Default timeout of 30s documented but never applied — requests hang indefinitely

[thomasht86]

Summary

Client.__init__ and the README both document a default request timeout of 30 seconds, but no timeout is ever applied. A default httpr.Client() will block indefinitely on a server that accepts the connection and never responds.

Cause

httpr/__init__.py:203 declares timeout: float | None = 30, but the body of Client.__init__ is just super().__init__() — it forwards nothing. The real constructor is PyO3's #[new] on RClient (src/lib.rs:118), whose signature defaults to timeout=None. With None, src/lib.rs:183 never calls client_builder.timeout(...), so reqwest applies no timeout at all.

Reproduction

import httpr
print(httpr.Client().timeout)   # -> None, not 30.0

Against a local server that accepts the request and then sleeps for 120s:

httpr.Client()  [docs say default timeout=30]  raised RequestError after 120.0s
httpx.Client()  [default timeout=5]            raised ReadTimeout   after 5.1s

httpr did not time out — it returned only when the server finally closed the socket at 120s.

Expected

httpr.Client().timeout == 30.0, and a request against an unresponsive server raises a timeout exception after ~30s.

Proposed fix

Change the default in the Rust #[new] signature (src/lib.rs:118) from timeout=None to timeout=30.0, so the documented default is the one that actually applies.

I audited all 16 constructor parameters — timeout is the only one where the Python-documented default disagrees with the Rust default. cookie_store, referer, follow_redirects, max_redirects, verify, https_only and http2_only all agree.

While here, it's worth making Client.__init__'s signature non-misleading (it currently lists every parameter and forwards none of them; the values reach Rust via __new__).

Suggested tests

• assert httpr.Client().timeout == 30.0
• assert httpr.Client(timeout=5).timeout == 5.0
• A test against a deliberately slow endpoint asserting a timeout is raised.

Size

~30 minutes.