@terryweiss/jsm v2.0.20
JSON Schema Model (jsm)
Generates TypeScript decorated with class-validator decorators based on a json-schema.
State of Development
This project is envisioned as a way of gaining access to strongly typed definitions/components from
the Swagger json-schema dialect, with pure json-schema a nice-to-have. But some differences from Swagger
to json-schema were difficult to reconcile and led to headaches.
Finally, class-validator accepts json-schema in its validation tools, but the decorators
are a disjoint language against Swagger and json-schema. Meaning there are 3 potential sources of validation information.
In the end, the specific keywords supported were
driven as much by what was possible as what I thought was "right".'
Which might beg the question: why do all this? Well, one answer is that this project makes sense if you consider json-schema/Swagger as a means of validating serialized objects in transit or at rest. But once the JSON was hydrated, there is a role for the object that contains the hydrated instance to continue the original schema's validation with run time checks as well as the values it mutates along its lifetime.
This library does not try to connect the lifetimes of the JSON and the object in any way. If you use
Swagger, the swagger-tools will pull these objects out of definitions and paths and return you a
plain old JS object - you get it in the request object. You can pass that POJO into one of these constructors and once validated, serialize
it again knowing it has to be valid. Anyway, that's the idea.
Contributors: You are welcome! Some of what this thing lacks:
- unit tests
- A means for allowing consumers to substitute their own generation templates
- Code review
- Better documentation
To get a good feel for what this does, check the repo's models/ and out/ directories and consider each file pair .yaml -> .ts
Installation
This is a CLI tool, not a library so you will need to install it globally.
sudo npm install @terryweiss/jsm -gTechnical Docs
A description of the internal API for contributors can be found on github or locally.
CLI
This is a static code generator, there is no API, just a CLI. The options are:
--helpShow help--versionShow version number--log-levelSets the logging level for the process. Choices are "trace", "debug", "warn", "data", "log", "info", "warn", "error"-i, --infilesThe path to your schema files, can be a glob to yaml or json files or a mix thereof. It uses a globbing library, so you can be creative.-o, --outpathThe path to write the files to. Class paths are appended to this-r, --rootNameWhen dealing with a schema that does not contains a root element, you can name the resultant object with this. This is only valuable when you generating from a single file.
Examples
You can find a set of hcard derived models in the models/ folder. Output is found in out/ in the repo.
Validation keywords
json.schema.model supports the common subset of keywords from swagger and jon-schema. Where the interpretations are
in conflict, this tends to prefer swagger.
Common keywords
nameThis is specific tojson.schema.modeland is used to name a class from the key of the object definition. it is auto-populated and is available in the template that generates the code.
{
"MyModel": { // <-- this is the name property of the runtime generator
"type": "object",
"properties":{}
}
}title:stringas in json-schemadescription:stringas in json-schemax-is-defined:booleanExtension for class-validator, checks if value is defined (!== undefined, !== null)required:arrayas in json-schemaconst:stringas in json-schema. Also asx-equals. Checks if value equals ("===") comparison.x-not-equalsas in class-validator. Checks if value not equal ("!==") comparison.x-empty:booleanas in class-validator. Checks if given value is empty (=== '', === null, === undefined).x-not-empty:booleanas in class-validator. Checks if given value is not empty (!== '', !== null, !== undefined).x-in:arrayas in class-validator. Checks if value is in a array of allowed values.x-not-in:arrayas in class-validator. Checks if value is not in a array of disallowed values.x-defaultspecify a default value for the propertystring,number,integer,boolean,array,objectdata type keywords as in json-schema
Object keywords
x-model-name:stringspecify a name for the generated classx-model-path:stringspecify a path (starting fromoutFilesoption) that the module should be writtenx-namespace:stringUsed for documentation only when defining the@modulekeywordx-extends:stringIf this is a subclass of some other class, put that class path here
Array keywords
minItems,maxItems,uniqueItemsas in json-schemax-contains:arrayas in class-validator. Checks if array contains all values from the given array of values.x-not-contains:arrayas in class-validator. Checks if array does not contain any of the given values.items:arrayas in swagger which is always an array
Date keywords
x-min-date:Date|stringas in class-validator. Checks if the value is a date that's after the specified date.x-max-date:Date|stringas in class-validator. Checks if the value is a date that's before the specified date.
Integer/Number keywords
minimum,maximum,multipleOfas json-schemax-positive-number:booleanas in class-validator. Checks if the value is a positive number.x-negative-number:booleanas in class-validator. Checks if the value is a negative number.
String keywords
minLength,maxLength,patternas in json-schemax-contains:stringas in class-validator. Checks if the string contains the seed.x-not-containsas in class-validator. Checks if the string not contains the seed.
String formats
This tool tries to supply the union of what json-schema and class-validator defines for strings. Admittedly, not well. But there it is. json_schema
hostnameemailipv4ipv6uridate-time,date,timeThese all create aDateobject
class-validator
base64asciialphanumericalphanumber-stringdate-stringboolean-stringiso8601militaryTimejsonlowerCaseupperCasemongoIduuid
Using JSM
As stated above, this is a CLI tool. At it's simplest you can create models from a directory like so:
jsm -f "models/*.yaml" -o ../src/models From a swagger file:
jsm -f "api.yaml" -o ../src/models
# or for a json file
jsm -f "api.json" -o ../src/modelsUsing globs you can get a whole directory and its subdirectories:
jsm -f "models/**/*.yaml" -o ../src/modelsWhen you produce your files, you can access them like any other file:
const {Address} = require("./models/Address");
const adr = new Address();Let's say you have an existing object you got from a CSV file. You can pass that object in (once it has been shaped) and validate it right away:
const {Address} = require("./models/Address");
const val = someJsonFromSomewhere();
const adr = new Address(val);
adr.validate().then((validationErrors)=>{
if (validationErrors.length === 0){
// do something cool
}else{
console.log("Validation failed:\n", validationErrors);
}
}).catch((e)=>{
console.error(e);
})To get all this delicious sugar in your bloodstream, you must install jsm as above. But you must also install three other dependencies in the project that gets the output:
sudo npm install @scrawl/json.schema.model -g
# then install local runtime dependencies
npm install class-transformer-validator class-transformer class-validator --saveSome more notes
If you are not using TypeScript, you will need to take one further step to use the generated files. You must install the TypeScript compiler and use it turn the generated files into something usable. It does require a few special flags to compile correctly.
tsc --target es2015 --lib 'ES2015' --module 'commonjs' --sourceMap --outDir ./myDir myFile # or filesAdd this command to your package.json, webpack, gulp, or grunt file.
Some more more notes
The files that are generated are completely generated every time. Any changes you make to a generated file will be lost the next time you run jsm. But, these are just classes, and simple classes at that. To customize the behavior of these classes, just inherit from them and customize your behavior there.