Fix serializer negative infinity

[aryxnn]

What does this PR do?

PR title must follow Conventional Commits, for example feat: add dataset scoring helper or fix(openai): preserve trace context.

Fixes #

Type of change

• Bug fix
• New feature
• Breaking change
• Refactor
• Documentation update
• Tooling, CI, or repo maintenance

Verification

List the main commands you ran:

poetry run pytest tests/unit/test_serializer.py

Checklist

• I self-reviewed the diff using code_review.md.
• I added or updated tests for behavior changes.
• I updated docs, examples, or .env.template if needed.
• I did not hand-edit generated files; if generated files changed, I used the upstream regeneration path.
• I did not commit secrets or credentials.

Greptile Summary

This PR fixes two bugs: negative infinity floats were always serialised as "Infinity" instead of "-Infinity", and propagate_attributes rejected non-string metadata values instead of coercing them.

• serializer.py: one-line fix to return "-Infinity" when math.isinf(obj) is true and obj < 0.
• propagation.py: widens the metadata parameter type to Dict[str, Any] and adds str(value) coercion for non-string values; None values are now silently skipped rather than propagated.
• Tests: new test_float_special_values covers all three float special values; a new propagation test verifies int/float/bool coercion; test-data fixture classes renamed from Test* to Mock* to avoid unintentional pytest collection.

Confidence Score: 4/5

Safe to merge — the serializer fix is a minimal one-liner with a clear test, and the propagation coercion change is well-covered by the new test.

Both fixes are small and targeted. The only gap is the docstring for propagate_attributes which still tells callers that non-string metadata values are dropped, when they are now silently coerced via str(). A caller relying on the old drop-with-warning behaviour could be surprised.

langfuse/_client/propagation.py — the docstring for the metadata parameter needs to reflect the new coercion behaviour.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[EventSerializer.default called with float] --> B{math.isnan?}
    B -- yes --> C[return 'NaN']
    B -- no --> D{math.isinf?}
    D -- no --> E[normal float path]
    D -- yes --> F{obj < 0?}
    F -- yes --> G[return '-Infinity']
    F -- no --> H[return 'Infinity']

    I[propagate_attributes metadata dict] --> J{value is None?}
    J -- yes --> K[skip / continue]
    J -- no --> L{isinstance str?}
    L -- yes --> M[use as-is]
    L -- no --> N[str coerce via str value]
    M --> O[_validate_string_value]
    N --> O
    O -- valid --> P[add to validated_metadata]
    O -- invalid --> Q[drop with warning]

Loading

Prompt To Fix All With AI

Fix the following 2 code review issues. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 2
langfuse/_client/propagation.py:126-130
The docstring still says non-string metadata values "will be dropped with warning" (`AVOID: ... non-string values (will be dropped with warning)`), but the new code now coerces them to strings via `str(value)`. A caller who reads the docstring before passing `int`/`float`/`bool` values will incorrectly expect them to be silently dropped.

```suggestion
        metadata: Additional key-value metadata to propagate to all spans.
            - Keys must be US-ASCII strings
            - String values must be ≤200 characters; non-string values are coerced via ``str()``
            - ``None`` values are silently skipped
            - Use for dimensions like internal correlating identifiers
            - AVOID: large payloads, sensitive data
```

### Issue 2 of 2
tests/unit/test_serializer.py:307
The file is missing a trailing newline. Most linters and editors (and git) expect text files to end with a newline character.

```suggestion
    assert serializer.encode(float("nan")) == '"NaN"'
```

_{Reviews (1): Last reviewed commit: "fix: serialize negative infinity as -Inf..." | Re-trigger Greptile}

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

Aryan Srivastava seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}