Question: Is there a way to supply multiple schemas for compilation?

I have a set of growing and somewhat redundant schemas that I’d like to break down into smaller “main” schemas with shareable “subschemas” using the $ref keyword. The schemas needed by the application would all be known at initialization time; I have no need for dynamic loading of schemas.

With my current (large) schemas, I call the compile function once at initialization/module load time as suggested here.

    const ajv = new Ajv({ allErrors: true })
    const validate = ajv.compile(myBigOriginalSchema1) 

After reading the documentation and the API reference I’m unclear (I’m sorry) on whether there is a way to have ajv compile multiple schemas at once (or whether that is even advisable). I notice the compile function only takes a single schema.

For example, I tried this, with addSchema:

    const ajv = new Ajv({ allErrors: true })
    ajv.addSchema(mySubschema1)
    ajv.addSchema(mySubschema2)
    const validate = ajv.compile(mySmallerMainSchema1) 

…and I looking at the source, I don’t believe it compiles the other schemas added via addSchema at the time of the compile call.

I also read here that the addSchema function itself doesn’t actually compile the schemas by design, until they are first needed. In my use case I KNOW these subschemas will very likely used on the first validation attempt. Breaking the large single schema down into a bunch (e.g. a dozen) subschemas will likely mean I am deferring most/all the compilation to the first validation attempt instead of compiling at initialization. Is my understanding correct? Is what I have above the most efficient approach? Is there a way to compile multiple schemas at initialization time instead?

I also experimented with the following, but I’m not sure it’s any different…I thought the constructor simply called addSchema on each schema given:

    const ajv = new Ajv({ allErrors: true, schemas: [mySubschema1, mySubschema2] })
    const validate = ajv.compile(mySmallerMainSchema1) 

Just to further try and illustrate what I’m asking about, I almost want to do something like this, but the compile function needs a schema argument:

    const ajv = new Ajv({ allErrors: true, schemas: [mySmallerMainSchema1, mySubschema1, mySubschema2] })
    const validate = ajv.compile() //needs a schema here, obviously

…OR like this, but I know this argument isn’t supported either:

    const ajv = new Ajv({ allErrors: true })
    const validate = ajv.compile({ schemas: [mySmallerMainSchema1, mySubschema1, mySubschema2] })  //only one allowed

Any advice is appreciated. And if I am missing the point of how subschemas and $ref are supposed to be used, please let me know.

Thank you for creating & maintaining a useful library in ajv.

[wcaneira]

I have a sample working with addSchema/getSchema, so I can close this issue, but before I do, one related follow-up, please - are code functions appropriate for database/HTTP service call validation? I see your range example but I feel like that’s a schema-based validation rule with no external dependencies. I’m struggling to see how my use case would be a good fit with a function that ends with ctx.fail or ctx.pass and string templates. Is a validate function (what I’m currently using) more appropriate for database/HTTP service-based validation?

This link was broken so I couldn’t see the additional examples

I assume it leads to the ajv-keyword examples you mentioned earlier, which I did look over, but I only see a couple examples there, and I’m still unclear. I’m sorry.

I’ll close this issue after receiving your reply.

[epoberezkin]

This is the folder where you can see the examples: https://github.com/ajv-validator/ajv/tree/master/lib/vocabularies/validation

are code functions appropriate for database/HTTP service call validation

they are used for all standard keywords, but they are only used during schema compilation. HTTP service call validation is the main use case, it’s all ok.