Skip to content

AGENTS.md

Latest commit

 

History

History
112 lines (96 loc) · 3.75 KB

AGENTS.md

File metadata and controls

112 lines (96 loc) · 3.75 KB
Raw

Greybox

Project Summary

This is a Python port of the R package greybox - a toolbox for model building and forecasting. The original R package implements functions and instruments for regression model building and its application to forecasting.

Main Features

  • Regression Models: Advanced Linear Model (ALM), Scale Model (SM), Dynamic Linear Regression
  • Model Selection: Stepwise regression based on information criteria, partial correlations. Combined ALM (CALM)
  • Variable Processing:
    • Lag/lead expansion (xregExpander)
    • Mathematical transformations (log, sqrt, etc.)
    • Cross-product generation
    • Temporal dummies, outlier dummies
  • Forecasting: Rolling origin evaluation, error measures
  • Statistical Measures: AIC, BIC, determination coefficients, association measures
  • Bootstrap Methods: Parameter bootstrap, bootstrap for time series
  • Distributions: Normal, Laplace, LogitNormal, Beta-Normal, etc.

Target Users

  • Data scientists and forecasters
  • Marketing analysts
  • Econometric modelers
  • Anyone building predictive models with time series data

Build, Lint, and Test Commands

These commands ensure your code adheres to standards and functions as expected.

IMPORTANT: Always use the python/.venv environment when running Python commands! Example:

cd python && .venv/bin/python -m pytest tests/

Build Command:

make build

This command compiles the project. Ensure you're in the root directory of the project before running it. Lint Command:

flake8 .

The flake8 tool checks your Python code for issues, including errors and deviations from PEP 8 style guidelines. It's a combination of PyFlakes, pycodestyle (formerly pep8), and McCabe script. Test Commands:

  • Run all tests: pytest
    • Run tests in watch mode: ptw This command runs pytest in a loop, re-running the tests whenever one of your files changes.
  • Run a specific test file: pytest [path_to_test_file]
    • Run a specific test within a test file: pytest [path_to_test_file]::[TestClassOrFunctionName]

Code Style Guidelines

Imports

  • Place standard library imports first, then third-party libraries, and lastly the local application/library-specific imports. Each section should be separated by an empty line.

Example:

import os
import sys
import requests
from flask import Flask
from myapp.forms import MyForm

Formatting

  • Use 4 spaces per indentation level.
  • Ensure lines do not exceed 80 characters to prevent horizontal scrolling.
  • Separate top-level function and class definitions with two blank lines, and method definitions inside a class body with one blank line.
  • Surround binary operators (+, -, *, / etc) with a single space on either side for readability. Types Enforce type hints in your codebase to improve readability and maintainability. This helps tools like mypy (used for static type checking). Example:
def multiply(a: int, b: int) -> int:
    return a * b

##Naming Conventions

  • Variables: Use lowercase with words separated by underscores. max_retries = 3
    • Constants: Capitalize each word and separate using underscores. MAX_CONNECTIONS = 100
    • Functions & Methods: Lowercase with words separated by underscores.
def send_email(to_address: str, body: str) -> None:
  pass
  • Classes: Capitalize the first letter of each word (CamelCase).
class UserAuthenticator:
  pass

Error Handling

Handle exceptions gracefully. Use specific exceptions rather than catching a broad Exception which can hide many errors.

Example:

try:
    response = requests.get('https://api.example.com/data')
    response.raise_for_status()
except requests.exceptions.HTTPError as e:
    print(f"HTTP error occurred: {e}")