go-jsonschema is a tool to generate Go data types from JSON Schema definitions.
This tool generates Go data types and structs that corresponds to definitions in the schema, along with unmarshalling code that validates the input JSON according to the schema's validation rules.
-
Download: Get a release here.
-
Install from source: To install with Go 1.16+:
$ go get github.com/atombender/go-jsonschema/...
$ go install github.com/atombender/go-jsonschema@latest
- Install with Brew: To install with Homebrew:
$ brew tap omissis/go-jsonschema
$ brew install go-jsonschema
This project makes use of go workspaces in order to ease testing of the generated code during development while keeping the codebase as tidy and maintainable as possible. It is bit of an unusual choice, but it allows to not only test the code-generation logic, but also the generated code itself.
At its most basic:
$ go-jsonschema -p main schema.json
This will write a Go source file to standard output, declared under the package main
.
You can generate code for multiple schemas in the same invocation, optionally writing to different files inside different packages:
$ go-jsonschema \
--schema-package=https://example.com/schema1=github.com/myuser/myproject \
--schema-output=https://example.com/schema1=schema1.go \
--schema-package=https://example.com/schema2=github.com/myuser/myproject/stuff \
--schema-output=https://example.com/schema2=stuff/schema2.go \
schema1.json schema2.json
This will create schema1.go
(declared as package myproject
) and stuff/schema2.go
(declared as package stuff
). If schema1.json
refers to schema2.json
or vice versa, the two Go files will import the other package that it depends on. Note the flag format:
--schema-package=https://example.com/schema1=github.com/myuser/myproject \
^ ^
| |
schema $id full import URL
In a few cases, special types are used to help with serializing/deserializing data frrom JSON. Namely a custom types is provided for the following semantic types:
SerializableDate
SerializableTime
These types are needed because there is no native type provided by Go which properly handles them.
While not finished, go-jsonschema can be used today. Aside from some minor features, only specific validations remain to be fully implemented.
- Core (RFC draft)
- Data model (§4.2.1)
-
null
-
boolean
-
object
-
array
-
number
- Option to use
json.Number
- Option to use
-
string
-
- Location identifiers (§8.2.3)
- References against top-level names:
#/$defs/someName
- References against nested names:
#/$defs/someName/$defs/someOtherName
- References against top-level names in external files:
myschema.json#/$defs/someName
- References against nested names:
myschema.json#/$defs/someName/$defs/someOtherName
- References against top-level names:
- Comments (§9)
- Data model (§4.2.1)
- Validation (RFC draft)
- Schema annotations (§10)
-
description
-
default
(only for struct fields) -
readOnly
-
writeOnly
-
(N/A)title
-
(N/A)examples
-
- General validation (§6.1)
-
enum
-
type
(single) -
type
(multiple; note: partial support, limited validation) -
const
-
- Numeric validation (§6.2)
-
multipleOf
-
maximum
-
exclusiveMaximum
-
minimum
-
exclusiveMinimum
-
- String validation (§6.3)
-
maxLength
-
minLength
-
pattern
-
- Array validation (§6.4)
-
items
-
maxItems
-
minItems
-
uniqueItems
-
additionalItems
-
contains
-
- Object validation (§6.5)
-
required
-
properties
-
patternProperties
-
dependencies
-
propertyNames
-
maxProperties
-
minProperties
-
- Conditional subschemas (§6.6)
-
if
-
then
-
else
-
- Boolean subschemas (§6.7)
-
allOf
-
anyOf
-
oneOf
-
not
-
- Semantic formats (§7.3)
- Dates and times
- Email addresses
- Hostnames
- IP addresses
- Resource identifiers
- URI-template
- JSON pointers
- Regex
- Schema annotations (§10)
MIT license. See LICENSE
file.