rstudio / r-system-requirements

System requirements for R packages

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

System Requirements for R Packages

CI Status

R packages can depend on one another, but they can also depend on software external to the R ecosystem. On Ubuntu 18.04, for example, in order to install the curl R package, you must have previously run apt-get install libcurl. R packages often note these dependencies inside their DESCRIPTION files, but this information is free-form text that varies by package.

This repository contains a catalog of "rules" that can be used to systematically identify these dependencies and generate commands to install them.

You may be expecting to see a list like:

Package SystemRequirements Field Dependency
rgdal "for building from source: GDAL >= ..." libgdal-dev

Storing this information as a table in this format is not efficient. Many R packages do not have any system dependencies, so the table would be very sparse. Moreover, R packages are added at an exponential rate, so maintaining this data would be nearly impossible.

Instead, this repository contains a set of rules that map a SystemRequirements field, e.g. rgdal's "for building from source: GDAL >= 1.11.4 and <= 2.5.0, library from ..." to a platform specific install command: apt-get install libgdal-dev gdal-bin libproj-dev.

Usage

The primary purpose of this catalog is to support RStudio Package Manager which knows how to translate these rules into install steps for specific packages or repositiories. However, the community is free to use and contribute to these rules subject to the MIT license.

RStudio Package Manager is professionally supported, but RStudio does not offer support for these rules. Please file questions in RStudio Community or open an issue in this repository.

A similar project is maintained by R-Hub. The two catalogs have different data formats, test coverage, and target different operating systems.

Rule Coverage

The rules presented in this repository are extensively tested with the following process:

  1. A Docker container is started with a minimal base R image.
  2. A target R package is identified. The catalog of rules is applied to install any known requirements for the package into the Docker container.
  3. The package is installed.

If the package install is successful, there is a high chance the existing rules are sufficient. If the install fails, there is an indication that a rule is missing. This process is repeated for all CRAN packages across 6 Linux distributions: Ubuntu 16/18, CentOS 7/8, openSUSE 42/15.

The results are summarized below:

Percentage of CRAN Packages that Install Successfully

Ubuntu 16 Ubuntu 18 CentOS 7 CentOS 8 openSUSE 42.3 openSUSE 15.0
No Rules 78% 78.1% 77.8% 77.7% 78.2%
With Rules 93.5% 95.8% 93.7% 88.5% 89.7%

Percentage Weighted by Downloads

This table contains similar results as the table above, but adjusted by download. This metric indicates how good the rules are for the majority of packages R users are likely to install, discounting the long tail of packages that have system requirements but are not frequently used.

Ubuntu 16 Ubuntu 18 CentOS 7 CentOS 8 openSUSE 42.3 openSUSE 15.0
No Rules 90.1% 90.1% 90.1% 90% 90.2%
With Rules 98.5% 99.2% 98.6% 96.1% 96.3%

Both tests run with R 3.5.3 for all CRAN packages as of April 4, 2019.

Operating Systems

The rules in this catalog support the following operating systems:

  • Ubuntu 20.04, 22.04
  • CentOS 7
  • Rocky Linux 8*, 9
  • Red Hat Enterprise Linux 7, 8, 9
  • openSUSE 15.4, 15.5
  • SUSE Linux Enterprise 15 SP4, 15 SP5
  • Debian 10, 11, 12, unstable
  • Fedora 36, 37, 38
  • Windows (for R 4.0+ only)

* Rocky Linux 8 is specified as centos8 for backward compatibility. CentOS 8 reached end of support on December 31, 2021.


For Developers

We welcome contributions to this catalog! To report a bug or request a rule, please open an issue in this repository. To add or update a rule, fork this repository and submit a pull request.

Overview

Each system requirement rule is described by a JSON file in the rules directory. The file is named rule-name.json, where rule-name is typically the name of the system dependency.

For example, here's an excerpt from a rule for the Protocol Buffers (protobuf) library at rules/libprotobuf.json.

{
  "patterns": ["\\blibprotobuf\\b"],  // regex which matches "libprotobuf" or "LIBPROTOBUF; libxml2"
  "dependencies": [
    {
      "packages": ["protobuf-devel"],  // to install the package: "yum install protobuf-devel"
      "pre_install": [
        {
          "command": "yum install -y epel-release"  // add the EPEL repository before installing
        }
      ],
      "constraints": [
        {
          "os": "linux",
          "distribution": "centos",  // make these instructions specific to CentOS 7
          "versions": ["7"]
        }
      ]
    }
  ]
}

Other examples:

  • Simple rule: git.json
  • OS version constraints (package names vary by OS version): libmysqlclient.json
  • Pre-install steps (adding the EPEL repo on CentOS/RHEL): gdal.json
  • Post-install steps (reconfiguring R for Java): java.json

