R/getPR.R

In malariaAtlas: An R Interface to Open-Access Malaria Data, Hosted by the 'Malaria Atlas Project'

#' Download PR points from the MAP database
#'
#' \code{getPR} downloads all publicly available PR points for a specified country (or countries) and returns this as a dataframe.
#'
#' \code{country} and \code{ISO} refer to countries and a lower-level administrative regions such as Mayotte and French Guiana.
#' While we cannot direectly distribute DHS coordinates, we can distribute the number of examined and positive. If the coordinates
#' are needed they can be downloaded from www.measuredhs.com, via the rdhs package or using malariaAtlas:fillDHSCoordinates().
#'
#' @param country string containing name of desired country, e.g. \code{ c("Country1", "Country2", ...)} OR \code{ = "ALL"}. (Use one of \code{country} OR \code{ISO} OR \code{continent}, not combined)
#' @param ISO string containing ISO3 code for desired country, e.g. \code{c("XXX", "YYY", ...)} OR \code{ = "ALL"}. (Use one of \code{country} OR \code{ISO} OR \code{continent}, not combined)
#' @param continent string containing continent (one of "Africa", "Americas", "Asia", "Oceania") for desired data, e.g. \code{c("continent1", "continent2", ...)}. (Use one of \code{country} OR \code{ISO} OR \code{continent}, not combined)
#' @param species string specifying the Plasmodium species for which to find PR points, options include: \code{"Pf"} OR \code{"Pv"} OR \code{"BOTH"}
#' @param extent an object specifying spatial extent within which PR data is desired, as returned by sf::st_bbox() - the first column has the minimum, the second the maximum values; rows 1 & 2 represent the x & y dimensions respectively (matrix(c("xmin", "ymin","xmax", "ymax"), nrow = 2, ncol = 2, dimnames = list(c("x", "y"), c("min", "max"))))
#' @param start_date string object representing the lower date to filter the PR data by (inclusive) e.g. '2020-01-01'
#' @param end_date string object representing the upper date to filter the PR data by (exclusive) e.g. '2020-01-01'
#' @param version (optional) The PR dataset version to return. If not provided, will just use the most recent version of PR data. (To see available version options, 
#' use listPRPointVersions)
#' 
#' @return \code{getPR} returns a dataframe containing the below columns, in which each row represents a distinct data point/ study site.
#'
#' \enumerate{
#' \item \code{dhs_id} The dhs survey id if appropriate.
#' \item \code{site_id} Unique site identifier
#' \item \code{site_name} Name of site.
#' }
#'
#' @examples
#' #Download PfPR data for Nigeria and Cameroon and map the locations of these points using autoplot
#' \dontrun{
#' NGA_CMR_PR <- getPR(country = c("Nigeria", "Cameroon"), species = "Pf")
#' autoplot(NGA_CMR_PR)
#'
#' #Download PfPR data for Madagascar and map the locations of these points using autoplot
#' Madagascar_pr <- getPR(ISO = "MDG", species = "Pv")
#' autoplot(Madagascar_pr)
#'
#' getPR(country = "ALL", species = "BOTH")
#' }
#'
#'
#' @seealso \code{autoplot} method for quick mapping of PR point locations (\code{\link{autoplot.pr.points}}).
#'
#'
#' @export getPR

getPR <- function(country = NULL,
                  ISO = NULL,
                  continent = NULL,
                  species = NULL,
                  extent = NULL,
                  start_date = NULL,
                  end_date = NULL,
                  version = NULL) {
  if (is.null(country) &
      is.null(ISO) &
      is.null(continent) & is.null(extent)) {
    stop("Must specify one of: 'country', 'ISO', 'continent', or 'extent'.")
  }
  
  if (!is.null(extent)) {
    extent <- getSpBbox(extent)
  }
  
  if (is.null(version)) {
    version <- getLatestPRPointVersion()
    message('Please Note: Because you did not provide a version, by default the version being used is ', version, 
            ' (This is the most recent version of PR data. To see other version options use function listPRPointVersions)')
  } else {
    df_available_versions <- listPRPointVersions(printed = FALSE)
    if (!version %in% df_available_versions$version) {
      stop(
        paste0(
          'Version provided is not valid. Valid versions for PR point data can be found using listPRPointVersions() and are ["',
          paste(df_available_versions$version, collapse = '", "'),
          '"]. Otherwise, you can choose to not specify a version, and the most recent version will be automatically selected'
        )
      )
    }
  }
  
