ruzicka / to-json-schema Goto Github PK

Seems like you accidentally put mocha in dependencies instead of devDependencies.

Hi, interesting project.

I was wondering if you had given any thought about how to do anyOf support in maybe a custom function

strings which are not "style" are inferred as styles.
for example, given this code:

import toJsonSchema from 'to-json-schema'

const json = {
  'notAStyle': '10:00',
}

console.log(toJsonSchema(json))

the following json schema is generated:
{
"type": "object",
"properties": {
"notAStyle": {
"type": "string",
"format": "style"
}
}
}

I think the regex here can be improved to avoid this mistakes: https://github.com/ruzicka/to-json-schema/blob/master/src/jsonSchemaHelpers.js#L74

I would love to upload PR for this one if possible

Is there any way to show the nested objects which I do not know the schema of?

Im getting:

{
  type: 'object',
  properties: {
    bpi: { type: 'object', properties: [Object] },
    disclaimer: { type: 'string' },
    time: { type: 'object', properties: [Object] }
  }
}

where I want:

{
  type: 'object',
  properties: {
    bpi: { type: 'object', properties: { valueIdontknowA: { type: 'string' } , valueIdontknowB: { type: 'string' } } },
    disclaimer: { type: 'string' },
    time: { type: 'object', properties: 
        { valueA: { type: 'string' } , valueB: { type: 'string' } }  
  }
  }
}

Conversion attempt of ["1", "two", "3", "four"] to schema throws exception with message "Failed to load schema: Cannot read properties of null (reading 'type')".
The cause is that mergeSchemaObjs returns null because list items schemas restored as {"type": "string", "format": "utc-millisec"} for "1" and "3", and {"type": "string"} for the rest.

Hi there -- not sure if this project is very active but figured I'd post this for the benefit of anyone else wondering.

I believe the required: true notation used to indicate properties are required is not valid JSON-schema syntax in any draft. I tested this by pasting the generated output of to-json-schema into a JSON schema validator and checking it against the existing JSON schema drafts!

That said, there appears to be an easy fix, which would be to use the example you define farther down:

const options = {
  objects: {
    postProcessFnc: (schema, obj, defaultFnc) => ({...defaultFnc(schema, obj), required: Object.getOwnPropertyNames(obj)})
  }
};

Thanks for this library @ruzicka, it's nicely engineered!

See test case here: https://github.com/ruzicka/to-json-schema/compare/master...forivall:forivall%2Ftest-array-of-arrays?diff=unified

If i add a string to the second array (like [['foo'], ['bar']]), then the test case passes. As is, the generated schema is:

{
  "type": "array",
  "items": {
    "type": "array"
  }
}

options.object, options.string, options.postProcessFnc are not flexible enough to handle complex cases, such as patch/merge some schema fileds on toJsonSchema processing.

For details:

1. Lack of array options.

2. Lack of options.preProcessFnc

3. Lack of object.keys and array.index hook

What I really need is the node path during toJsonSchema iteration process, something like:

options.postProcessFnc(nodePath = ['a', 1, 'b'], value = 'myValue', schema) {
  // patch on result schema
  if (nodePath.join('.') === 'a.1.b') {
    schema.example = 'my code';
    schema.anything = 'xxxx';
  }
  return schema;
}

But at this point, documented API seems not working for this case (especially array index), any idea or workaround would be also appreciated.

Hi,

I'd like to define a JSON file such as

{
 "user" : "sample"
}

And save it as user.json and then run something such as to-json-schema users.json userschema.json in order to create a schema file for the users.json file