In mlr-org/mlrCPO: Composable Preprocessing Operators and Pipelines for Machine Learning

CPO Vignette Navigation

# output format should be of the form
#> output
#> output
knitr::opts_chunk$set(collapse = TRUE, comment = "#>")

# initialize: load library, make everything deterministic
library("mlrCPO")
set.seed(123)

# get the path of the parent document
# path = names(knitr::opts_knit$get("encoding"))[1]
base = knitr::opts_knit$get("output.dir")
file = sys.frame(min(grep("^knitr::knit$|^knit$", sapply(sys.calls(), function(x) as.character(x)[1]))))$input
file = basename(file)
path = file.path(base, file)
rpath = gsub("\\.[^.]*$", ".R", path)

# strip whitespace from lines in tangle (R file) output for lintr
knitr::knit_hooks$set(document = function(x) {
  if (file_test("-f", rpath)) {
    lines = readLines(rpath)
    lines = gsub(" *(\n|$)", "\\1", lines)
    cat(lines, file = rpath, sep = "\n", append = FALSE)
  }
  x
})

#############################
# do the trans-vignette ToC #
#############################
fullfile = file

allfiles = list.files(path = base, pattern = ".*\\.Rmd$")
stopifnot(file %in% allfiles)

# collect information (title, url, main / compact) for each file in vignette dir
fileinfolist = list()
for (cf in allfiles) {
  ismain = TRUE
  if (grepl("^z_", cf)) {
    infoslot = gsub("^z_", "", cf)
    infoslot = gsub("_terse\\.Rmd$", "", infoslot)
    subslot = "compact"
  } else {
    infoslot = gsub("^a_", "", cf)
    infoslot = gsub("\\.Rmd$", "", infoslot)
    subslot = "main"
  }

  content = scan(paste(base, cf, sep = "/"), what = "character", quiet = TRUE)
  pos = min(c(which(content == "title:"), Inf))
  if (is.infinite(pos)) {
    stop(sprintf("parsing error: %s", cf))
  }
  infolist = list(title = content[pos + 1], url = cf, iscurrent = cf == file)

  applist = list(infolist)
  names(applist) = subslot
  fileinfolist[[infoslot]] = c(fileinfolist[[infoslot]], applist)
}

# helper function that creates a link for all files except the current one
linkify = function(info, title) {
  if (info$iscurrent) {
    title
  } else {
    sprintf("[%s](%s)", title, gsub("\\.Rmd$", ".html", info$url))
  }
}

# output ToC
for (idx in seq_along(fileinfolist)) {
  content = fileinfolist[[sort(names(fileinfolist))[idx]]]
  if (!is.null(content$compact)) {
    if (paste(sub("[0-9]\\. ", "", content$main$title), "(No Output)") != sub("^z ", "", content$compact$title)) {
      stop(sprintf("File %s and its compact version %s have incompatible titles\nThe compact version must be paste(main_title, \"(No Output)\"). Is: '%s', expected: '%s'",
        content$main$url, content$compact$url, content$compact$title, paste(content$main$title, "(No Output)")))
    }
    line = sprintf("%s (%s)", linkify(content$main, content$main$title), linkify(content$compact, "compact version"))
  } else {
    line = linkify(content$main, content$main$title)
  }
  cat(sprintf("%s. %s\n", idx, line))
  if (content$main$iscurrent || content$compact$iscurrent) {
    fullfile = content$main$url
  }
}

fullpath = file.path(base, fullfile)

#############################
# Optional Document TOC     #
#############################

# print everything up to level `print.level`.
# level is the number of '#' prefixes. The lowest level is usually 2.
printToc = function(print.level = 3) {
  owncontent = readLines(fullpath)
  tripletic = grepl("^```", owncontent)
  owncontent = owncontent[cumsum(tripletic) %% 2 == 0]  # exclude ```-delimited code
  headlines = grep("^#+ +", owncontent, value = TRUE)
  headlevels = nchar(gsub(" .*", "", headlines))
  headlines = gsub("^[#]+ +", "", headlines)

  links = gsub("[^-a-z. ]", "", tolower(headlines))
  links = gsub(" +", "-", links)
  links = gsub("-$", "", links)

  if (!sum(headlevels <= print.level)) {
    return(invisible(NULL))
  }

  cat("<h", headlevels[1], ">Table of Contents</h", headlevels[1], ">\n<div id=\"TOC\">\n", sep = "")

  lastlevel = headlevels[1] - 1
  for (idx in seq_along(headlines)) {
    line = headlines[idx]
    level = headlevels[idx]
    link = links[idx]
    if (level > print.level) {
      next
    }
    if (level < headlevels[1]) {
      stop("First headline level must be the lowest one used, but '", line, "' is lower.")
    }
    lvldiff = level - lastlevel
    if (lvldiff > 1) {
      stop("Cannot jump headline levels. Error on: ", line)
    }
    if (lvldiff > 0) {
      # higher level -> open a <ul>
      cat("<ul>")
    } else {
      cat("</li>\n")
    }
    if (lvldiff < 0) {
      # lower level -> close a few <ul>
      for (l in seq_len(-lvldiff)) {
        cat("</ul></li>")
      }
    }
    cat("<li><a href=\"#", link, "\">", line, "</a>", sep = "")
    lastlevel = level
  }
  # if the last level is greater than the first level, close a few <ul>
  lvldiff = lastlevel - headlevels[1]

  cat("</li></ul>\n</div>\n")
}



#############################
# Some output settings      #
#############################

options(width = 80)


replaceprint = function(ofunc) {
  force(ofunc)
  function(x, ...) {
    cu = capture.output({ret = ofunc(x, ...)})
    cu = grep("time: [-+e0-9.]{1,6}", cu, value = TRUE, invert = TRUE)
    cat(paste(cu, collapse = "\n"))
    if (!grepl("\n$", tail(cu, 1))) {
      cat("\n")
    }
    ret
  }
}

for (pfunc in grep("print\\.", ls(asNamespace("mlr")), value = TRUE)) {
  ofunc = get(pfunc, asNamespace("mlr"))
  assign(pfunc, replaceprint(ofunc))
}

mlr-org/mlrCPO documentation built on Nov. 18, 2022, 11:25 p.m.