Site creation fails when using Japanese characters for site name

[benlower]

Trying to create a site with the following:

    sites = "遊び場"
    new_site = TSC.SiteItem(name=site, content_url=site.replace(" ", ""), admin_mode=TSC.SiteItem.AdminMode.ContentAndUsers)
    server.sites.create(new_site)

This errors with:

409002: Resource Conflict
A site already exists with the content URL '遊び場'

The site doesn't exist on my server.

[graysonarts]

@benlower Did you try this using postman directly? That error is coming directly from the REST API, so I don't think this is a bug in the server client but in the REST api. Can you verify via postman against the rest api directly, and update this issue once you have?

[t8y8]

@benlower specify a different content_url, that content_url is becoming "" (because you can't have Japanese characters as the content_url) and "" is the ID of the default site.

There might be a bug in Server, where it's telling you the content_url that you provided and not the actual sanitized content_url that it tried (We fixed a bug in tabcmd related to this on my team recently)

[t8y8]

Idea: we could validate that content_url only contains valid ascii characters and throws before making the call to Server

    @content_url.setter
    @property_not_nullable
    def content_url(self, value):
        if not is_ascii(value): raise ValueError
        self._content_url = value

And/or make it a new decorator. Py2/3 might make this a little bit of code, but still probably just a one line change and add a helper function

[graysonarts]

I think making it a decorator would be better. Something like "OnlyAscii" or something.

[t8y8]

@RussTheAerialist, I think according to SO this is the best py2/py3 way to do it:

def property_is_url_safe(func):
    @wraps(func)
    def wrapper(self, value):
        try:
            value.decode('ascii')
        except UnicodeDecodeError:
            raise ValueError("content_url must be ascii only and url safe")
        return func(self, value)

    return wrapper

[LGraber]

Why do you call is "is_url_safe"? That doesn't really describe what it is doing (I still have to go to the code to figure out that safe == ascii). Are you planning on adding other things here? It is also an odd property name given that it will be on a property which is not an url. Besides that I am fine with it after talking with Russell

[graysonarts]

Maybe have it be a property that takes an argument and call it something like "allowed_characters"? And then define "allowed_content_url_characters" somewhere which includes a string (or list/tuple) of all of the allowed characters in a content url?

Or maybe a regex validation method?

[t8y8]

Resolved in #64