ResponseNotRead when using raise_for_status hook

[nicklas-delphai]

In my setup I configure the AsyncClient to call response.raise_for_status() for every call using the raise_on_4xx_5xx response hook as stated in the documentation.

When catching HTTPStatusError as ex I was very confused to find that ex.response.json() would raise ResponseNotRead exception instead of giving me the expected json data.

After doing some digging in the source code I realized two things:

• The hooks are called right after the request was made here
• This is before response.read() is called for non-streaming calls

This will mean that if a user uses the raise_on_4xx_5xx as shown in the documentation they might unexpectedly run into the same issue as I did, where read() has not been called yet.

I believe this is very confusing. At least I can say it confused the hell out of me and cost me the better part of 2 hours now to figure out.

My solution to this problem was to extend the raise_on_4xx_5xx hook:

async def raise_on_4xx_5xx(response: httpx.Response):
    if response.status_code >= 400:
        await response.aread()
    response.raise_for_status()

Also it's worth to note that I didn't need this workaround previous to v0.19. Before v0.19 everything worked as I would expect and I could call ex.response.json() without encountering this exception.

Edit

I probably should have read the note in the documentation:

Response event hooks are called before determining if the response body should be read or not.
If you need access to the response body inside an event hook, you'll need to call response.read(), or for AsyncClients, response.aread().

That's on me. Anyway, I still find it pretty unpredictable.

[lovelydinosaur]

Thanks Nicklas, that's useful feedback - might be worth us looking at if it's possible to improve on the user experience here one way or another.