JSON Fields

{
  "patterns": [...],
  "dependencies": [
    {
      "packages": [...],
      "constraints": [
        {
          "os": ...,
          "distribution": ...,
          "versions": [...]
        }
      ],
      "pre_install": [
        {
          "command": ...,
          "script": ...
        }
      ],
      "post_install": [
        {
          "command": ...,
          "script": ...
        }
      ]
    }
  ]
}

Top-level fields

Field Type Description
patterns Array Regular expressions to match SystemRequirements fields. Case-insensitive. Note that the escape character must be escaped itself (\\. to match a dot). Use word boundaries (\\b) for more accurate matches.
Example: ["\\bgnu make\\b", "\\bgmake\\b"] to match GNU Make or gmake; OpenSSL
dependencies Array Rules for installing the dependency on one or more operating systems. See dependencies.

Dependencies

Field Type Description
packages Array Packages installed through the default system package manager (e.g. apt, yum, zypper). Examples: ["libxml2-dev"], ["tcl", "tk"]
constraints Array One or more operating system constraints. See constraints.
pre_install Array Optional commands or scripts to run before installing packages (e.g. adding a third-party repository). See pre/post-install actions.
post_install Array Optional commands or scripts to run after installing packages (e.g. cleaning up). See pre/post-install actions.

Constraints

Field Type Description
os String Operating system. Only "linux" is supported for now.
distribution String Linux distribution. One of "ubuntu", "debian", "centos", "redhat", "opensuse", "sle", "fedora"
versions Array Optional set of OS versions. If unspecified, the rule applies to all supported versions. See systems.json for supported values by OS. Example: ["18.04"] for Ubuntu.

Pre/post-install actions

Pre-install and post-install actions can be specified as either a command or script. Commands are preferred unless there's complicated logic involved.

Field Type Description
command String A shell command. Example: "yum install -y epel-release"
script String A shell script found in the scripts directory. Example: "centos_epel.sh"

Adding a rule

A typical workflow for adding a new rule:

  1. Come up with regular expressions to match all R packages with the system dependency. See sysreqs.json for a sample list of CRAN packages and their SystemRequirements fields. Note that the applicable R packages don't have to be on CRAN; they can be on GitHub or other repositories, such as Bioconductor and rOpenSci.

  2. Determine the system packages and any pre/post-install steps if needed. The more operating systems covered, the better, but it's fine if only some operating systems are covered.

    Useful resources for finding packages across different OSs:

    Or to search for packages on each OS:

    # Ubuntu/Debian
    apt-cache search <package-name>
    
    # CentOS/RHEL/Fedora
    yum search <package-name>
    
    # openSUSE/SLE
    zypper search <package-name>
  3. Add the new rule as a rule-name.json file in the rules directory.

  4. Run the schema tests and (optionally) the system package tests locally.

  5. Submit a pull request.

Testing

Schema tests

To lint and validate rules against the schema, you'll need Node.js.

# Install dependencies
npm install

# Run the tests
npm test

To list R packages and system requirements matched by a rule:

# List matching system requirements for a rule
npm run test-patterns -- rules/libcurl.json --verbose

# List matching system requirements for all rules
npm run test-patterns-all -- --verbose

# Fail if a rule doesn't match any system requirements
npm run test-patterns-all -- --strict

To update the list of R packages and system requirements used for testing, run:

make update-sysreqs

System package tests

Docker images are provided to help validate system packages on supported OSs.

Available tags:

  • focal (Ubuntu 20.04)
  • jammy (Ubuntu 22.04)
  • buster (Debian 10)
  • bullseye (Debian 11)
  • bookworm (Debian 12)
  • sid (Debian unstable)
  • centos7 (CentOS 7)
  • centos8 (Rocky Linux 8)
  • rockylinux9 (Rocky Linux 9)
  • opensuse154 (openSUSE 15.4)
  • opensuse155 (openSUSE 15.5)
  • fedora36 (Fedora 36)
  • fedora37 (Fedora 37)
  • fedora38 (Fedora 38)

To build the images:

# Build a specific image (e.g. focal)
make build-focal

# Build all images
make build-all

To test the rules:

# Test a specific rule on an OS (e.g. focal)
make test-focal RULES=rules/libcurl.json

# Test a specific rule on all OSs
make test-all RULES=rules/libcurl.json

# Test all rules on all OSs
make test-all

Schema

The JSON schema is defined in the file schema.json. Do not modify this file directly, since it is automatically generated. Instead, modify schema.template.json and then run npm run generate-schema. The generate-schema target is automatically run when running npm test.

If you need to modify the distros and/or versions supported in the schema definitions, modify systems.json.

About

System requirements for R packages

License:MIT License


Languages

Language:Dockerfile 33.5%Language:Shell 33.0%Language:JavaScript 21.9%Language:Makefile 7.0%Language:R 4.6%