rmarcink
59ccbb40ca
Replace eager wildcard-imports with on-demand attribute resolution. Cold import drops from ~930 ms to ~280 ms because no indicator function runs at import time. Public API is preserved. * `_lazy_subpackage.install_lazy_subpackage` installs `__getattr__` / `__dir__` on every category package so `from <cat> import <name>` imports `<cat>/<name>.py` only on first access. Supports aliases (e.g. `math.max` -> `math.rolling_max`) and special targets (e.g. `candles.cdl` lives inside `candles/cdl_pattern.py`). * `_indicator_loader._make_ta_wrapper` builds a per-indicator `AnalysisIndicators` method from the underlying function signature, invoked lazily by the new `AnalysisIndicators.__getattr__`. The old hand-written wrapper class (2400+ lines) is gone. * Top-level `pandas_ta_classic.__getattr__` resolves bare names like `ta.rsi` and individual candle submodules (`ta.cdl_2crows`). It also exposes `ALL_PATTERNS` and emits a `DeprecationWarning` for the legacy `CDL_PATTERN_NAMES` alias. * `core.pyi` + `tools/gen_core_stub.py` give IDEs and mypy full signatures for every accessor method; the stub is generated from Category at build time. * All ten category `__init__.py` files collapse to a single `install_lazy_subpackage(__name__)` call. * `math` joins `_VALID_CATEGORIES` so the loader discovers its operators alongside other indicators. * `tests/test_indicator_candle.py` switches to the new public `ALL_PATTERNS` name.
4 lines
106 B
Python
4 lines
106 B
Python
from pandas_ta_classic._lazy_subpackage import install_lazy_subpackage
|
|
|
|
install_lazy_subpackage(__name__)
|