JSON Table Schema

A utility library for working with JSON Table Schema in Javascript.

Version v0.2.0 has renewed API introduced in NOT backward-compatibility manner. Previous version could be found here.

Schema - a javascript model of a JSON Table Schema with useful methods for interaction
Field - a javascript model of a JSON Table Schema field
Infer - a utility that creates a JSON Table Schema based on a data sample
Validate - a utility to validate a schema as valid according to the current spec
Table
Goals
Contributing

Installation

npm install jsontableschema

Library requires Promise to work properly, and need to be sure that Promise available globally. You are free to choose any Promise polyfill.

Components

Let's look at each of the components in more detail.

Schema

A model of a schema with helpful methods for working with the schema and supported data. Schema instances can be initialized with a schema source as a url to a JSON file or a JSON object. The schema is initially validated (see validate below), and will raise an exception if not a valid JSON Table Schema.

var Schema = require('jsontableschema').Schema;

var model = new Schema('http://someurl.com/remote.json')

var model = new Schema({JSON OBJECT})

instance always returns Promise

model.then(function(schema) {
    // working code to use schema model
    var fields = schema.fields;
   
}).catch(function(error) {
    // something went wrong and error variable has explanations
})

Following methods are available on Schema instances:

castRow(items, failFast = false, skipConstraints = false) - convert the arguments given to the types of the current schema ¹
descriptor - JSON representation of Schema description
fields - returns an array of Field instances of the schema's fields
foreignKeys - returns the foreign key property for the schema
getField(fieldName, index = 0) - returns an instance of Field by field name (fieldName) ²
hasField(fieldName) - checks if the field exists in the schema by it's name. Returns a boolean
headers - returns an array of the schema headers
primaryKey - returns the primary key field for the schema as an array
save(path) - saves the schema JSON to provided local path. Returns Promise

¹ Where the option failFast is given, it will raise the first error it encounters, otherwise an array of errors thrown (if there are any errors occur)
² Where the optional index argument is available, it can be used as a positional argument if the schema has multiple fields with the same name

Field

Class represents field in the Schema

castValue(value, skipConstraints) - returns a value cast against the type of the field and it's constraints ¹
constraints - returns the constraints object for a given fieldName
format - returns the format of the field
name - returns the name of the field
required - returns boolean
testValue(value, skipConstraints) - returns boolean after a check if value can be casted against the type of the field and it's constraints ¹
type - returns the type of the field

¹ Skip constraints if set to false, will check all the constraints set for field while casting or testing the value

Field types

Data values can be cast to native Javascript types. Casting a value will check the value is of the expected type, is in the correct format, and complies with any constraints imposed by a schema.

{
    'name': 'birthday',
    'type': 'date',
    'format': 'default',
    'constraints': {
        'required': True,
        'minimum': '2015-05-30'
    }
}

Following code will not raise the exception, despite the fact our date is less than minimum constraints in the field, because we do not check constraints of the field descriptor

var dateType = field.castValue('2014-05-29')

And following example will raise exception, because we set flag 'skip constraints' to false, and our date is less than allowed by minimum constraints of the field. Exception will be raised as well in situation of trying to cast non-date format values, or empty values

try {
    var dateType = field.castValue('2014-05-29', false)
} catch(e) {
    // uh oh, something went wrong
}

Values that can't be cast will raise an Error exception.
Casting a value that doesn't meet the constraints will raise an Error exception.
Note: the unique constraint is not currently supported.

Available types, formats and resultant value of the cast:

Type	Formats	Casting result
string	default¹, uri, email, binary	String
integer	default	Number
number	default, currency	Number²
boolean	default	Boolean
array	default	Array
object	default	Object
date	default, any, fmt	Date object
time	default, any, fmt	Date object
datetime	default, any, fmt	Date object
geopoint	default, array, object	Accordingly to format³
geojson	default, topojson	Accordingly to format^3,4

¹ default format can be not specified in the field descriptor
² in case value has 00 after point (1.00), it will return Number(1).toFixed(2), which is actually String '1.00'
³ default format returns String
⁴ topojson is not implemented

Infer

Given headers and data, infer will return a JSON Table Schema as a JSON object based on the data values. Given the data file, example.csv:

id,age,name
1,39,Paul
2,23,Jimmy
3,36,Jane
4,28,Judy

Call infer with headers and values from the datafile:

var parse = require('csv-parse');
var fs = require('fs');
var infer = require('jsontableschema').infer;
 
fs.readFile('/path/to/example.csv', function(err, data) {
  parse(data, function(error, values) {
    var headers = values.shift()
        , schema = infer(headers, values);
  });
});

The schema variable is now a JSON object:

{
  fields: [
    {
      name: 'id',
      title: '',
      description: '',
      type: 'integer',
      format: 'default'
    },
    {
      name: 'age',
      title: '',
      description: '',
      type: 'integer',
      format: 'default'
    },
    {
      name: 'name',
      title: '',
      description: '',
      type: 'string',
      format: 'default'
    }
  ]
}

