smvr

Simple implementation of Semantic Versioning 2.0.0 on R.

smvr provides a lightweight, fully vectorized R class for Semantic Versioning. It enables type-safe manipulation, comparison, and sorting of version numbers, and integrates smoothly with tidyverse tools.

Installation

Install the latest release version from CRAN:

install.packages("smvr")

The development version can be installed from GitHub:

pak::pak("eitsupi/smvr")

Example

library(smvr)

# Parse version characters into smvr objects
v <- parse_semver(c("1.0.0", "1.0.0-alpha.2", "1.0.0-alpha.10", "1.0.1+20250621", "0.9.0"))

# Sort versions
sort(v)
#> <smvr[5]>
#> [1] 0.9.0          1.0.0-alpha.2  1.0.0-alpha.10 1.0.0          1.0.1+20250621

# Can compare with string notation
v["1.0.0-alpha" < v & v < "1.0.0"]
#> <smvr[2]>
#> [1] 1.0.0-alpha.2  1.0.0-alpha.10

# Works with tibble data frame and dplyr
tibble::tibble(version = v) |>
  dplyr::arrange(version) |>
  dplyr::mutate(
    `>= 1.0.0` = version >= "1.0.0",
    `pre-release` = is_pre_release(version),
  )
#> # A tibble: 5 × 3
#>          version `>= 1.0.0` `pre-release`
#>           <smvr> <lgl>      <lgl>        
#> 1          0.9.0 FALSE      FALSE        
#> 2  1.0.0-alpha.2 FALSE      TRUE         
#> 3 1.0.0-alpha.10 FALSE      TRUE         
#> 4          1.0.0 TRUE       FALSE        
#> 5 1.0.1+20250621 TRUE       FALSE

Features

• Fully vectorized Semantic Versioning class.
• Type-safe comparison and sorting.
• Tidyverse compatibility ({tibble}, {dplyr}, etc.).
• No dependencies except {vctrs}.

Related Works

• The numeric version class vector can be crated with numeric_version() in base R works well for versions that only have MAJOR.MINOR.PATCH. But it does not support pre-release identifiers of SemVer, so in the case of including pre-release versions, it is not suitable.
• The semver package is a wrapper for a C++ SemVer parser. The class provided by this package is a special list, which does not work well with {tibble} and {dplyr}.
• The semverutils package has a SemVer parser written in R. It creates a single version as an {R6} object, which is not vectorized. Also, at the moment (version 0.1.0, published 2020-02-22 on CRAN), it has a bug in comparing pre-release versions¹, so it does not work correctly.