_targets.R

library(targets)
library(tarchetypes)
library(here)
library(crew)

tar_option_set(
  packages = c("tibble"),
  format = "qs",
  error = "continue",
  controller = crew_controller_local(
    workers = 4
  ),
  storage = "worker",
  retrieval = "worker"
)

tar_source()

# Targets ---------------------
list(
  # which JECFA IDs are to consider
  tar_target(
    name = jecfa_ids,
    command = seq_len(10000)
  ),

  # Scraping JECFA
  tar_target(
    name = jecfa_list,
    command = compose_jecfa_list(jecfa_ids),
    pattern = map(jecfa_ids),
    iteration = "list",
    cue = tar_cue("always")
  ),

  tar_target(
    name = jecfa_kos,
    command = get_error(jecfa_list),
    pattern = map(jecfa_list),
    iteration = "list"
  ),

  tar_target(
    name = jecfa_oks,
    command = get_result(jecfa_list),
    pattern = map(jecfa_list),
    iteration = "list"
  ),

  # Creates a single, cleaned, data frame with metadata
  tar_target(
    name = jecfa_raw,
    command = create_df(jecfa_oks)
  ),
  tar_target(
    name = jecfa,
    command = process_df(jecfa_raw)
  ),
  tar_target(
    name = jecfa_augmented,
    command = add_metadata(jecfa)
  ),

  # Descriptive tables -----------
  tar_target(tbl1, compose_tbl1(jecfa_augmented)),
  tar_target(tbl2, compose_tbl2(jecfa_augmented)),
  tar_target(tbl3, compose_tbl3(jecfa_augmented)),
  tar_target(tbl4, compose_tbl4(jecfa_augmented)),


  # TRS -------------------------
  # compose URLS to download
  tar_target(
    urls_list,
    command = compose_urls(jecfa_augmented)
  ),

  tar_target(
    urls_ok,
    command = remove_null_urls(urls_list)
  ),

  # output path for TRS (with id)
  tar_target(
    trsPaths,
    compose_filepaths(
      urls_ok,
      here::here("data/TRS")
    )
  ),

  # output path for TRS (without id)
  tar_target(
    trsPathsNoid,
    compose_filepaths(
      urls_ok,
      here::here("data/TRS_unique"),
      noid = TRUE
    )
  ),

  # define the relationship between TRS urls and JECFA
  tar_target(
    trsMapToJecfa,
    compose_maptojecfa(trsPaths)
  ),

  # extract the valid URLs as proper URL format for targets,
  # this is usefull to avoid download multiple time urls with
  # unchanged destination file.
  tar_target(
    trsUrls,
    command = purrr::map_chr(urls_ok, "url"),
    format = "url"
  ),

  # download TRS
  tar_target(
    trsDownload,
    download_trs(
      trsUrls, trsPaths, trsPathsNoid
    ),
    pattern = map(
      trsUrls, trsPaths, trsPathsNoid
    ),
    format = "file"
  ),

  tar_target(
    trsUniqueAux,
    unique(trsDownload)
  ),

  # this additional step is required to convert the character into
  # a patter of file paths for targets
  tar_target(
    trsUnique,
    trsUniqueAux,
    pattern = map(trsUniqueAux),
    format = "file"
  ),

  # parse TRS to text (only once if the file is the same)
  tar_target(
    trsParsed,
    parse_pdf(trsUnique, dpi = 75),
    pattern = map(trsUnique),
    iteration = "list"
  ),


  # same for FAS (which are provided and not scraped/downloaded)
  tar_files(
    fasPaths,
    list.files(
        here::here("data/FAS"),
        pattern = "\\.pdf$",
        full.names = TRUE
    )
  ),
  tar_target(
    fasMapToJecfa,
    jecfa_augmented |>
      dplyr::mutate(
        file = FAS |>
          stringr::str_extract("^\\d+") |>
          stringr::str_c(".pdf")
      ) |>
      dplyr::select(file, ref_id)
  ),
  tar_target(
    fasParsed,
    parse_pdf(fasPaths, dpi = 75),
    pattern = map(fasPaths),
    iteration = "list"
  ),


  # Text mining ---------------------------------------------------

  # define the keywords of interest for text-mining
  tar_target(
    keywords,
    c(
      "dose-response",
      "dose response",
      "modelling",
      "modeling",
      "bmd",
      "bmr",
      "benchmark dose",
      "benchmark-dose"
    )
  ),

  # match keywords in the parsed PDFs
  tar_target(
    TRSKeywordMatching,
    match_keywords(
      trsUnique, trsParsed, keywords, "TRS"
    ) |>
      dplyr::left_join(
        trsMapToJecfa,
        relationship = "many-to-many"
      ),
    pattern = map(trsUnique, trsParsed)
  ),
  tar_target(
    FASKeywordMatching,
    match_keywords(
      fasPaths, fasParsed, keywords, "FAS"
    ) |>
      dplyr::left_join(
        fasMapToJecfa,
        relationship = "many-to-many"
      ),
    pattern = map(fasPaths, fasParsed)
  ),

  tar_target(
    keywordMatching,
    dplyr::bind_rows(FASKeywordMatching, TRSKeywordMatching)
  ),

  # compose a dataset with all the information, metadata and keywords'
  # matching
  tar_target(
    jecfa_tm_full,
    jecfa_augmented |>
      dplyr::left_join(
        keywordMatching,
        relationship = "many-to-many"
      ) |>
      dplyr::distinct()
  ),

  # tables tm -------------------
  tar_target(tbl5, compose_tbl5(keywordMatching)),
  tar_target(tbl6, compose_tbl6(keywordMatching)),
  tar_target(tbl7, compose_tbl7(keywordMatching)),
  tar_target(tbl8, compose_tbl8(jecfa_tm_full)),
  tar_target(tbl9, compose_tbl9(jecfa_tm_full)),
  tar_target(tbl10, compose_tbl10(jecfa_tm_full)),


  tar_target(
    jecfaDistiller,
    create_distiller_jecfa(jecfa_tm_full)
  ),

  # just reorganized files in folders
  tar_target(
    fileUsed,
    get_file_used(
      jecfaDistiller, fasPaths, trsUnique
    )
  ),

  tar_target(
    moveUsed,
    move_file_used(
      fileUsed,
      here::here("data/used")
    ),
    format = "file"
  ),

  # Report ----------------------
  tar_render(
    jecfa_ws,
    here("report/jecfa_ws.Rmd")
  ),
  tar_render(
    pdf_tm_2,
    here("report/pdf_tm_2.Rmd")
  ),


  # Share -----------------------
  tar_target(
    objectToShare,
    list(
      jecfa_raw = jecfa_raw,
      jecfa = jecfa,
      jecfa_augmented = jecfa_augmented,
      jecfa_distiller = jecfaDistiller
    )
  ),
  tar_target(
    shareOutputNow,
    share_objects(
      objectToShare,
      last = FALSE
    ),
    format = "file",
    pattern = map(objectToShare)
  ),
  tar_target(
    shareOutput,
    share_objects(objectToShare),
    format = "file",
    pattern = map(objectToShare)
  )
)