Validation-Server

Web service to validate data with support of multiple schema languages

Large parts of practical data science or other data processing work is spent by bringing dirty data into shape. Data formats define desired shapes of data. To check whether data conforms to a data format, it must be validated. This application helps to validate data against data formats. Validation errors show if, where and how a format is violated so data can be cleaned or rejected. Data formats can be configured with schemas in multiple schema languages.

Background
Install
- From GitHub
- Configuration
Usage
API
Maintainers
Contributing
License

Background

Data Formats

Data formats supported by validation server are described as JSON Object with:

id mandatory local format identifier
title optional title
short optional short title (abbreviation or acronym)
description optional short textual description or definition
url optional URL with information about the format
wikidata optional Wikidata identifier of the format
mimetypes optional array of content types
base optional identifier of a base format (e.g. json for JSON-based formats)
versions an optional object with format versions, the keys used as version identifier (is default if unknown) and each values an object with:
- schemas an optional array with schemas, each an object with:
  - type mandatory schema type (identifier of a schema language, e.g. json-schema)
  - url optional URL to retrieve the schema file from
  - value schema as string (mandatory if no url given)

This meta-format is defined as Data About Data Formats with a JSON Schema at https://format.gbv.de/validate/format-schema.json.

API endpoint /formats or command line argument --list return data formats supported by an instance of validation server.

Schema Languages

Schema languages (also known as schema formats or schema types) are data formats used to define other data formats. Formats defined by a schema language all share a common base format. For instance JSON Schema is a schema language to define JSON-based formats, XML Schema is a schema language to define XML-based formats, and regular expressions can be used as schema language to describe character-based formats.

Schema languages supported by validation server are described as data formats with additional mandatory keys restricts referencing the common base format(s).

The following schema languages are supported for validation of other formats. The list is available via API endpoint /languages:

JSON Schema (json-schema)
- Supports draft-04, draft-06, and draft-07
- Supports format keywords from https://github.com/ajv-validator/ajv-formats and https://github.com/luzlab/ajv-formats-draft2019
XML Schema (xsd)
- Requires xmllint executable to be installed.
Regular Expressions (regexp)
- Supports ECMAScript variant with Unicode flag enabled automatically and flags other than i, m, s ignored.
- Can be given as plain pattern or in form /${pattern}/${flags}

Locator Languages

Locator languages such as XPath and JSON Pointer are used to reference parts of a document. Validation server supports the following languages to locate validation errors and/or to select which parts of a document to validate:

JSON Pointer (jsonpointer) for error locations in JSON documents
Line & Column (linecol) for error locations in documents: e.g. 3:7 for line 3 column 7
Character position (rfc5147) for error locations given by character position as defined in RFC 5147: e.g. char=23
JSONPath (jsonpath) limited to value $.* to select to validate elements of a JSON array

Install

Requires at least Node v14.8.0.

Support of XML Schema based formats requires xmllint to be installed:

sudo apt install libxml2-utils

From GitHub

git clone https://github.com/gbv/jskos-server.git
cd jskos-server
npm ci

Configuration

The service must be customized via configuration files. By default, this configuration file resides in config/config.json (or config/config.test.json for tests). Missing keys are defaulted from config/config.default.json:

{
  "title": "Validation Service",
  "description": "Web service to validate data against schemas",
  "version": "X.X.X",
  "port": 3700,
  "proxies": [],
  "postLimit": "20MB",
  "timeout": 2000,
  "allErrors": false
  "verbosity": "info",
  "formats": [],
  "cache": "../cache",
}

Keys version and description are defaulted to its value in package.json. In addition the environment variable NODE_ENV is respected with development as default. Alternative values are production, test, and debug.

Key formats must contain an array of data formats or a file containing such array (JSON, NDJSON or YAML format). The list of formats is automatically extended by some hardcoded formats and schema languages.

Some validators stop after finding the first error. Key allErrors configures validators to continue validation to find more errors.

Additional keys provider and links can be used to adjust the HTML footer.

A JSON Schema to validate configuration is included at config/config-schema.json.

Usage

This node package is mainily designed to be run as web service but it can also be used as module. See API Reference for usage of a running server instance instead.

Run Server

# Development server with hot reload and auto reconnect at localhost:3700 (default)
npm run start

# To run the server in production, run this:
NODE_ENV=production node ./server.js

On startup all configured schemas are downloaded and cached in directory configured with cache (set to false to use a temporary directory). Addition and updates require to restart the server.

Deployment

To provide the service behind a nginx web server at path /validate/ (like at http://format.gbv.de/validate/), add this to nginx configuration file:

location /validate {
    port_in_redirect off;
    return /validate/;
}
location /validate/ {
    proxy_pass http://127.0.0.1:3700/;
}

We recommend to use PM2 to start and update the server:

pm2 start ecosystem.config.json

Updates

To update an instance deployed with PM2:

# get updates from repository
git pull

# install dependencies
npm ci

# restart the process (adjust process name if needed)
pm2 restart validation-server

The server needs to be restarted to reflect updates in configuration, including formats and schemas to be supported. Script bin/update.js can help updating formats and local schema files without breaking things.

