Installation¶
Requirements¶
- Python 3.10+
- numpy, pandas, scipy
Install from Git¶
# Basic
pip install git+https://github.com/marketneutral/alpha-parser.git
# With all optional dependencies
pip install "qex[all] @ git+https://github.com/marketneutral/alpha-parser.git"
Install for Development¶
Project layout¶
src/qex/ the language
parser.py reads an expression into a tree of Signal nodes
signal.py the Signal base class: evaluate, to_weights
context.py compute_context: the shared cache
data.py fields, close() and friends, market(), LazyData
primitives.py returns, volatility, volume, adv; price_returns, forward_returns
operators.py arithmetic, comparisons and logic, math, is_valid, is_finite, None
conditional.py where, coalesce
timeseries.py time-series operators, events, fill_forward, latch
crosssection.py cross-sectional operators, regression_neut, pca_factor
groups.py group operators and labels
universe.py the cross-sectional universe
kinds.py numbers versus labels, checked when an expression is read
missing.py explain_missing
adjust.py the look-ahead check for adjusted data
distributed.py SignalGraph and Dask
risk.py the factor risk model
src/evaluation/ books and backtests
book.py BookSettings, build_book, hold_untradable
backtest.py Backtest, BacktestResult
quantile.py QuantileAnalysis
metrics.py Sharpe, drawdowns, turnover, fitness, margin
tests/ pytest; test_docs.py checks the documentation's examples
docs/ this site (mkdocs)
examples/ cookbook.py, risk_model_example.py
data/fetch_fmp.py the FMP data fetcher
Optional Dependencies¶
| Group | Packages | Use Case |
|---|---|---|
dev |
pytest, pyarrow | Running tests |
data |
requests, python-dotenv, pyarrow | FMP data fetcher |
distributed |
dask, distributed, graphviz | Dask evaluation, graph pictures |
risk |
statsmodels | Factor risk model |
docs |
mkdocs, mkdocs-material, mkdocstrings | Building this site |
all |
Everything above | Full installation |
Install specific groups (from cloned repo):
pip install -e ".[dev]" # Just testing
pip install -e ".[risk]" # With risk model
pip install -e ".[all]" # Everything