All development should be on a feature or development branch. Do not merge back into master until devtools::check() is passing for R code and XXX is passing for python code.
To install from local source:
conda create -n countland -c conda-forge python==3.10
conda activate countland
cd countland-py
pip install .[dev]
pip install jupyter
This will reload countland functions (e.g. after an edit)
import importlib
importlib.reload(countland)
To run tests for countland-py/:
conda activate countland
python -m pytest
Our automated code review draws on this strategy.
To manually run it, cd to countland\countland-py\countland and run:
black .
flake8 .
This package is built with the excellent devtools. Extensive explanations on using devtools to create R packages is available in the book R Packages.
Development typically involves cding to the package directory countland-R/, launching R, and running some combination of the following:
options(error=traceback) # Get line numbers for errors
library(devtools)
devtools::load_all()
devtools::test()
devtools::document()
devtools::check() # A wrapper for R CMD check, see http://r-pkgs.had.co.nz/check.html#check
devtools::build() # Create package bundle, including executed vignettes
To regenerate the pdf manual, run the following shell command in the package directory:
R CMD Rd2pdf . --force --output=countland-manual.pdf
Construction of vignettes follows https://r-pkgs.org/vignettes.html .
This works nicely for developing countland R code.