Support for coercing data types

[alexiskat]

Hello,

I was wondering if there is any support or plans to support coercing of data types ? I am referring to the feature that can be seen here for AJV. A scaler value is a string in the data passed to the validator but the json schema declares that it should be an integer so it tries to convert it to an integer

I want to validate a queryStringParameters in a lambda event as an integer value (has been converted to a string by lambda) and would like to use this feature to convert the string to an integer using the json schema

thanks

[leandrodamascena]

Hi @alexiskat! Thanks for opening this discussion.

We already have the validation utility that can validate input against a JSON schema, but it doesn't support data type coercion before validating it. In this utility we use fastjsonschema, but it also does not support coercion operations before validating the type. Still in this JSON schema space, I was playing around with Cerberus a few months ago and it looks like it supports data coercion before validating, so you can bring it in as a dependency (~300kb) and use it.

But, if you want a more robust library to do that, I would suggest to take a look into Pydantic and see if it make sense for your use case. Pydantic supports Data validation and you can coerce types before of after the validation.

For now, we have no plans to change the implementation of the validation utility because that would require a major release and writing an RFC to understand the impact, but we are always open to hearing suggestions from customers on how we can improve our utilities.

Please let me know your thoughts.

[alexiskat]

hi @leandrodamascena,
thanks for the quick response.

I may be missing something but it seem like pretty useful feature to have, especially since the event dict in python lambda means that all the values get converted to string and using the json schema you can validate it them against diffrent types. Like you mention a few frameworks do this already.

At the moment we really don't want to add another dependency to our lambda. As a work around we are using the following in the schema to validate integer numbers.

"foo": {
   "type": "string",
   "pattern": "^[0-9]+$"
} 

thanks
alexis