R/rating_get_candidate_ratings.R

In decktools/votesmart: Wrapper for the Project 'VoteSmart' API

#' Get SIG (Special Interest Group) ratings for candidates
#'
#' @param candidate_ids A vector of candidate ids.
#' @param sig_ids A vector of SIG ids. Default is \code{""} for all SIGs.
#' @param all Boolean: should all possible combinations of the variables be searched for, or just the exact combination of them in the order they are supplied?
#' @param verbose Should cases when no data is available be messaged?
#'
#' @return A dataframe with the columns \code{rating_id, candidate_id, sig_id, rating, rating_name, timespan, categories, rating_text}.
#' @export
#'
#' @examples
#' \dontrun{
#' pelosi_id <- "26732"
#' rating_get_candidate_ratings(pelosi_id)
#' }
rating_get_candidate_ratings <- function(
    candidate_ids,
    sig_ids = "",
    all = TRUE,
    verbose = TRUE) {
  candidate_ids %<>%
    as_char_vec()

  sig_ids %<>%
    as_char_vec()

  if (all) {
    query_df <-
      expand_grid(
        candidate_id = candidate_ids,
        sig_id = sig_ids
      ) %>%
      mutate(
        query = glue::glue("&candidateId={candidate_id}&sigId={sig_id}")
      )
  } else {
    arg_lengths <-
      c(length(candidate_ids), length(sig_ids)) %>%
      magrittr::extract(
        !. == 1
      )

    if (length(arg_lengths) > 1 && (max(arg_lengths) - min(arg_lengths) != 0)) {
      stop("If `all` is FALSE, lengths of inputs must be equivalent to each other, or 1.")
    }

    query_df <-
      tibble(
        candidate_id = candidate_ids,
        sig_id = sig_ids
      ) %>%
      mutate(
        query = glue::glue("&candidateId={candidate_id}&sigId={sig_id}")
      )
  }

  r <- "Rating.getCandidateRating?"

  out <- tibble()

  for (i in 1:nrow(query_df)) {
    candidate_id <- query_df$candidate_id[i]
    sig_id <- query_df$sig_id[i]
    q <- query_df$query[i]

    message(glue::glue(
      "Requesting data for {{candidate_id: {candidate_id}, sig_id: {sig_id}}}."
    ))

    suppressWarnings(
      this <-
        get(
          req = r,
          query = q,
          level_one = "candidateRating",
          level_two = "rating"
        )
    )

    if (all(is.na(this))) {
      if (verbose) {
        message(glue::glue(
          "No results found for query {q}."
        ))
      }

      this <-
        query_df %>%
        select(-query) %>%
        vs_na_if("")
    } else {
      suppressWarnings({
        # For the case where we fixed up the JSON which didn't end with `}}`
        if ("categories" %in% names(this) && nrow(this) > 1) {
          this$categories %<>%
            purrr::map(as_tibble) %>%
            purrr::map(distinct) %>%
            purrr::map(mutate, rn = row_number()) %>%
            purrr::map(
              tidyr::pivot_wider,
              values_from = c(categoryId, name),
              names_from = rn
            )

          this %<>%
            tidyr::unnest(categories) %>%
            clean_df()
        } else {
          this %<>%
            rename_all(
              stringr::str_remove,
              "categories_"
            ) %>%
            rename_all(
              stringr::str_remove,
              "_category"
            )

          this %<>%
            # Distinct all the category values which are sometimes doubled up
            tidyr::pivot_longer(contains("category")) %>%
            group_by(rating_id) %>%
            distinct(value, .keep_all = TRUE) %>%
            ungroup() %>%
            tidyr::drop_na(value) %>%
            mutate(
              rating_id_nester = rating_id
            ) %>%
            # Split into individual tibbles by rating_id, apply chunk_it to each, and then recombine
            tidyr::nest(-rating_id_nester) %>%
            pull(data) %>%
            purrr::map(chunk_it, n_per_chunk = 2) %>%
            bind_rows() %>%
            rowwise() %>%
            # Rename categories now that we've deduped
            mutate(
              type =
                case_when(
                  stringr::str_detect(value, "[0-9]") ~ "id",
                  TRUE ~ "name"
                ),
              name = glue::glue("category_{type}_{chunk}")
            ) %>%
            select(-chunk, -type) %>%
            # Back to wide format
            tidyr::pivot_wider() %>%
            ungroup() %>%
            tidyr::unnest()
        }
      })

      this %<>%
        mutate(
          candidate_id = candidate_id
        ) %>%
        select(
          rating_id,
          candidate_id,
          sig_id,
          rating,
          rating_name,
          timespan,
          rating_text,
          everything()
        )

      out %<>%
        bind_rows(this)
    }
  }

  if ("categories" %in% names(out)) {
    out %<>%
      select(-categories)
  }

  suppressWarnings(
    out %>%
      tidyr::unnest() %>%
      distinct()
  )
}