fix(env): preserve '=' in complex dict cast values (#565)

[jbbqqf]

Summary

The complex-dict cast path (cast={'value': ...}) used val.split('=') with no maxsplit, so a value that itself contains = (e.g. base64 padding, query strings, JWT segments) was silently truncated at the second =. The simple-dict cast path (cast=dict) already used split('=', 1). This PR aligns the two paths.

Fixes #565.

Context

Issue #565 noticed the inconsistency in environ/environ.py:

• Complex dict (line 554, before fix): val.split('=') — drops the rest of the value after the first internal =
• Simple dict (line 557): v.split('=', 1) — preserves it

Changes

• environ/environ.py — single-line change: val.split('=') → val.split('=', 1) in the complex-dict branch of parse_value. Added a code comment pointing at the simple-dict line and at issue Complex dict value cannot have equal sign, but simple dict can. #565 so a reviewer doesn't have to re-derive the reasoning.
• tests/test_env.py — extends the existing parametrised test_dict_parsing matrix with a dict_value_contains_equal_sign case ('a=1;b=v=more' with dict(value=str)).

Reproduce BEFORE/AFTER yourself (copy-paste)

# --- one-time setup ---
git clone https://github.com/joke2k/django-environ.git /tmp/dj-environ-565 \
  && cd /tmp/dj-environ-565
python3 -m venv .venv && . .venv/bin/activate
pip install -e ".[testing]" >/dev/null

# --- BEFORE: origin/develop, expect the bug ---
git checkout origin/develop -- environ/ tests/
python3 -c "
import environ
e = environ.Env()
print(e.parse_value('a=1;b=v=more', {'value': str}))
"
# Expected: {'a': '1', 'b': 'v'}            <-- truncated, BUG

# --- AFTER: this branch, expect the fix ---
git fetch https://github.com/jbbqqf/django-environ.git fix/565-complex-dict-equal-sign
git checkout FETCH_HEAD -- environ/ tests/
python3 -c "
import environ
e = environ.Env()
print(e.parse_value('a=1;b=v=more', {'value': str}))
"
# Expected: {'a': '1', 'b': 'v=more'}       <-- preserved, FIXED

# Regression test:
pytest tests/test_env.py -k dict_value_contains_equal_sign -v
# Expected: PASSED in TestEnv / TestFileEnv / TestSubClass (3 cases)

What I ran locally

$ pytest tests/test_env.py -k test_dict_parsing -v
... 24 passed, 267 deselected in 0.08s

$ pytest
... 458 passed in 0.26s

The new test was also verified to FAIL on develop (only the environ.py change reverted): assert {'a': '1', 'b': 'v'} == {'a': '1', 'b': 'v=more'}.

Edge cases

#	Scenario	Input	Expected	Verified by
1	Existing simple key=value	'a=1' with dict(value=int)	{'a': 1}	dict_int (unchanged)
2	Existing list value	'a=1,2,3' with dict(value=[int])	{'a': [1, 2, 3]}	dict_int_list (unchanged)
3	Value contains a single =	'a=1;b=v=more' with dict(value=str)	{'a': '1', 'b': 'v=more'}	new dict_value_contains_equal_sign
4	Multiple = in value	'k=a=b=c' with dict(value=str)	{'k': 'a=b=c'}	covered by same path (split with maxsplit=1)

Risk / blast radius

The change touches a single line in Env.parse_value's complex-dict branch. The previous behaviour silently truncated values at the first internal =, so any user actively relying on that truncation would have been doing so unintentionally — they would have been getting partial values without warning. All existing test_dict_parsing cases pass unchanged (none of them placed = inside a value), and the broader test suite (458 tests) is green.

---

PR drafted with assistance from Claude Code (Anthropic). The change was reviewed manually against django-environ's source. The reproducer block above is the one I used during development; reviewers can paste it verbatim.