Skip to content

Latest commit

 

History

History
79 lines (51 loc) · 3.44 KB

DEVELOPING.md

File metadata and controls

79 lines (51 loc) · 3.44 KB
Raw

Developing in COMB

Guidelines

Formatting

styler is an R code formatter. For example, to style everything in a sub-directory:

style_dir(here("spatial/src/"))

For more information instructions follow the link.

Before making a pull request containing .R source files, it is strongly recommended to style all that have changed under styler's default settings, as this can meaningfully reduce the rate of merge conflicts due to formatting. (Differences due to editor handling of tab expansion and trailing white space are the most common).

Development Tools

renv

First-time setup

renv is a dependency management system for R. Using it is optional but highly recommended as a way of keeping all contributors' package versions in sync. There is not much to do to use it. When you run R from the project directory, you will notice a prompt suggesting

renv::restore()

to sync with the lockfile. Running this command will install the lockfile-specified package versions into renv/library. This can take about half an hour the first time, but after that R will start up as fast as before, and syncs as new packages are added will be quicker.

Adding packages

If you add a library() line for a new package during development and find that you needed to install a package, please also run renv::snapshot() and include the changes to the lockfile in the commit that adds the library() statement.

Removing redundant calls to packages

Sometimes when rewriting scripts don't call all the packages that are loaded. To figure out what packages or functions you need in a script so you can eliminate any unnecessary library calls use NCmisc. First install the package:

  1. install.packages("NCmisc")
  2. library(NCmisc)

Then run following command, for example on one particular script eda_forest_data_points:

  1. NCmisc::list.functions.in.file(here("spatial/src/eda_forest_data_points.R"))

And compare the output to the list of library calls in the script or other code. This is not strictly necessary but helps when breaking scripts down into sub-parts to optimize our code.

Jupyter Notebook

(Optional) If you would like to be able to run R .ipynb files from other contributors or to author your own, you can run a local Jupyter notebook. The IRkernel package is already installed in the project R environment by renv. For the rest:

  1. Install Jupyter.
  2. Run the R script infrastructure/install_jupyter_r_kernel.R from the project R environment.
  3. Run jupyter notebook or jupyter lab according to your preference and which are installed.
  4. In your browser, find or create the .ipynb you are interested in.

Other issues

Data are stored in Google Drive and synchronized to your local computer using code (see our drive_sync() function).

All 'raw' data would be synchronized (copied from Google Drive) into a relevant local /input/ directory. Outputs are also synchronized (copied to) the relevant /output directory and mirrored on Google Drive. Subsequent downstream steps symlink from previous /output directories to /input directories and in general /output(s) that are synchronized to Google Drive are not read back in except for extremely computationally expensive steps we won't want to regenerate each time we run the code.