xingwudao/open-xquant/create-portfolio-optimizer

Create PortfolioOptimizer

You create allocation logic that returns target weights.

Scope

Default built-in paths:

• source: src/oxq/portfolio/{snake_name}.py
• tests: tests/portfolio/test_{snake_name}.py
• package export: src/oxq/portfolio/__init__.py
• built-in registry: src/oxq/core/registry.py

Existing built-ins live in src/oxq/portfolio/optimizers.py; read that file before choosing whether to add a new module or extend the existing built-in module. Prefer a new module for a new component unless project maintainers ask otherwise.

Phase 1: Read Existing Patterns

Read before editing:

• src/oxq/core/types.py
• src/oxq/portfolio/optimizers.py
• one existing test in tests/portfolio/
• src/oxq/portfolio/__init__.py
• the portfolio registration block in src/oxq/core/registry.py

Phase 2: Define Behavior

State before coding:

• allocation formula
• constructor parameters
• whether it reads signals, indicators, or both
• required indicator columns
• fallback when no valid inputs exist
• whether weights can include CASH
• max/min weight constraints
• whether the optimizer is stateful. If it consumes categorical signals such as

BUY, SELL, and HOLD, define how HOLD preserves or resets prior target weights. SignalToPosition is the built-in reference pattern.

Ask the user if allocation logic is ambiguous.

Phase 3: Test First

Write tests with deterministic DataFrames:

• protocol compliance with PortfolioOptimizer
• empty input returns {"CASH": 1.0}
• weights sum to 1.0
• multi-symbol behavior
• hand-calculated allocation
• invalid or NaN input behavior
• name value

Run the new test and confirm the missing implementation fails before coding.

uv run pytest tests/portfolio/test_{snake_name}.py -v

Phase 4: Implement

Skeleton:

"""Short description portfolio optimizer."""
from __future__ import annotations
import pandas as pd
class ClassName:
    """Short allocation description."""
    name = "ClassName"
    def optimize(
        self,
        signals: dict[str, pd.DataFrame],
        indicators: dict[str, pd.DataFrame],
    ) -> dict[str, float]:
        """Return target weights that sum to 1.0."""
        ...

Rules:

• every path returns a non-empty dict
• invalid input returns {"CASH": 1.0}
• weights must sum to 1.0
• include CASH for unused capital
• do not mutate input DataFrames
• use only latest available row unless design says otherwise

Phase 5: Register

For a built-in component:

• import it in src/oxq/portfolio/__init__.py
• add it to __all__
• add it to _load_builtins() in src/oxq/core/registry.py

Verify:

uv run python - <<'PY'
import oxq
assert "ClassName" in oxq.list_portfolio_optimizers()
print("registered")
PY
uv run pytest tests/portfolio/test_{snake_name}.py -v

Run engine-level tests when optimizer behavior changes order generation.

Red Lines

• Do not return weights that sum above or below 1.0.
• Do not return an empty dict.
• Do not ignore invalid data silently; fall back to CASH or exclude the

symbol deliberately.

• Do not register before invariant tests pass.