Fixes #600

Problem

python-dotenv incorrectly parsed inline comments when a key had an empty value. Instead of treating the comment as a comment, it was included as part of the value.

For example:

EMPTY_VALUE= # comment

Expected: EMPTY_VALUE should be an empty string ('')
Actual: EMPTY_VALUE was set to '# comment'

This only happened when there was whitespace between = and #. The library correctly handled comments when values were non-empty (e.g., KEY=hello # comment worked fine), making this behavior inconsistent.

Root Cause

The parser had two issues:

1. The _equal_sign regex pattern (=[^\S\r\n]*) was consuming whitespace after the = sign, making it impossible to detect whether a # followed whitespace or came right after =
2. The parser didn't check if a # appeared after whitespace following the = sign (which should be treated as an inline comment, not a value)

Solution

I made two changes to src/dotenv/parser.py:

1. Fixed the regex pattern (line 24): Changed _equal_sign from r"(=[^\S\r\n]*)" to r"(=)" to stop consuming whitespace
2. Added inline comment detection (lines 158-165): After reading the = sign, check if there's whitespace followed by #. If so, treat it as an inline comment and set the value to empty string

Now the parser correctly handles these cases:

• a= # comment → '' (whitespace before # = inline comment)
• a=# comment → '# comment' (no whitespace = part of value)
• a=## comment → '## comment' (no whitespace = part of value)

Testing

Added comprehensive tests in tests/test_parser.py to verify:

• Empty value with inline comment (a= # comment)
• No whitespace before # (a=# comment)
• Tab character before # (a=\t# comment)
• Explicitly quoted empty value (a="")
• Multiple # characters in various positions

All existing tests pass (210/210).

Impact

This change makes python-dotenv's behavior consistent with other dotenv implementations (JavaScript, Bash) and fixes a parsing inconsistency within the library itself.