Script for finding loaded but unused packages in R scripts

analyze_pkg_usage.R

#!/usr/bin/env Rscript

suppressPackageStartupMessages({
    library(globals)
    library(readr)
    library(stringr)
    library(rex)
    library(magrittr)
    library(rlang)
    library(knitr)
    library(assertthat)
    library(withr)
    library(glue)
    library(BiocParallel)
    register(MulticoreParam())
})

match_identical <- function(x, table, nomatch=NA_integer_) {
    assert_that(is_list(table))
    for (i in seq_along(table)) {
        if (identical(x, table[[i]])) {
            return(i)
        }
    }
    return(nomatch)
}

get_Rmd_source <- function(fname) {
    with_tempfile("temp", {
        purl(fname, output=temp, quiet=TRUE)
        read_file(temp)
    })
}

analyze_pkg_usage <- function(fname) {
    if (str_detect(fname, regex(rex(".Rmd", end), ignore_case=TRUE))) {
        ## Rmd file
        file_text <- get_Rmd_source(fname)
    } else {
        # R file
        file_text <- read_file(fname)
    }
    lib_regex <- rex(
        "library",
        zero_or_more(space),
        "(",
        zero_or_more(space),
        capture(one_or_more(any, type="lazy")),
        zero_or_more(space),
        one_of(",", ")"))
    loaded_pkgs <- unique(do.call(rbind, str_match_all(file_text, lib_regex))[,2])
    missing_pkgs <- character(0)
    for (pkg in loaded_pkgs) {
        if (!suppressMessages(suppressWarnings(
            library(package=pkg, character.only=TRUE, logical.return = TRUE)))) {
            missing_pkgs <- c(missing_pkgs, pkg)
        }
    }
    loaded_pkgs %<>% setdiff(missing_pkgs)
    package_envs <- loaded_pkgs %>% set_names %>%
        lapply(. %>% str_c("package:", .) %>% as.environment)
    package_namespaces <- loaded_pkgs %>% set_names %>% lapply(asNamespace)

    file_exprs <- parse(text=file_text)
    expr_globals <- lapply(file_exprs, globalsOf, mustExist=FALSE)
    used_envs <- unique(do.call(c, lapply(expr_globals, attr, "where")))
    used_pkgs <- character(0)
    for (env in used_envs) {
        for (envlist in list(package_envs, package_namespaces)) {
            found <- match_identical(env, envlist)
            if (!is.na(found)) {
                used_pkgs <- c(used_pkgs, names(envlist)[found])
                break()
            }
        }
    }

    pkg_usage <- list(Used = used_pkgs,
                               Unused = setdiff(loaded_pkgs, used_pkgs))
    if (length(missing_pkgs) > 0) {
        pkg_usage$Missing <- missing_pkgs
    }
    pkg_usage
}

fnames <- commandArgs(TRUE)

pkg_usage <- bplapply(set_names(fnames), function(fname) {
    message("Analyzing ", deparse(fname))
    try(analyze_pkg_usage(fname))
})

message("Package usage:")
print(pkg_usage)
saveRDS(pkg_usage, "pkg_usage.RDS")

Thanks, I don't know how I missed that! I could install BiocParallel using the Bioconductor manager. The script is missing some of the packages that I know for sure are mandatory in my script, but it can still be helpful to find some of the useless ones. Thank you.

If you think my script is missing some cases, I'd like to know about it so I can fix it.

My script depends on functions from the packages:

• pacman (function p_load())
• ggspatial (layer_spatial())
• data.table (fread())
• dplyr (group_by())
• ggrepel (geom_text_repel())
• rgdal
• raster (shapefile())
• lubridate (ymd())

It fails if I don't require one of those packages. The analysis script found that sf is unused and this is correct (in fact, the line library(sf) was commented out in the R script, but the analysis script found it). However, it only listed one used package. I can't remember which one now but I am happy to check later when I'm back at that computer.

It is entirely possible that I just didn't use the script correctly or didn't quite understand what it aims to output, I'm sorry if that is the case.