R/info-ci.R

In ropensci-review-tools/pkgcheck: rOpenSci Package Checks

# CI checks have no print method, as they are currently only put into the full
# html report, and otherwise only appear in summary.

#' Get all CI badges from a repository, and check that jobs currently pass.
#'
#' This extracts badges directly from README files. If none are found and a repo
#' is GitHub, then an additional function is called to check CI status directly
#' from GitHub workflow files.
#'
#' @param u URL of repo
#' @return Character vector of hyperlinked badge images
#' @noRd
pkgchk_ci_badges <- function (u) {

    if (!curl::has_internet ()) {
        return (NULL)
    }

    orgrepo <- strsplit (u, "\\/") [[1]]
    org <- utils::tail (orgrepo, 2) [1]
    repo <- utils::tail (orgrepo, 1)
    # note: default branch is github only, so will only work if repo is also
    # mirrored on github!
    branch <- get_default_github_branch (org, repo)

    if (grepl ("github", u)) {

        u_readme <- paste0 (
            "https://raw.githubusercontent.com/",
            org,
            "/",
            repo,
            "/",
            branch,
            "/README.md"
        )

    } else if (grepl ("gitlab", u)) {

        u_readme <- paste0 (
            "https://gitlab.com/",
            org,
            "/",
            repo,
            "/-/raw/",
            branch,
            "/README.md"
        )
    }

    if (!url_exists (u_readme, quiet = TRUE)) {
        return (NULL)
    }

    f <- tempfile (fileext = ".md")
    chk <- utils::download.file (u_readme, destfile = f, quiet = TRUE) # nolint
    readme <- rm_html_comments (readLines (f, encoding = "UTF-8"))

    badges <- unlist (regmatches (
        readme,
        gregexpr ("https.*\\.svg", readme)
    ))
    if (length (badges) == 0) {
        return (NULL)
    }

    platforms <- paste0 ("https\\:\\/\\/", c ("github", "gitlab"))
    badges <- badges [grep (
        paste0 (platforms, collapse = "|"),
        badges
    )]

    for (p in platforms) {

        index <- grep (p, badges)
        p_u <- p
        p_no_regex <- gsub ("\\\\", "", p)

        if (p == "https\\:\\/\\/github") {

            wf_nms <- vapply (badges [index], function (i) {
                utils::tail (strsplit (i, "/") [[1]], 2) [1]
            },
            character (1),
            USE.NAMES = FALSE
            )

            p_u <- paste0 (
                "https://github.com/",
                org,
                "/",
                repo,
                "/actions"
            )

        }

        badges [index] <- paste0 (
            "[![",
            wf_nms,
            "](",
            badges [index],
            ")](",
            p_u,
            ")"
        )
    }

    return (badges)
}

#' Remove all html-comment chunks from readme
#'
#' See #109: Some READMEs may have badges within html comments, and these should
#' be removed prior to checking presence of CI badges.
#' @param x Character vector of the README file.
#' @noRd
rm_html_comments <- function (x) {

    # First remove single-line comments:
    x <- gsub ("<!\\-\\-.*\\-\\->", "", x)

    # Then identify and remove multi-line html comments:
    cmt_open <- which (regexpr ("<!\\-\\-", x) > 0L)
    cmt_close <- which (regexpr ("\\-\\->", x) > 0L)

    if (length (cmt_open) == length (cmt_close) & length (cmt_open) > 0L) {
        g <- cbind (cmt_open, cmt_close)

        index <- unlist (apply (g, 1, function (i) seq (i [1], i [2])))
        if (methods::is (index, "matrix")) {
            index <- sort (as.vector (index))
        }

        x <- x [-index]
    }

    return (x)
}


#' CI results for GitHub only
#' @inheritParams pkg_uses_roxygen2
#' @return A 'data.frame' with one row for each GitHub workflow, and columns for
#' name, the 'conclusion' status, the git 'sha', and the date.
#' @noRd
ci_results_gh <- function (path) {

    u <- pkginfo_url_from_desc (path)
    if (length (u) == 0L) {
        return (NULL)
    }

    if (grepl ("github\\.io", u)) {

        u <- pkginfo_url_from_desc (path, type = "BugReports")
        if (nzchar (u)) {
            u <- gsub ("\\/issues(\\/?)$", "", u)
        } else {
            return (NULL)
        }
    }

    url <- strsplit (u, "\\/") [[1]]
    org <- utils::tail (url, 2) [1]
    repo <- utils::tail (url, 1)

    url <- paste0 (
        "https://api.github.com/repos/",
        org,
        "/",
        repo,
        "/actions/runs"
    )

    runs <- jsonlite::fromJSON (url)

    if (!"total_count" %in% names (runs)) {
        return (NULL)
    }

    if (runs$total_count == 0) {
        return (NULL)
    }

    dat <- data.frame (
        id = runs$workflow_runs$id,
        name = runs$workflow_runs$name,
        status = runs$workflow_runs$status,
        conclusion = runs$workflow_runs$conclusion,
        sha = runs$workflow_runs$head_sha,
        run_number = runs$workflow_runs$run_number,
        time = runs$workflow_runs$created_at,
        stringsAsFactors = FALSE
    )

    dat$time <- strptime (dat$time, "%Y-%m-%dT%H:%M:%SZ")
    dat$time_dbl <- as.double (dat$time)
    # non-dply group_by %>% summarise:
    dat <- lapply (
        split (dat, f = as.factor (dat$name)),
        function (i) {
            i [which.max (i$time_dbl), ]
        }
    )
    dat <- do.call (rbind, dat)

    dat$sha <- substring (dat$sha, 1, 6)
    dat$date <- strftime (dat$time, "%Y-%m-%d")
    rownames (dat) <- dat$time_dbl <- dat$time <- dat$status <- NULL

    return (dat)
}