R/GIFT_traits.R

In GIFT: Access to the Global Inventory of Floras and Traits (GIFT)

#' Trait values at the species level
#'
#' Retrieve specific trait values.
#'
#' @param trait_IDs a character string indicating which trait you want to
#' retrieve. Traits must belong to the available list of traits.
#' 
#' @param agreement Percentage of resources that agree on an aggregated trait
#' value, entries below this threshold will be omitted. 
#' 
#' @param bias_ref When FALSE, exclude entries that are only based on a
#' resource that potentially introduces a bias (e.g. a resource only including
#' trees).
#' 
#' @param bias_deriv When FALSE, exclude entries that are only based on a
#' derivation that potentially introduces a bias (e.g. all phanerophytes being
#' woody but some life forms being ambiguous).
#' 
#' @template GIFT_version_api
#' 
#' @return
#' A long-format data frame with 6 columns: `trait_ID`, `work_ID`, `species`,
#' `trait_value`, `agreement` and `references`.
#'
#' @details Here is the detail of each column:
#' 
#' \emph{trait_ID} - Identification number of the trait\cr
#' \emph{work_ID} - Identification number of the taxonomically harmonized
#'  species\cr
#' \emph{species} - Species name\cr
#' \emph{trait_value} - Value of the trait\cr
#' \emph{agreement} - Agreement score between the different sources for that
#' trait value, only for categorical traits\cr
#' \emph{cv} - Coefficient of variation for the different sources for that
#' trait value, only for numeric traits\cr
#' \emph{n} - Number of sources leading to the trait value\cr
#' \emph{references} - ref_ID from which we got the trait information
#'
#' @references
#'      Denelle, P., Weigelt, P., & Kreft, H. (2023). GIFT—An R package to
#'      access the Global Inventory of Floras and Traits. Methods in Ecology
#'      and Evolution, 14, 2738-2748.
#'      https://doi.org/10.1111/2041-210X.14213
#' 
#'      Weigelt, P, König, C, Kreft, H. GIFT – A Global Inventory of Floras and
#'      Traits for macroecology and biogeography. J Biogeogr. 2020; 47: 16– 43.
#'      https://doi.org/10.1111/jbi.13623
#'
#' @seealso [GIFT::GIFT_traits_meta()]
#'
#' @examples
#' \donttest{
#' self_fertilization <- GIFT_traits(trait_IDs = "3.1.1", agreement = 0.66,
#' bias_ref = FALSE, bias_deriv = FALSE)
#' }
#' 
#' @importFrom jsonlite read_json
#' @importFrom dplyr bind_rows left_join relocate mutate_at
#' @importFrom tidyr pivot_wider
#' @importFrom utils setTxtProgressBar txtProgressBar
#' 
#' @export

GIFT_traits <- function(
    trait_IDs = "", agreement = 0.66, bias_ref = TRUE, bias_deriv = TRUE,
    api = "https://gift.uni-goettingen.de/api/extended/",
    GIFT_version = "latest"){
  
  # 1. Controls ----
  api_check <- check_api(api)
  if(is.null(api_check)){
    return(NULL)
  } else{
    # Arguments
    check_trait_IDs(trait_IDs)
    
    if(!is.numeric(agreement)){
      stop(
        "agreement must be a numeric between 0 and 1 indicating the proportion 
         of original trait values that needs to support the aggregated value in 
         case of categorical traits.")
    } else if(agreement > 1 | agreement < 0){
      stop(
        "agreement must be a numeric between 0 and 1 indicating the proportion 
         of original trait values that needs to support the aggregated value in 
         case of categorical traits.")
    }
    
    check_bias_ref(bias_ref)
    check_bias_deriv(bias_deriv)
    GIFT_version <- suppressMessages(check_gift_version_simple(GIFT_version))
    
    # Load traits_metadata to check if the provided IDs are available
    tmp <- GIFT_traits_meta(api = api, GIFT_version = GIFT_version)
    if(!all(trait_IDs %in% tmp$Lvl3)){
      stop(
        "trait_IDs must belong to the available list of traits. To see which
           traits are available, run 'GIFT_traits_meta() and look at column
           'Lvl3'.")
    }
    
    if(GIFT_version == "1.0" & (bias_ref == FALSE | bias_deriv == FALSE)){
      message(
        "Warning: In GIFT version 1.0 it is not yet possible to filter trait 
      values for biases. bias_ref and bias_deriv arguments are ignored.")
    }
    
    # 2. Function ----
    # Get species names
    message("\nRetrieving species' names.\n")
    
    species <- suppressMessages(GIFT_species(GIFT_version = GIFT_version, 
                                             api = api))
    
    message(paste0("Preparing the download of trait data for ",
                   length(unique(trait_IDs)),
                   " trait(s).\n"))
    