  wfs_client <- get_wfs_clients()$Malaria
  wfs_cap <- wfs_client$getCapabilities()
  
  bbox_filter <- NULL
  time_filter <- NULL
  location_filter <- NULL
  
  #Building time filter
  start_date <-
    convert_to_date_with_trycatch(start_date, 'start_date')
  end_date <- convert_to_date_with_trycatch(end_date, 'end_date')
  
  if (!is.null(start_date) | !is.null(end_date)) {
    time_filter <- build_cql_time_filter(start_date, end_date)
  }
  
  #Building location filter or bbox filter
  available_countries_pr <-
    listPRPointCountries(printed = FALSE,
               version = version)
  
  if ("all" %in% tolower(c(country, ISO))) {
    location_filter <- NULL
    
    message("Data downloaded for all available locations.")
  } else{
    if (any(c(!is.null(country), !is.null(ISO), !is.null(continent)))) {
      if (!(is.null(country))) {
        location_cql_col <- location_col <- "country"
      } else if (!(is.null(ISO))) {
        location_cql_col <- location_col<- "country_id"
      } else if (!(is.null(continent))) {
        location_cql_col <- "continent_id"
        location_col <- "continent"
      }
      
      checked_availability_pr <-
        isAvailable_pr(
          sourcedata = "pr points",
          country = country,
          ISO = ISO,
          continent = continent,
          full_results = TRUE
        )
      
      message(paste(
        "Attempting to download PR point data for",
        paste(available_countries_pr$country[available_countries_pr[, location_col] %in% checked_availability_pr$location[checked_availability_pr$is_available ==
                                                                                                                                1]], collapse = ", "),
        "..."
      ))
      
      valid_locations <-
        checked_availability_pr$location[checked_availability_pr$is_available ==
                                           1]
      location_filter <-
        build_cql_filter(location_cql_col, valid_locations)
      
    } else if (!is.null(extent)) {
      bbox_filter <- build_cql_bbox_filter(extent)
    }
  }
  
  cql_filters <- list(time_filter, location_filter, bbox_filter)
  cql_filter <- combine_cql_filters(cql_filters)
  
  pf_dataset_id <- getPfPRPointDatasetIdFromVersion(version)
  pv_dataset_id <- getPvPRPointDatasetIdFromVersion(version)
  
  if (tolower(species) == "both") {
    pf_wfs_feature_type <-
      wfs_cap$findFeatureTypeByName(pf_dataset_id)
    pv_wfs_feature_type <-
      wfs_cap$findFeatureTypeByName(pv_dataset_id)
    pf_df <-
      callGetFeaturesWithFilters(pf_wfs_feature_type, cql_filter)
    pv_df <-
      callGetFeaturesWithFilters(pv_wfs_feature_type, cql_filter)
    df <- dplyr::bind_rows(pf_df, pv_df)
    
  } else if (tolower(species) == "pf") {
    wfs_feature_type <-
      wfs_cap$findFeatureTypeByName(pf_dataset_id)
    df <-
      callGetFeaturesWithFilters(wfs_feature_type, cql_filter)
    
  } else if (tolower(species) == "pv") {
    wfs_feature_type <-
      wfs_cap$findFeatureTypeByName(pv_dataset_id)
    df <-
      callGetFeaturesWithFilters(wfs_feature_type, cql_filter)
    
  } else {
    stop("Species not recognized, use one of: \n   \"Pf\" \n   \"Pv\" \n   \"BOTH\"")
  }

