#' Reads docx file and splits each table into list
#'
#' @param path file path
#' @param data.type character vector. Could be "paragraph" or "table cell".
#'
#' @return
#' @export
#'
#' @examples
docx2ds <- function(path = here::here("data-raw/Deltagerliste Skirva 2024.docx"),
                    data.type = "table cell", verbose = TRUE) {
  # Ref: https://www.r-bloggers.com/2020/07/how-to-read-and-create-word-documents-in-r/
  doc <- officer::read_docx(path)

  content <- doc |> officer::docx_summary()

  if (verbose) {
    message("Content types in the current document are as follows:")
    print(content$content_type |> unique())
  }

  table_cells <- content |> dplyr::filter(content_type %in% data.type)

  # .x <- split(table_cells, table_cells$doc_index)[[4]]

  split(table_cells, table_cells$doc_index) |> purrr::map(function(.x) {
    table_data <- .x |>
      dplyr::filter(!is_header) |>
      dplyr::select(row_id, cell_id, text)

    # split data into individual columns
    splits <- split(table_data, table_data$cell_id)
    splits <- lapply(splits, function(.y) .y$text)
    splits <- splits |> 
      purrr::keep(function(.y) length(.y)>1) 

    # If a footer has been added, it is considered part of the first column,
    # and will result in unequal col lengths.
    # This solution does not handle merged cells
    col_lengths <- lengths(splits)
    
    if (col_lengths[1] > col_lengths[2]){
      splits[[1]] <- splits[[1]][seq_len(col_lengths[2])]
    }
    
    # combine columns back together in wide format
    table_result <- splits |> 
      dplyr::bind_cols()
    
    # get table headers
    cols <- .x |> dplyr::filter(is_header)
    names(table_result) <- cols$text
    table_result
  })
}

get_coefs <- function(path, 
                      index.table = 1) {
  data = docx2ds(
    path = path
  )
  
  data |>
    purrr::pluck(index.table) |>
    setNames(c(
      "variable",
      lapply(c("drop", "hop"),
        paste,
        c("median", "mean"),
        sep = "_"
      ) |>
        purrr::list_c()
    )) |>
    dplyr::select(variable, tidyselect::ends_with("median")) #|>
  # dplyr::mutate(dplyr::across(tidyselect::ends_with("median"),~as.numeric))
  # setNames(c("variable","decrease","increase"))
}

gtsummary2docx <- function(data, path) {
  data |>
    gtsummary::as_flex_table() |>
    flextable::save_as_docx(path = path)
}