    # Initiating list
    trait_list <- list()
    
    n <- ceiling(tmp$count[match(trait_IDs, tmp$Lvl3)]/10000)
    
    progress <- utils::txtProgressBar(min = 0, max = sum(n)+1, initial = 0) 
    
    count <- 1
    utils::setTxtProgressBar(progress, count)
    
    for(i in seq_along(trait_IDs)){
      trait_list_i <- list()
      
      for (k in seq_len(n[i])){
        trait_list_i[[k]] <- jsonlite::read_json(
          paste0(api, "index", ifelse(GIFT_version == "beta", "",
                                      GIFT_version),
                 ".php?query=traits&traitid=",
                 trait_IDs[i], "&biasref=", as.numeric(bias_ref),
                 "&biasderiv=", as.numeric(bias_deriv), 
                 "&startat=", as.integer((k-1)*10000)),
          simplifyVector = TRUE)
        count <- count + 1
        utils::setTxtProgressBar(progress, count)
      }
      trait_list[[i]] <- dplyr::bind_rows(trait_list_i)
      trait_list[[i]]$trait_ID <- trait_IDs[i]
      
    }
    
    # trait_list <- purrr::map(
    #   .x = seq_along(trait_IDs),
    #   .f = function(x){
    #     trait_list_x <- list()
    #     
    #     trait_list_x <- purrr::map(
    #       .x = seq_len(n[x]),
    #       .f = function(y){
    #         jsonlite::read_json(
    #           paste0(api, "index", ifelse(GIFT_version == "beta", "",
    #                                       GIFT_version),
    #                  ".php?query=traits&traitid=",
    #                  trait_IDs[x], "&biasref=", as.numeric(bias_ref),
    #                  "&biasderiv=", as.numeric(bias_deriv),
    #                  "&startat=", as.integer((y-1)*10000)),
    #           simplifyVector = TRUE)
    #       },
    #       .progress = paste0("Retrieving trait number ", x))
    #     
    #     trait_list_x <- dplyr::bind_rows(trait_list_x)
    #     trait_list_x$trait_ID <- trait_IDs[x]
    #     return(trait_list_x)
    #   },
    #   .progress = TRUE)
    
    message("\n")
    
    # Formatting trait_list as a data.frame
    trait_list <- dplyr::bind_rows(trait_list)
    trait_list <- trait_list[which(trait_list$agreement >= agreement |
                                     is.na(trait_list$agreement)), ]
    
    # Make certain columns numeric
    trait_list <- dplyr::mutate_at(
      trait_list, c("work_ID", "agreement", "coeff_var", "n"), as.numeric)
    
    # Add species names
    trait_list <- dplyr::left_join(trait_list,
                                   species[, c("work_ID", "work_species", 
                                               "work_author")],
                                   by = "work_ID")
    
    # Round agreement score
    trait_list$agreement <- round(trait_list$agreement, 3)
    
    # Reordering columns
    trait_list <- trait_list[, c("work_ID", "work_species", "work_author",
                                 "trait_ID", "trait_value", "agreement",
                                 "coeff_var", "n", "references")]
    
    # Renaming column
    colnames(trait_list)[colnames(trait_list) == "coeff_var"] <- "cv"
    
    # Wider format
    trait_list <- tidyr::pivot_wider(
      trait_list, names_from = "trait_ID",
      values_from = c("trait_value", "agreement", "cv", "n", "references"))
    
    # Make data.frame
    trait_list <- as.data.frame(trait_list)
    
    # Remove agreement columns for continuous traits and coeff_var for
    # categorical traits
    numeric_traits <- tmp[which(tmp$type == "numeric"), "Lvl3"]
    numeric_columns <-
      paste("agreement", numeric_traits, sep = "_")[
        paste("agreement", numeric_traits, sep = "_") %in%
          colnames(trait_list)]
    
    categorical_traits <- tmp[which(tmp$type != "numeric"), "Lvl3"]
    categorical_columns <-
      paste("cv", categorical_traits, sep = "_")[
        paste("cv", categorical_traits, sep = "_") %in% colnames(trait_list)]
    
    trait_list <- trait_list[, !(colnames(trait_list) %in% numeric_columns)]
    trait_list <- trait_list[, !(colnames(trait_list) %in%
                                   categorical_columns)]
    
    # Make numeric trait values numeric
    numeric_columns <-
      paste("trait_value", numeric_traits, sep = "_")[
        paste("trait_value", numeric_traits, sep = "_") %in%
          colnames(trait_list)]
    trait_list <- dplyr::mutate_at(trait_list, numeric_columns, as.numeric)
    
    return(trait_list)
  }
}