Use as Module

The internal API is not stable yet!

const { loadConfig, validateConfig, createService, knownFormats } = require("validation-server")

const config = loadConfig()

createService(config).then(service => {
  const format = service.getFormat("json-schema", { version: "draft-07" })

  // asynchronous validation
  format.valid(data)
    .then(() => console.log("ok"))
    .catch(e => console.error(e.errors))

  // validate a stream of records
  const resultStream = inputStream.pipe(format.validateStream)
})

loadConfig

Returns a valid configuration object. Location of an optional configuration file can be given as argument (or by setting environment variable NODE_ENV to debug or test):

const config = loadConfig("./config.json")

An error is thrown on invalid configuration.

validateConfig

Validate a configuration object. Throws an error if configuration is not valid. Returns the passed configuration otherwise.

createService

Given a configuration object, createService returns a promise to an initialized service object with methods:

getFormat returns format (or undefined), queried by format identifier and optional version (set to default if not specified)
listFormats returns a list of formats, optionally filtered (see GET /formats for query parameters)

Schemas of format objects can have an additional validate method to validate data in this format.

knownFormats

Object with predefined data formats.

Command Line Interface

The module includes an experimental command line interface:

npm run validate -- --help  # run from repository
validate --help             # if installed via npm

Run Tests

npm test
npm run debug    # test with logging enabled
npm run coverage # test with code coverage report

API

The response status code should always be 200 (possibly including validation errors, unless there was an API error such as wrong request parameters or unexpected internal failures.

GET /validate

Endpoint to validate records passed via query parameter or URL.

Query parameters

format format identifier
url Optional URL to load data from
data Data to be validated. Ignored when parameter url is given.
select Optional selection of records within the posted data. Only supported for JSON-based formats with JSONPath select=$ (default: data is one record) and select=$.* (records are array or object elements).
encoding Optional alternative encoding for (experimental). Supports yaml and ndjson for json-based formats, so far.

Response

Array of same length as the posted data and validation result formeach record. An element is true when the object passed validation, or an array of validation errors when the object failed validation.

Examples

Check whether a simple string such as {} or [x] is valid or invalid JSON:

curl -g 'http://format.gbv.de/validate/validate?format=json&data={}'

[
  true
]

curl -g 'http://format.gbv.de/validate/validate?format=json&data=[x]'

[
  [
    {
      "message": "Unexpected token x in JSON at position 1",
      "position": { "rfc5147": "char=1", "linecol": "1:2" }
    }
  ]
]

JSON parsing errors are returned with location.

The service does not guarantee to return all validation errors but it may stop at the first error.

POST /{format}

Validate records like GET /validate but data is send via HTTP POST payload or as multipart/form-data.

Query parameters or multipart form fields

format format identifier. Can also be specified in the URL path, e.g. /json is identical to /?format=json.
version optional version identifier. Can also be given as part of format as format@version
select optional selection of records within the posted data.
file File to be validated (form data only)
data Data to be validated (form data only)
encoding Optional alternative encoding (experimental)

Response

Same as response of GET /validate.

Examples

This file schema.json contains valid JSON but not a valid JSON Schema:

{
  "properties": []
}

curl -X POST 'http://format.gbv.de/validate/json-schema' --data-binary @schema.json

[
  {
    "message": "must be object",
    "position": {
      "jsonpointer": "/properties"
    }
  }
]

This file articles.json contains two records in vzg-article format, one invalid and one valid:

[
  { },
  {
    "primary_id": {
      "id": "123",
      "type": "unknown"
    },
    "title": "An example article",
    "lang_code": [ "ger" ],
    "journal": {
      "title": "Journal of Examples",
      "year": "2022"
    }
  }
]

To validate both records in one query, parameter select=$.* must be added:

curl -X POST 'http://format.gbv.de/validate/vzg-article?select=$.*' --data-binary @articles.json

[
  [
    {
      "message": "must have required property 'primary_id'",
      "position": {
        "jsonpointer": ""
      }
    }
  ],
  true
]

GET /formats

Lists all data formats, optionally filtered by identifier, version, and/or schema type.

Query Parameters

format=[id] select format with given format identifier
version=[id] version to filter for
type=[string] schema type filter for

Success Response

JSON Array of format objects.

GET /languages

List schema languages as array of data formats. The result is a subset of GET /formats with same query parameters and response format.

GET /schema

Get a schema file.

Query Parameters

format=[id] format identifier
version=[string] optional version (set to default by default)
type=[string] optional schema type

Success Response

The schema file is served with corresponding content type.

Error Resonse

An API error with status code 404 is returned in no corresponding schema was found.

Validation Errors

Validation results (see GET /validate and POST /{format}) can include validation errors. Each error is a JSON object with

message mandatory error message
error optional type of error
position optional object mapping locator formats to locators (e.g.rfc5147 to locate character positions in a string or jsonpointer to reference elements in a JSON document)

Errors may contain additional keys but these may change with future versions of the service.

For instance the following validation error indicates that value of JSON key authors was not given as array:

{
  "message": "must be array",
  "position": {
    "jsonpointer": "/authors"
  }
}

API Errors

Non-validation errors such as wrong request parameters or unexpected internal failures are returned as JSON object such as the following:

{
  "error": "MalformedRequest",
  "status": 400,
  "message": "Missing query parameter: format"
}

A stack trace is included in development mode.

Maintainers

@nichtich

Contributing

PRs accepted against the dev branch. Never directly work on the main branch.

For releases (maintainers only) make changes on dev and then run the release script:

npm run release:patch # or minor or major