Maybe we should always offer a "team_id" parameter?

[ricpol]

Several (Pygrister) apis don't expose a "team_id" parameter to allow changing the org/team id at runtime. Internally, they simply rely on "self.configurator.server" to construct the required url.
In fact, the equivalent Grist apis tend to be rather lenient: https://myteam.getgrist.com/api/profile/apikey will work, but https://bogus.getgrist.com/api/profile/apikey will work too :-/
However, Grist may become stricter in future and, even now, calling inspect() and finding out that we called a different url than expected may confuse the user.
For consistency, we should always expose a "team_id" argument, even if it is not necessary. Each function that now works like this:

def foo(self, ...):
    url = f'{self.configurator.server}/...'

should work like

def foo(self, ..., team_id=''):
    url = f'{self.configurator.make_server(team_id)}/...'

instead.

OR, all of this is actually wrong and we shouldn't offer a parameter that is not actually needed?

(besides, the opposite would be tricky too... removing the team_id parameter every time it's not actually needed in the underlying Grist api... ugh)

[raph-topo]

I do not use

    grist.foo(
        ...
        doc_id="...",
        team_id="...",
    )

at all anymore, and wouldn't use team_id='...' either.

Rather, I use one instance of pygrister.api.GristApi for each single doc, configured with a dynamic config dict.

    @property
    def grist(self) -> GristApi:
        self._grist.apicaller.request_options["timeout"] = env.API_TIMEOUT
        self._grist.reconfig(self.grist_config())
        if not self._grist.apicaller.session:
            self._grist.open_session()
        return self._grist

Much simpler & reusable in multi-action scripts. And connections to entirely different Grist docs (even different servers!) can coexist in the same code, without reconfigs all over.

[ricpol]

yeah, between the two of us, this is the smart thing to do - and offering a doc_id/team_id parameter every single time was not a smart design decision, perhaps. But it seemed to make sense at the time, and now it's too late and I need to stay consistent with my own api. (besides, Gry heavily relies on "-d" and "-t" options everywhere).