  # Just to avoid visible binding notes
  pf_pr <- pv_pr <- permissions_info <- NULL
  
  # removing points that are publicly available but are for the opposite species to what is currently queried.
  if (tolower(species) == "pf") {
    df <-
      dplyr::filter(df, !(is.na(pf_pr) &
                            permissions_info %in% c("", NA)))
  } else if (tolower(species) == "pv") {
    df <-
      dplyr::filter(df, !(is.na(pv_pr) &
                            permissions_info %in% c("", NA)))
  }
  
  if (!nrow(df) > 0 & !is.null(extent)) {
    stop(
      "Error in downloading data for extent: (",
      paste0(extent, collapse = ","),
      "),\n try query using country or continent name."
    )
  }
  
  if (nrow(df) == 0) {
    stop(
      "PR data points are not available for the specified species in requested countries."
    )
  }
  
  if (!(length(country) == 1 && tolower(country) == "all") && any(c(!is.null(country), !is.null(ISO), !is.null(continent)))) {
    message(
      "Data downloaded for ",
      paste(checked_availability_pr$location[checked_availability_pr$is_available == 1], collapse = ", "), "."
    )
  } else if (!is.null(extent)) {
    message("Data downloaded for extent: ",
            paste0(extent, collapse = ","))
  } 
  
  
  
  if (tolower(species) == "both") {
    if (all(is.na(unique(df$pv_pos)))) {
      message(
        "NOTE: All available data for this query was downloaded for both species,\n but there are no PR points for P. vivax in this region in the MAP database. \nTo check endemicity patterns or to contribute data, visit malariaatlas.org OR email us at map@bdi.ox.ac.uk."
      )
    } else if (all(is.na(unique(df$pf_pos)))) {
      message(
        "NOTE: All available data for this query was downloaded for both species,\n but there are no PR points for P. falciparum in this region in the MAP database. \nTo check endemicity patterns or to contribute data, visit malariaatlas.org OR email us at map@bdi.ox.ac.uk."
      )
    }
  }
  
  if (any(grepl("dhs", df$permissions_info))) {
    message(
      "NOTE: Downloaded data includes data points from DHS surveys. \nMAP cannot share DHS survey cluster coordinates, but these are available from www.measuredhs.com, via the rdhs package or using malariaAtlas:fillDHSCoordinates()."
    )
  }
  
  df <- sf::st_drop_geometry(df)
  class(df) <- c("pr.points", class(df))
  
  df <- pr_wide2long(df)
  
  return(df)
  
}


pr_wide2long <- function(object) {
  object$species <- NA
  
  if (all(c("pv_pr", "pf_pr") %in% colnames(object))) {
    pf <-
      rbind(object[is.na(object$pv_pos) &
                     !is.na(object$pf_pos),-which(names(object) %in% grep("pv", names(object), value = T))],
            object[!is.na(object$pv_pos) &
                     !is.na(object$pf_pos),-which(names(object) %in% grep("pv", names(object), value = T))])
    pv <-
      rbind(object[!is.na(object$pv_pos) &
                     is.na(object$pf_pos),-which(names(object) %in% grep("pf", names(object), value = T))],
            object[!is.na(object$pv_pos) &
                     !is.na(object$pf_pos),-which(names(object) %in% grep("pf", names(object), value = T))])
    conf <-
      object[is.na(object$pv_pos) &
               is.na(object$pf_pos),-which(names(object) %in% grep("pf", names(object), value = T))]
    
    names(pf)[which(names(pf) %in% grep("pos", names(pf), value = T))] <-
      "positive"
    names(pv)[which(names(pv) %in% grep("pos", names(pv), value = T))] <-
      "positive"
    names(conf)[which(names(conf) %in% grep("pos", names(conf), value = T))] <-
      "positive"
    names(pf)[which(names(pf) %in% grep("pr", names(pf), value = T))] <-
      "pr"
    names(pv)[which(names(pv) %in% grep("pr", names(pv), value = T))] <-
      "pr"
    names(conf)[which(names(conf) %in% grep("pr", names(conf), value = T))] <-
      "pr"
    
    if (nrow(pf) != 0) {
      pf$species <- "P. falciparum"
    }
    if (nrow(pv) != 0) {
      pv$species <- "P. vivax"
    }
    if (nrow(conf) != 0) {
      conf$species <- "Confidential"
    }
    