It possible to provide additional options to build the JSON schema as 3rd argument of infer function. It is an object with following possible values:

rowLimit (integer) - limit number of rows used by infer
explicit (boolean) - add required constraints to fields
primaryKey (string, array) - add primary key constraints
cast (object) - object with cast instructions for types in the schema. For example:

var parse = require('csv-parse');
var fs = require('fs');
var infer = require('jsontableschema').infer;
 
fs.readFile('/path/to/example.csv', function(err, data) {
  parse(data, function(error, values) {
    var headers = values.shift(),
        options = {
          rowLimit: 2,
          explicit: true,
          primaryKey: ['id', 'name'],
          cast: {
            string : { format : 'email' },
            number : { format : 'currency' },
            date: { format : 'any'}
          } 
        },
        schema = infer(headers, values, options);
  });
});

The schema variable will look as follow:

{
  fields: [
    {
      name: 'id',
      title: '',
      description: '',
      type: 'integer',
      format: 'default',
      required: true
    },
    {
      name: 'age',
      title: '',
      description: '',
      type: 'integer',
      format: 'default',
      required: true
    },
    {
      name: 'name',
      title: '',
      description: '',
      type: 'string',
      format: 'default',
      required: true
    }
  ],
  primaryKey: ['id', 'name']
}

In this example:

rowLimit: only two rows of values from example.csv will be proceed to set field type. It can be useful in cases when data in CSV file is not normalized and values type can be different in each row. Consider following example:

id,age,name
1,39,Paul
2,23,Jimmy
3,thirty six,Jane
four,28,Judy

In this case by limiting rows to 2, we can build schema structure with correct field types

cast: every string value will be casted using email format, number will be tried as a currency format, and date - as any format

Validate

Given a schema as JSON object, validate returns Promise, which success for a valid JSON Table Schema, or reject with array of errors.

var validate = require('jsontableschema').validate;
var schema = {
   fields: [
     {
       name: 'id',
       title: '',
       description: '',
       type: 'integer',
       format: 'default'
     },
     {
       name: 'age',
       title: '',
       description: '',
       type: 'integer',
       format: 'default'
     },
     {
       name: 'name',
       title: '',
       description: '',
       type: 'string',
       format: 'default'
     }
   ]
};
 
validate(schema).then(function() {
  // do something with valid schema here
}).catch(function(errors) {
  // uh oh, some validation errors in the errors array
})

Note: validate() validates whether a schema is a validate JSON Table Schema accordingly to the (specifications)[http://schemas.datapackages.org/json-table-schema.json]. It does not validate data against a schema.

Table

A javascript model of a table (schema+source of data)

Instance always returns Promise. In case if schema object is not valid, it will reject promise.

Source of data can be:

array of objects with values, represent the rows
local CSV file
remote CSV file (URL)
readable stream

Following methods are available on Table instances:

iter(callback, failFast, skipConstraints)^1,2 - iterate through the given dataset provided in constructor and returns converted data
read(keyed, extended, limit) - Read part or full source into array.
- keyed: row looks like {header1: value1, header2: value2}
- extended: row looks like [row_number, [header1, header2], [value1, value2]].
  - Low-level usage: when you need all information about row from stream but there is no guarantee that it is not malformed. For example, in goodtables you cannot use keyed because there is no guarantee that it will not fail - https://github.com/frictionlessdata/goodtables-py/blob/master/goodtables/inspector.py#L205
  - High-level usage: useful when you need to get row + row number. This row number is exact row number of source stream row. It's not like counted or similar. So if you skip first 9 rows using skipRows first row number from iter(extended=True) will be 10. It's not possible to get this information on client code level using other approach - iter() index in this case will start from 0.
- limit: limt the number of rows return to limit
save(path) - Save source to file locally in CSV format with , (comma) delimiter. Returns Promise

¹ If failFast is set to true, it will raise the first error it encounters, otherwise an array of errors thrown (if there are any errors occur). Default is false
² Skip constraints if set to true, will check all the constraints set for field while casting or testing the value. Default is false

var jts = require('jsontableschema');
var Table = jts.Table;
 
var model = new Table({SCHEMA}, {SOURCE})
var callback = function(items) {
    // ... do something with converted items
    // iter method convert values row by row from the source
}
model.then(function (table) {
    table.iter(callback, true, false).then(function() {
          // ... do something when conversion of all data from source is finished 
    }, function (errors) {
          // something went wrong while casting values from source
          // errors is array with explanations
    })
}, function(error) {
    // Table can't instantiate for some reason
    // see error for details
})

Goals

A core set of utilities for working with JSON Table Schema
Use in other packages that deal with actual validation of data, or other 'higher level' use cases around JSON Table Schema (e.g. Tabular Validator)
Be 100% compliant with the the JSON Table Schema specification (we are not there yet)

Contributing

Please read the contribution guideline:

How to Contribute

Thanks!