defined by the engines key in package.json. Now, install dependencies and run the new_schema Grunt task:

cd src
npm install
npm run grunt new_schema

You will be prompted for the name of the schema. Once you enter your schema name, the task will:

  • Create a new schema file at src/schemas/json/<schemaName>.json
  • Create a positive test file at src/test/<schemaName>/<schemaName>.json
  • Print a string for you to add to the Schema Catalog

If you do not wish to use the new_schema Grunt task, the manual steps are listed below 👇

Manual Steps
  1. Create a schema file in src/schemas/json/<name>.json:

    {
      "$id": "https://json.schemastore.org/<schemaName>.json",
      "$schema": "http://json-schema.org/draft-07/schema#",
      "additionalProperties": true,
      "properties": {},
      "type": "object"
    }
  2. Add positive test files at src/test/<schemaName>/<testFile>: (optional, but strongly recommended)

    File extensions .json, .toml, .yml, and .yaml are supported.

  3. Add negative test files at src/negative_test/<schemaName>/<testFile> (optional)

  4. Register your schema (in alphabetical order) in the schema catalog:

    {
      "description": "Schema description",
      "fileMatch": ["list of well-known filenames matching schema"],
      "name": "Friendly schema name",
      "url": "https://json.schemastore.org/<schemaName>.json"
    }

Finally, validate your changes. See How to Validate a JSON Schema for details.

How to add a JSON Schema that's self-hosted/remote/external

You may wish to serve a schema from https://json.schemastore.org/<schemaName>.json, but keep the content of the schema file at a place you control (not this repository).

See this PR as an example. Simply register your schema in the schema catalog, with the url field pointing to your schema file:

{
  "name": "hydra.yml",
  "description": "ORY Hydra configuration file",
  "fileMatch": [
    "hydra.json",
    "hydra.yml",
    "hydra.yaml",
    "hydra.toml"
  ],
  "url": "https://raw.githubusercontent.com/ory/hydra/master/.schema/version.schema.json"
},

How to edit an existing JSON Schema

First, clone the repository:

git clone https://github.com/SchemaStore/schemastore
cd schemastore

Be sure that NodeJS is installed. The minimum required NodeJS version is defined by the engines key in package.json.

Now, modify the schema you intend to modify. Schemas are located under src/schemas/json.

Finally, validate your changes. See How to Validate a JSON Schema for details.

How to add a JSON Schema with multiple versions

Refer to this agripparc PR as an example. First, your schema names should be suffix with the version number.

  • src/schemas/json/agripparc-1.2.json
  • src/schemas/json/agripparc-1.3.json
  • src/schemas/json/agripparc-1.4.json

Then, use the versions field to list each of them. Add the latest version to the url field:

{
  "description": "JSON schema for the Agrippa config file",
  "fileMatch": [".agripparc.json", "agripparc.json"],
  "name": ".agripparc.json",
  "url": "https://json.schemastore.org/agripparc-1.4.json",
  "versions": {
    "1.2": "https://json.schemastore.org/agripparc-1.2.json",
    "1.3": "https://json.schemastore.org/agripparc-1.3.json",
    "1.4": "https://json.schemastore.org/agripparc-1.4.json"
  }
}

How to move a JSON Schema from SchemaStore to somewhere that's self-hosted

Simply changing the url field in the schema catalog (as described here) is not enough. You must also:

  • Keep the original schema files in the repository and point to your schema with $ref
  • Add an entry under skiptest so the remaining schema file isn't tested

See this PR for a full example.

How to add a $ref to a JSON Schema that's hosted in this repository

$ref from schema_x.json to schema_y.json

  • Both schemas must exist locally in SchemaStore.
  • Both schemas must have the same draft (ex. draft-07)
  • schema_y.json must have id or $id with this value "https://json.schemastore.org/schema_y.json"
  • In schema_x.json, add ref to schema_y.json: "$ref": "https://json.schemastore.org/schema_y.json#..."
  • Within schema-validation.json, in "options": [], add an entry: { "schema_x.json": {"externalSchema": ["schema_y.json"] } }

How to add a $ref to a JSON Schema that's self-hosted

This currently isn't possible. This is tracked by issue #2731.

How to validate a JSON Schema

This repository validations JSON Schemas in multiple ways:

  • "Meta validation"
    • Check there are no unused files or directories
    • Check that schema is valid JSON
    • Check that the entry in catalog.json is valid
    • etc.
  • AJV strict mode validation
    • Checks schema to prevent any unexpected behaviors or silently ignored mistakes
    • Fixing strict mode errors does not change validation results, it only serves to improve schema quality
    • More info at Ajv Strict mode docs
  • Schema validation
    • Actually uses the schema against any test files
    • Checks that schemas properly constrain the object as schema authors intended

To validate all schemas, run:

npm run grunt

Because there are hundreds of schemas, you may only want to validate a single one to save time. To do this, run:

npm run grunt -- --SchemaName=<schemaName.json>

For example, to validate the ava.json schema, run npm run grunt -- --SchemaName=ava.json

Note that <schemaName.json> refers to the filename that the schema has under src/schemas/json. If the task succeeds, your changes are valid and you can safely create a PR.

How to ignore validation errors in a JSON Schema

Note Please only do this if you must. Validating in strict mode catches many common errors by schema authors and improves schema quality.

Sometimes, the build fails due to a failed validation check. See a list of validation checks here. An error may look like:

>> compile              | schemas/json/prefect-deploy.json (draft-07)(FullStrictMode)
>> Error: strict mode: use allowUnionTypes to allow union type keyword at "#/definitions/prefect_docker.deployments.steps.push_docker_image/properties/credentials" (strictTypes)

To ignore most validation errors, you need to modify src/schema-validation.json:

  • If a strict error fails, you need to add your JSON Schema to the ajvNotStrictMode array
  • If you are getting "unknown format" or "unknown keyword" errors, you need to add your JSON Schema to the options array
  • If you are using a recent version of the JSON Schema specification, you will need to add your JSON Schema to the highSchemaVersion array

How to name schemas that are subschemas (partial-)

Often, it is useful to extract a subschema into its own file. This can make it easier to write tests, find schemas pertaining to a particular project, and logically separate extremely large schemas. The partial- prefix makes it easier for SchemaStore developers and subschema consumers to identify that the schema is a subschema.

A subschema should be extracted to its own file based on the following rules:

  • If a schema represents an existing project that could be its own file, then simply use that file for the "subschema". In other places, $ref that file where appropriate.
    • For example, mypy reads configuration from both mypy.ini and pyproject.toml's tool.mypy key. Because mypy.ini is its own file, then name the schema mypy.json like you usually would.
    • Same with Prettier. It reads from .prettierrc.json (among other files) and package.json's prettier key.
  • If the schema cannot be its own file, then extracting the subschema may be an improvement
    • For example, Poetry reads configuration only from pyproject.toml's tool.poetry key. Because the Poetry subschema is relatively complex and a large project, it has been extracted to its own file, partial-poetry.json.
  • If the schema must exist locally to workaround issue #2731, then the subschema should be extracted
    • In a top-level $comment, you must add the date at which you copied the original. See #3526 for an example

Use your best judgement; if the project or schema is small, then the drawbacks of extracting the subschema to its own file likely outweigh the benefits.