Skip to contents

Contributing to spicyR

Source: CONTRIBUTING.md

spicyR’s cell-level test runs in a C++17 core (src/core, with Eigen) that is shared verbatim with the Python package spicyr. The files in src/*.cpp are the R bindings only, and the R code in R/cell.R and R/spicy_main.R prepares inputs and assembles results without doing any statistics.

  • The cell method’s loop over images lives in the core (src/core/src/stream.cpp, spicyglm/stream.hpp): the passes over the image and type codes, the gathering of a batch of whole images and their neighbour sums, each image in a Dataset of its own, the images shared among threads. src/glm_bindings.cpp only describes R’s vectors to it; reading the cells (data frames, Zarr stores via Rarr in R/store.R) stays in R.
  • Change the core here, then copy it to spicyr with its tools/sync_core.sh. spicyr’s tests check the copy.
  • tests/testthat/test-cell.R checks the core against a slow plain-R implementation written from the definitions (tests/testthat/helper-reference.R). Keep them in step.
  • spicyr’s tests/shared_cases/make_cases.R writes cases from this package that spicyr must reproduce; rerun it after any change to the core or to R/cell.R or R/spicy_main.R.
  • R CMD check and BiocCheck should stay clean.