Allow client to use external Session object

[lavigne958]

Allow new gspread.client.Client to be initialized using an external requests.Session object.

closes #1383

[alifeee]

I am not really clear on how the client works.

Is this something that was possible in v5 but was made impossible in v6?

Either way, nice fix.

[alifeee]

This is good.

Is this something we would've caught if we'd tried to use gspread for some common actions?

[lavigne958]

I am not really clear on how the client works.

The client is the gspread entry point, it required some credentials and the Type of http client you want to use (standard/BackOff).

Is this something that was possible in v5 but was made impossible in v6?

Yes it is 😑😞 that's a regression out of the box right here.

In v5 it used to take the following arguments to instantiate a new client:

• credentials
• session object.

Now with the rework and new client it takes:

• credentials
• the Type of http client to use

So no more ways to pass you own session object 😬

Either way, nice fix.

Thanks ! 🙃

[lavigne958]

This is good.

Is this something we would've caught if we'd tried to use gspread for some common actions?

I don't think so, this can be caught only by having a test with a custom session object. Otherwise you don't see that you miss something in your interface.

[alifeee]

The client is the gspread entry point, it required some credentials and the Type of http client you want to use (standard/BackOff).

ok I think I'm getting confused between "client" (gspread client) and "client" (http client)

I don't think so, this can be caught only by having a test with a custom session object. Otherwise you don't see that you miss something in your interface.

Then we should also add a regression test...

[lavigne958]

The client is the gspread entry point, it required some credentials and the Type of http client you want to use (standard/BackOff).

ok I think I'm getting confused between "client" (gspread client) and "client" (http client)

yes sorry ! that's a very good point ! we should use client for the gspread client and http_client for the actual class that handles all HTTP request (the HTTPClient or BackOffHTTPClient)

I don't think so, this can be caught only by having a test with a custom session object. Otherwise you don't see that you miss something in your interface.

Then we should also add a regression test...

yes, I propose we add it later exceptionally if that's ok for you ? that would allow us to release first for the users

[lavigne958]

rebased on master branch before merge