fix: override_if_not_empty should not skip falsy primitives (0, False, 0.0)

[gaoflow]

Problem

TypeConflictStrategies.strategy_override_if_not_empty uses a bare truthiness check:

return nxt if nxt else base

This treats any falsy value as "empty", including 0, False, and 0.0.
These are valid, intentional values — they are not empty, and they are not
null — so they should override base.

from deepmerge import Merger
m = Merger([], "override", "override_if_not_empty")

m.merge("base", 0)      # returns "base"  ← wrong, should be 0
m.merge("base", False)  # returns "base"  ← wrong, should be False
m.merge("base", 0.0)    # returns "base"  ← wrong, should be 0.0

Fix

Replace the truthiness check with an explicit null/empty test:

1. None → keep base (null)
2. len(nxt) == 0 (dict, list, set, str, …) → keep base (empty container)
3. Anything else (including 0, False, 0.0) → return nxt

The empty check is isinstance(nxt, Sized) and len(nxt) == 0, so non-sized
types (ints, floats, bools) fall through to the override path.

The existing tests pass unchanged; a parameterized test covers the falsy
primitives and the empty/null cases.

[toumorokoshi]

Thanks! this looks right, but a quick comment.

Also although I agree this is a "fix", I think we'll do a major version bump - it's a breaking change to a merge strategy.

[gaoflow]

Parameterized in e16063a — 13 cases with ids, covering the falsy primitives plus the empty/null side so both directions are in one table. Your point about the early exit is exactly what shows up: against master the run reports all four broken cases instead of stopping at the first one.

FAILED test_merge_if_not_empty_falsy[zero int]
FAILED test_merge_if_not_empty_falsy[false]
FAILED test_merge_if_not_empty_falsy[zero float]
FAILED test_merge_if_not_empty_falsy[zero complex]
4 failed, 9 passed

It asserts the type as well as the value, since 0 == False would otherwise let a wrong return slip through. Whole suite is 34 passed, black --check and mypy clean.

Major version bump is your call, that makes sense to me. Also corrected the PR description: it still claimed a try/except TypeError, which 07e8b4f replaced with the Sized check.

[toumorokoshi]

great, thank you!