Do not add a plus sign to a negative zero float result

[andrewstellman]

Summary

Arithmetic on a float written with an explicit sign renders +-0.0 when the result is negative zero, and tomlkit cannot parse that output.

import tomlkit

doc = tomlkit.parse("x = -1.5")
doc["x"] *= 0
print(doc.as_string())          # actual: x = +-0.0    expected: x = -0.0
tomlkit.parse(doc.as_string())  # InvalidNumberError: Invalid number at line 1 col 9

Float._new (tomlkit/items.py:879 on master) adds + when self._sign and result >= 0. -0.0 >= 0 is True, and str(-0.0) is already "-0.0". This is the line #345 changed for #341, and negative zero is the case that change left open. A signed integer reaches the same line through true division (n = +0, then doc["n"] /= -1).

The fix adds the + only when str(result) does not already start with -. Negative zero is the only value that is >= 0 and prints with a leading -, so no other result renders differently.

Agent Drafting Metadata

• Agent: Quality Playbook, run with Claude Code
• Model: Claude Fable 5.1, Claude Opus 5.5 and Claude Sonnet 5.5
• Notes: Found by Quality Playbook, an AI code-review tool, with Claude; I reviewed the change. The agent also drafted the test, the fix and this text.