Changing the resolution base via 'id' does not work properly

[erayd]

Per issue #446. See below test case:

$data = json_decode(<<<'eot'
{
    "propertyOne": "blue"
}
eot
);

$base = json_decode(<<<'eot'
{
    "id": "schema.json",
    "properties": {
        "propertyOne": {
            "$ref": "propertyOne.json"
        }
    }
}
eot
);

$validator = new \JsonSchema\Validator();
var_dump($validator->validate($data, $base));

@bighappyface Could you please mark this as blocking for 6.0.0? I need to stare at the code some more, but at first look this may require significant refactoring of the $ref code.

[erayd]

@shmax Have you had anything to do with with the ref processing when the scope has been changed? Some insight into how this is intended to work would be handy and would save a lot of time tracing stuff. No worries if not.

The tests in draft4/refRemote.json all pass, including the ones with a nested scope change (e.g. folderInteger.json), but the logic that actually achieves this isn't clear, and it's not covering all cases. In the test code above, the resolution base gets completely dropped somewhere along the line, despite being explicitly set in the base schema.

[shmax]

No, not only was that before my time, but I never really totally figured out how to get $ref functionality working in my own code; I had to resort to overriding the UriResolver and just cheeseballing the resolve method. I looked over this issue, but I think I'm a bit out of my depth; I don"t even know what a "behat" is. I'll see if I can get caught up a bit more this weekend.

[erayd]

Bugger. Will get to tracing stuff then. Thanks anyway :-).

[erayd]

For what it's worth, this bug has nothing to do with Behat. Behat just happens to depend on this validator, and @dkarlovi happened to find the bug while using Behat.

[shmax]

Gotcha. I'll load up his sample app tomorrow and see if I can make any sense of it.

[erayd]

No need, unless you're curious - the test case at the top of this thread reproduces the bug using only the json-schema validator.

[dkarlovi]

Behat is irrelevant here, I used it to replicate the bug the only way I knew how. :)

[mdeboer]

Same problem here and I have no clue how to get it to work except for writing my own resolver I guess. Can't we just add a getter/setter for the base path for the resolver? This would make all relative file references relative to that path.

I can probably cook something up but I'm not sure if it would break stuff as this is a bug apparently.

[erayd]

@mdeboer This is definitely a bug. I will fix it.

Regarding your suggestion of a base path setter... what purpose would this serve? The spec already defines the id keyword as filling this role; it's just the implementation which is broken.

[mdeboer]

Alright, awaiting the fix then 👍

[erayd]

@mdeboer See #448. Assuming code review is OK, I will backport this fix to 5.x.x after merge.