[Feature Request] Automatic installation of packages detected via explicit package::function() calls in quarto and R projects

[fretwurst]

Problem

In CI/CD-Workflows, particularly for Quarto-based projects, missing package installations often cause rendering processes to fail. This is a common issue when:

• Packages are used explicitly via package::function() in .qmd files and are not pre-installed in the CI/CD environment.
• CI/CD pipelines (e.g., with Rocker Docker containers) halt after encountering the first missing package, leading to time-consuming debugging in large projects.

For Quarto projects, active chapters defined in _quarto.yml (under chapters or render) often determine the relevant .qmd files to render. Detecting and pre-installing the packages used in these files before rendering could significantly streamline the CI/CD workflow.

---

Proposed Solution

Enhance pak with a feature to:

1. Scan project files (e.g., .qmd, .Rmd, .R) for all explicitly used packages:
   • Detect package::function() calls.
   • Optionally scan for pak::pkg_install() or pak::pkg() calls in file headers.
2. Support Quarto project workflows:
   • Read _quarto.yml to identify active .qmd files (chapters or render keys).
   • Install all required packages before rendering begins.
3. Install missing packages efficiently:
   • Use pak's parallelized installation and caching to minimize installation time in CI/CD pipelines.
   • Avoid breaking on the first missing package.

---

Best Practice Alignment

Modern R style guides, such as the Google R Style Guide and the RStudio Tidyverse Style Guide, recommend using explicit package::function() calls over loading packages globally. This approach improves:

• Clarity: The source of each function is immediately clear.
• Conflict avoidance: Prevents naming conflicts between functions in different packages.
• Modularity: Ensures code runs independently of preloaded packages.

Given this trend, tools like pak should support workflows where packages are explicitly referenced, especially in CI/CD contexts where no preloaded environment exists.

#### Example Workflow
A new `pak` function, such as `pak::install_quarto_deps()`, could streamline this process:

```r
# Automatically scan a Quarto project and install dependencies
pak::install_quarto_deps(yml = "_quarto.yml")

This function would:

• Parse _quarto.yml to identify active .qmd files.
• Extract all packages used via package::function() in these files.
• Install any missing packages before rendering.

Alternatively, a more general function like pak::scan_and_install() could be used for non-Quarto workflows:

# Scan an arbitrary folder for used packages and install them
pak::scan_and_install(path = ".", pattern = "\\.qmd$")

---

Benefits

1. Streamlined CI/CD Pipelines:
   Avoid pipeline failures due to missing packages by ensuring all dependencies are installed in advance.

2. Efficiency for Large Projects:
   Automatically handle dependency management for Quarto projects with multiple .qmd files and dynamic dependencies.

3. Modern Style Alignment:
   Supports best practices by enabling workflows where package::function() is preferred over global package loading.

4. Broader Use Case:
   While the focus is on Quarto projects, this feature could benefit RMarkdown users or anyone working with R scripts in CI/CD environments.

5. Optimized for Docker:
   By leveraging pak’s caching and parallelized installation, it minimizes time and resources in containerized environments.

[gaborcsardi]

This is already happening here: r-lib/pkgdepends#390