bug: type order in 'allOf' has an impact

[kroc702]

Hello,
When validate with multiple schema using 'allOf', definition order has a behavior impact.
cf attached code to test
testSchemaValidator2.zip

Correct:

input value: "true"
schema: {
    "allOf": [
        {
            "type": "string"
        },
        {
            "type": "boolean"
        }
    ]
}
value is valid

reverse order lead to incorrect validation:

input value: "true"
schema: {
    "allOf": [
        {
            "type": "boolean"
        },
        {
            "type": "string"
        }
    ]
}
Boolean value found, but a string is required
Failed to match all schemas

[kroc702]

I forget to mention version:

            "name": "justinrainbow/json-schema",
            "version": "v5.2.13",
            "source": {
                "type": "git",
                "url": "https://github.com/justinrainbow/json-schema.git",
                "reference": "fbbe7e5d79f618997bc3332a6f49246036c45793"
            },

[DannyvdSluijs]

@kroc702 thanks for the report. Based on your schema provided it seems no input data will be valid ever as the a single value cannot be boolean and string at the same time. Using https://jsonschemalint.com/ you can see that both versions of you schema should raise errors whereas you mention that in both cases you expect them to be valid.

[kroc702]

Well it's a choice, your choice ;)
As php is loosely typed, it doesn't hurt me to validate the strings "true" or "1" as boolean or number. Changing this behavior may be a braking change for the lib.

We just need a consistent behavior with all types and order shouldn't impact behavior of any/all/oneOf.

[DannyvdSluijs]

@kroc702 I now realise how the allOf would match the schemas in combination with the type coercion.

It seems the issue can be reproduced in the 5.2.13 and 5.3.0 releases and in the 5.x.x branch. In the releases 6.x and the master branch I can't reproduce the issue. The fixes have been made in #384 and are part of the 6.x release. although a backport was considered it was skipped for 5.x versions.

I would recommend to either update to version 6.x or see if you can alter your schema.

[kroc702]

Hi, and thank to keep track of this issue.

I'm not sure what "not planned" mean... It's fixed in version 6.x ?
I didn't test it yet but I understood yes, so thank !