json-schema-merger

create json schema from pojo and merge schemas https://www.npmjs.com/package/json-schema-merger

Motivation

this was part of my smarthome development the problem description was having an "unknown" node provide a series of json strings and then through analysis of the different jsons being able to create a schema (https://json-schema.org/) that describes the dataset

example: emit1:

    {
        "test":123
    }

emit2:

    {
        "test":456
    }

then the resulting schema should describe an object that has a test property that can be either 123 or 456 or in ts:

   type schema={test:123|456}

if the value is a date or something hat frequently changes it should be collapsed from an enum/const/type union to a more generic type at latest after a few emits(i picked 10 as a default)

create Schema

create a json schema from a json object:

const myRandomObject={
    test: 123
}

const schema=createSchema(myRandomObject)
/*
{
    type: "object",
    properties: {
        test: {
            type: "number",
            enum: [
                123,
            ],
        },
    },
    additionalProperties: false,
    required: [
        "test",
    ],
}
*/

by default all values are treated as constant values

this can be changed by passing a second argument to createSchema

createSchema(myRandomObject,{disableAssumeConst:true})

schema merging

assuming 2 schemas of similar json objects this can be used to update and refine the initial schema

const schema1={...}
const schema2={...}

// currently modifies "old" schema
schema1 = mergeSchema({
    old: schema1,
    new: schema2,
    params: { mergeLength: 10 }
})

if the amount of different values for any given property exceed mergeLength it will be merged from an enum to a generic

example:

{
    type:"string",
    enum:["1","2","3","...10"]
}

will be merged to just

{
    type:"string"
    merged:true //indicator for future attempts
}

params

params support individual mergeLength per field in the form of a dot separated path string

{
    a:{
        b:123
    }
}

=>

{
    mergeLength: 10 
    "a.b":5
}

if enumKeyList is passed it will be filled with all found object paths (and can be used to generate the param list)

Schema Validation

let say you have a schema defining what is allowed as an input and a schema defining what is provided to that input

then u can use

validateJsonSchemas({
    assigning:providingSchema
    target:inputSchema
})

then if the providing schema cannot be assigned (isnt a subset) of the inputSchema it will throw an error

TODO:

• add more types to the schema and merging