Skip to content
This repository


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

jesse (JSon Schema Erlang) is an implementation of a json schema validator for Erlang.

branch: master

jesse Build Status

jesse (JSon Schema Erlang) is an implementation of a json schema validator for Erlang.

jesse implements Draft 03 of the specification. It supports almost all core schema definitions except:

  • format
  • $ref

Quick start

There are two ways of using jesse:

  • to use jesse internal in-memory storage to keep all your schema definitions In this case jesse will look up a schema definition in its own storage, and then validate given json.
  • it is also possible to provide jesse with schema definitions when jesse is called.


NOTE: jesse doesn't have any parsing functionality. It currently works with three
      formats: mochijson2, jiffy and jsx, so json needs to be parsed in advance,
      or you can specify a callback which jesse will use to parse json.

      In examples below and in jesse test suite jiffy parser is used.
  • Use jesse's internal in-memory storage:

(parse json in advance)

1> Schema = jiffy:decode(<<"{\"items\": {\"type\": \"integer\"}}">>).
2> jesse:add_schema(some_key, Schema).
3> Json1 = jiffy:decode(<<"[1, 2, 3]">>).
4> jesse:validate(some_key, Json1).
5> Json2 = jiffy:decode(<<"[1, \"x\"]">>).
6> jesse:validate(some_key, Json2).

The [1] in the error is the path in the original value to <<"x">> where the validation failed. See Validation errors below for the full error format.

(using a callback)

1> jesse:add_schema(some_key,
1>                  <<"{\"uniqueItems\": true}">>,
1>                  [{parser_fun, fun jiffy:decode/1}]).
2> jesse:validate(some_key,
2>                <<"[1, 2]">>,
2>                [{parser_fun, fun jiffy:decode/1}]).
{ok,[1, 2]}
3> jesse:validate(some_key,
3>                <<"[{\"foo\": \"bar\"}, {\"foo\": \"bar\"}] ">>,
3>                [{parser_fun, fun jiffy:decode/1}]).
  • Call jesse with schema definition in place (do not use internal storage)

(parse json in advance)

1> Schema = jiffy:decode(<<"{\"pattern\": \"^a*$\"}">>).
2> Json1 = jiffy:decode(<<"\"aaa\"">>).
3> jesse:validate_with_schema(Schema, Json1).
4> Json2 = jiffy:decode(<<"\"abc\"">>).
5> jesse:validate_with_schema(Schema, Json2).

(using a callback)

1> Schema = <<"{\"patternProperties\": {\"f.*o\": {\"type\": \"integer\"}}}">>.
<<"{\"patternProperties\": {\"f.*o\": {\"type\": \"integer\"}}}">>
2> jesse:validate_with_schema(Schema,
2>                            <<"{\"foo\": 1, \"foooooo\" : 2}">>,
2>                            [{parser_fun, fun jiffy:decode/1}]).
3> jesse:validate_with_schema(Schema,
3>                            <<"{\"foo\": \"bar\", \"fooooo\": 2}">>,
3>                            [{parser_fun, fun jiffy:decode/1}]).
  • Since 0.4.0 it's possible to say jesse to collect errors, and not stop immediately when it finds an error in the given json:
1> Schema = <<"{\"properties\": {\"a\": {\"type\": \"integer\"}, \"b\": {\"type\": \"string\"}, \"c\": {\"type\": \"boolean\"}}}">>.
<<"{\"properties\": {\"a\": {\"type\": \"integer\"}, \"b\": {\"type\": \"string\"}, \"c\": {\"type\": \"boolean\"}}}">>
2> jesse:validate_with_schema(Schema,
2>                            <<"{\"a\": 1, \"b\": \"b\", \"c\": true}">>,
2>                            [{parser_fun, fun jiffy:decode/1}]).

now let's change the value of the field "b" to an integer

3> jesse:validate_with_schema(Schema,
3>                            <<"{\"a\": 1, \"b\": 2, \"c\": true}">>,
3>                            [{parser_fun, fun jiffy:decode/1}]).

works as expected, but let's change the value of the field "c" as well

4> jesse:validate_with_schema(Schema,
4>                            <<"{\"a\": 1, \"b\": 2, \"c\": 3}">>,
4>                            [{parser_fun, fun jiffy:decode/1}]).

still works as expected, jesse stops validating as soon as finds an error. let's use the 'allowed_errors' option, and set it to 1

5> jesse:validate_with_schema(Schema,
5>                            <<"{\"a\": 1, \"b\": 2, \"c\": 3}">>,
5>                            [{parser_fun, fun jiffy:decode/1},
5>                             {allowed_errors, 1}]).

now we got a list of two errors. let's now change the value of the field "a" to a boolean

6> jesse:validate_with_schema(Schema,
6>                            <<"{\"a\": true, \"b\": 2, \"c\": 3}">>,
6>                            [{parser_fun, fun jiffy:decode/1},
6>                             {allowed_errors, 1}]).

we stil got only two errors. let's try using 'infinity' as the argument for the 'allowed_errors' option

7> jesse:validate_with_schema(Schema,
7>                            <<"{\"a\": true, \"b\": 2, \"c\": 3}">>,
7>                            [{parser_fun, fun jiffy:decode/1},
7>                             {allowed_errors, infinity}]).

Validation errors

The validation functions jesse:validate/2 and jesse:validate_with_schema/2,3 return {ok, Value} on success and {error, ListOfErrors} on failure. An error is either data_invalid or schema_invalid.

A data_invalid error is a tuple on the form {data_invalid, Schema, ErrorType, Value, Path} where

  • Schema is the part of the schema where validation failed
  • ErrorType is the type of error, usually an atom such as wrong_type, not_in_range or no_match
  • Value is The part of the value where failed validation agains Schema
  • Path is a path to where validation failed within the original value. The path is a list of property names and zero-based array indices referencing the properties and array items within a JSON document; e.g. in the JSON document {"foo": [42, 43, 44]}, the path [<<"foo">>, 0] refers to the value 42. An empty list refers to the whole JSON document.

A schema_invalid error is a tuple on the form {schema_invalid, Schema, ErrorType} where

  • Schema is the part of the schema which is invalid
  • ErrorType is an atom such as missing_id_field or a tuple such as {wrong_type_dependency, Dependency}.


  • pattern and patternProperty attributes:

    jesse uses standard erlang module re for regexp matching, therefore there could be some incompatible regular expressions in schemas you define.

    From erlang docs: "re's matching algorithms are currently based on the PCRE library, but not all of the PCRE library is interfaced"

    But most of common cases should work fine.


If you see something missing or incorrect, a pull request is most welcome!

Something went wrong with that request. Please try again.