---
title: "Contributing to ‘box’"
author: Konrad Rudolph
date: "`r Sys.Date()`"
output:
    rmarkdown::html_vignette:
        toc: true
    md_document:
        variant: gfm
vignette: >
    %\VignetteEngine{knitr::rmarkdown}
    %\VignetteIndexEntry{Contributing to ‘box’}
    %\VignetteEncoding{UTF-8}
---
# Installation

Compared to many other R packages, ‘box’ uses a somewhat different
development workflow:

While ‘box’ uses ‘devtools’ and ‘roxygen2’ to generate documentation and
package infrastructure, *generated code and data is not checked into
version control!* This means in particular that the project does not
contain a `NAMESPACE` file, since that is auto-(re)generated by tools.
Therefore, attempting to install the current development version of
‘box’ via `pak::pak('klmr/box')` or equivalent means *will fail!*

Instead, an up-to-date, automatically generated build of the development
version of ‘box’ should be installed from
[R-Universe](https://klmr.r-universe.dev/):

```{r eval = FALSE}
install.packages('box', repos = 'https://klmr.r-universe.dev')
```

… or from the corresponding `build` branch:

```{r eval = FALSE}
pak::pak('klmr/box@build')
```

Alternatively it can be built manually using the instructions below.

# Development

## Building

The project contains a `Makefile` written in the [GNU
Make](https://www.gnu.org/software/make/manual/make.html) dialect that
contains various development utilities. Invoking `make` without target
will show a list of available targets with short descriptions. Using
this `Makefile` isn’t *necessary*, but it helps. In particular:

- `make documentation` builds the `NAMESPACE` file, the shared C library
  and the package documentation.
- `make test` runs the unit tests.
- `make check` runs checks, and should run cleanly before submitting a
  pull request: of note, `make check` performs additional checks that
  are not performed by either `R CMD check` or `rcmdcheck::rcmdcheck()`
  (these checks can be found as individual scripts under `scripts/`;
  they roughly correspond to some undocumented checks performed
  internally on CRAN).

## Branches

All new code should be developed on a new branch with a name prefixed
`fix/` (for bug fixes), `feature/` (for new features/enhancements), and
`chore/` (for any other contributions: fixed typos, project
infrastructure, tests, etc.). Branches for pull requests will be merged
into the [`main` branch](https://github.com/klmr/box/tree/main).

## Attribution and copyright

All contributions to the project are licensed under the same licence as
the project itself ([MIT licence](https://opensource.org/license/mit)).
By contributing, you agree to put your contribution under the project’s
licence. By submitting a pull request, you are asserting that your
contribution is your original work and that you are releasing its
copyright to the project under the project’s licence.

In particular, LLM-generated content without significant creative human
input is not copyrightable under many legislations (see e.g. the
[briefing of the EU parliament research
service](https://www.europarl.europa.eu/thinktank/en/document/EPRS_BRI(2025)782585)).
Therefore, contributions that are generated by an LLM without
significant creative human input cannot be accepted. At the same time,
we recognise that the use of code-generation tools has a long-standing
tradition, and that merely using a tool to help generate text does not
necessarily negate the creative contribution of a human author. We
therefore stop short of outright banning the use of LLM-aided
contributions for the time being.

## Code style

The code style of ‘box’ is similar to the [Tidyverse style
guide](https://style.tidyverse.org/), but with several notable
differences:

- The file extension for R code files is lowercase `.r` (not uppercase
  `.R`). The file extension for R Markdown files is lowercase `.rmd`
  (not uppercase `.Rmd`).

  <div class="rationale">

  **Rationale:** using capital letters in file extensions is a pointless
  violation of established convention that creates unnecessary
  inconsistencies.

  </div>

- Use `=` for assignment, not `<-`.

  - On the (very rare!) occasions where assignment inside a function
    call is required, use additional parentheses to make `=`
    syntactically an assignment rather than named argument passing.

    ``` r
    # GOOD:
    x = 5

    # OK (rarely):
    if (is.null((x = function_call()))) {}

    # BAD:
    x <- 5
    if (is.null(x <- function_call())) {}
    ```

  - `<<-` is banned. Instead of `name <<- value`, use
    `env$name = value`, where `env` is a previously-defined name
    referring to the desired target environment. This pattern is used in
    various places in the code, and good, context-dependent names for
    `env` are `ns`, `self`, or similar. `assign()` should generally not
    be used when the name of the assignee is statically known.

    ``` r
    # GOOD:
    self = parent.frame()
    # … later, inside a nested closure:
    self$x = TRUE

    # BAD:
    x <<- TRUE

    # BAD:
    assign('x', TRUE, envir = caller)
    ```

    <div class="rationale">

    **Rationale:** `<<-` leaves it unclear where assignment will happen.
    Subset-assignment via `$` makes the assignment target scope
    explicit, which makes the code clearer and less error-prone.

    </div>

- Leave a space between `function` and the following opening
  parenthesis.

  ``` r
  # GOOD:
  f = function () {}

  # BAD:
  f = function() {}
  ```

  <div class="rationale">

  **Rationale:** a function declaration in R does not use function call
  syntax; rather, it’s a special syntactic construct equivalent to `if`
  and `while`, so the same formatting conventions apply.

  </div>

- Use single quotes, not double quotes, around strings. — Even when the
  string contains `'`, which should be escaped.

  ``` r
  # GOOD:
  'text'
  'text with \'quotes\''

  # BAD:
  "text"
  "text with 'quotes'"
  r'(text with 'quotes')'
  ```

  <div class="rationale">

  **Rationale:** raw strings cannot be used since ‘box’ supports R
  versions pre R 4.0.

  </div>

- Use four spaces for indentation. Do *not* add extra spaces to align
  assignments or named function call arguments.

  ``` r
  # GOOD:
  first = 1
  second = 2

  # BAD:
  first  = 1
  second = 2
  ```

- The line length is hard-limited to 120 columns. *Most* lines should be
  shorter, but there is no (soft, or otherwise) limit at 80 columns.

- Explicitly use integer literals where the logical type of the
  expression is integer.

  ``` r
  # GOOD:
  if (length(x) == 0L) {}

  # BAD:
  if (length(x) == 0) {}
  ```