Can't instantiate Validator() - TypeError

[dosena]

Trying the example for is_valid() from https://python-jsonschema.readthedocs.org/en/latest/validate.html

schema = {"maxItems" : 2}
jsonschema.Draft3Validator(schema).is_valid([2, 3, 4])
Traceback (most recent call last):
File "", line 1, in
File "/home/dn/.local/lib/python2.6/site-packages/jsonschema.py", line 176, in init
resolver = RefResolver.from_schema(schema)
File "/home/dn/.local/lib/python2.6/site-packages/jsonschema.py", line 997, in from_schema
return cls(schema.get("id", ""), schema, _args, *_kwargs)
File "/home/dn/.local/lib/python2.6/site-packages/jsonschema.py", line 983, in init
self.store = _URIDict(store, **_meta_schemas())
TypeError: init() keywords must be strings

[dosena]

jsonschema.version
u'1.1.0'

[dosena]

sys.version
'2.6.4 (r264:75706, Dec 7 2009, 18:43:55) \n[GCC 4.4.1]'

[Julian]

If I recall correctly, this is a bug in early versions of 2.6 (specifically that they didn't support unicode kwargs in early versions). I test (and have installed) 2.6.7, but if I get my hands on an early version I can try to verify that. I'm not sure exactly what to do about this other than that, it's a "bug" more or less. I'll try to investigate a bit more.

[Julian]

Well, I guess on second thought this wouldn't be too painful to workaround I don't think.

[gazpachoking]

Yeah, I believe 2.6.5 fixed this issue, had similar reports on our project when I switched to unicode literals.

[Julian]

@dosena it would be awesome if you attempted a pull request for this since you have the setup to test it already. It should be fairly simple, just adding some 'b' prefixes to the keys in a single dict.

[gazpachoking]

@Julian Do we want the keys in the meta_schemas dict to be bytes though?

[Julian]

I don't know, do we not (I didn't look at what that entails even)?

I think what we really want is for _meta_schemas to go away, since we don't even need it anymore if I remember correctly.

[gazpachoking]

Well, I don't like the idea of storing any strings as bytes, but I do believe you are correct that _meta_schemas can go away. We can probably just transfer meta_schemas data to the RefResolver store in some other way that **kwargs to solve this problem.

[Julian]

Technically these are neither strings nor bytes I think, they're arbitrary
URIs, although yeah you're right it looks like, all the ones we are
prepopulating look like they should be strings.

I think we want to remove this by just having the resolver walk all of the
validators, which I think we store elsewhere.
On Apr 3, 2013 6:43 PM, "Chase Sterling" [email protected] wrote:

Well, I don't like the idea of storing any strings as bytes, but I do
believe you are correct that _meta_schemas can go away. We can probably
just transfer meta_schemas data to the RefResolver store in some other
way that **kwargs to solve this problem.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/87#issuecomment-15870158
.

[gazpachoking]

Do we need to walk anything? Couldn't we just do something like:

self.store = meta_schemas.copy()
self.store.update(store)

And rip out _meta_schemas

[Julian]

Sounds good, I wasn't looking at the code but that sounds vaguely familiar.
On Apr 3, 2013 8:16 PM, "Chase Sterling" [email protected] wrote:

Do we need to walk anything? Couldn't we just do something like:

self.store = meta_schemas.copy()self.store.update(store)

And rip out _meta_schemas

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/87#issuecomment-15873546
.

[gazpachoking]

Well, the copy won't actually work, forgot that meta_schemas maps uris to Validators, not schemas. We can still do something similar though.