fix: import urllib.parse explicitly instead of relying on side effects

[skippdot]

Fixes #2803

Four modules import the bare urllib package but call urllib.parse.*: _helpers.py (4 call sites), discovery.py (6), http.py (6), model.py (1). import urllib does not import submodules — urllib.parse only becomes an attribute of the package after some module executes import urllib.parse. Today that happens as a side effect of httplib2 being imported transitively, which is why the bug is invisible in most runs.

_helpers.py has no third-party imports, so it crashes outright:

import googleapiclient._helpers as h
h._add_query_parameter("http://example.com", "a", "b")
# AttributeError: module 'urllib' has no attribute 'parse'

The same mechanism is why pytest tests/test__helpers.py in isolation fails 7 of 10 tests on main while the full-suite run passes.

Change

Replace import urllib with import urllib.parse in the four modules. Call sites are untouched — urllib.parse.urlparse(...) keeps its current spelling — so the diff is one line per module, +4/−4. The other three modules are included because they carry the same undeclared dependency on httplib2's import side effects, just without a currently-reachable crash.

Verification (Python 3.10 and 3.14)

• The two-line reproducer above: crashed before, returns http://example.com?a=b after.
• Fresh-subprocess import-and-use probe of each of the four modules: all OK.
• pytest tests/test__helpers.py in isolation: 7 failed / 3 passed → 10 passed.
• Full tests/ folder: 338 passed, 1 skipped on both Python versions.
• black --check clean; flake8 --select=E9,F63,F7,F82 (the set CI enforces) clean.

[gemini-code-assist]

Code Review

This pull request updates several files within the googleapiclient library to import urllib.parse directly instead of the top-level urllib module. There are no review comments, and I have no feedback to provide.

[daniel-sanche]

LGTM