    # To avoid no visible bindings. But maybe need to learn tidyeval.
    object <-
      dhs_id <-
      site_id <- site_name <- latitude <-  longitude <- NULL
    rural_urban <-
      country <- country_id <- continent_id <- month_start <- NULL
    year_start <-
      month_end <-
      year_end <- lower_age <- upper_age <- examined <- NULL
    positive <-
      pr <- species <- method <- rdt_type <- pcr_type <- NULL
    malaria_metrics_available <-
      location_available <- permissions_info <- NULL
    citation1 <- citation2 <- citation3 <- NULL
    
    object <- rbind(pf, pv, conf)
    object <-
      dplyr::select(
        object,
        dhs_id,
        site_id,
        site_name,
        latitude,
        longitude,
        rural_urban,
        country,
        country_id,
        continent_id,
        month_start,
        year_start,
        month_end,
        year_end,
        lower_age,
        upper_age,
        examined,
        positive,
        pr,
        species,
        method,
        rdt_type,
        pcr_type,
        malaria_metrics_available,
        location_available,
        permissions_info,
        citation1,
        citation2,
        citation3
      )
    
  } else if ("pv_pr" %in% colnames(object) &
             !("pf_pr" %in% colnames(object))) {
    pv <- object[!is.na(object$pv_pos),]
    conf <- object[is.na(object$pv_pos),]
    
    names(pv)[which(names(pv) %in% grep("pos", names(pv), value = T))] <-
      "positive"
    names(conf)[which(names(conf) %in% grep("pos", names(conf), value = T))] <-
      "positive"
    
    names(pv)[which(names(pv) %in% grep("pr", names(pv), value = T))] <-
      "pr"
    names(conf)[which(names(conf) %in% grep("pr", names(conf), value = T))] <-
      "pr"
    
    if (nrow(pv) != 0) {
      pv$species <- "P. vivax"
    }
    if (nrow(conf) != 0) {
      conf$species <- "Confidential"
    }
    
    object <- rbind(pv, conf)
    object <-
      dplyr::select(
        object,
        dhs_id,
        site_id,
        site_name,
        latitude,
        longitude,
        rural_urban,
        country,
        country_id,
        continent_id,
        month_start,
        year_start,
        month_end,
        year_end,
        lower_age,
        upper_age,
        examined,
        positive,
        pr,
        species,
        method,
        rdt_type,
        pcr_type,
        malaria_metrics_available,
        location_available,
        permissions_info,
        citation1,
        citation2,
        citation3
      )
    
  } else if (!("pv_pr" %in% colnames(object)) &
             "pf_pr" %in% colnames(object)) {
    pf <- object[!is.na(object$pf_pos),]
    conf <- object[is.na(object$pf_pos),]
    
    names(pf)[which(names(pf) %in% grep("pos", names(pf), value = T))] <-
      "positive"
    names(conf)[which(names(conf) %in% grep("pos", names(conf), value = T))] <-
      "positive"
    
    names(pf)[which(names(pf) %in% grep("pr", names(pf), value = T))] <-
      "pr"
    names(conf)[which(names(conf) %in% grep("pr", names(conf), value = T))] <-
      "pr"
    
    if (nrow(pf) != 0) {
      pf$species <- "P. falciparum"
    }
    if (nrow(conf) != 0) {
      conf$species <- "Confidential"
    }
    
    object <- rbind(pf, conf)
    object <-
      dplyr::select(
        object,
        dhs_id,
        site_id,
        site_name,
        latitude,
        longitude,
        rural_urban,
        country,
        country_id,
        continent_id,
        month_start,
        year_start,
        month_end,
        year_end,
        lower_age,
        upper_age,
        examined,
        positive,
        pr,
        species,
        method,
        rdt_type,
        pcr_type,
        malaria_metrics_available,
        location_available,
        permissions_info,
        citation1,
        citation2,
        citation3
      )
    
  }
  
  return(object)
}

#' Convert data.frames to pr.points objects.
#'
#' Will create empty columns for any missing columns expected in a pr.points data.frame.
#' This function is particularly useful for use with packages like dplyr that strip
#' objects of their classes.
#'
#' @param x A data.frame
#'
#'
#' @export
#'
#' @examples
#' #Download PfPR data for Nigeria and Cameroon and map the locations of these points using autoplot
#' \dontrun{
#' library(dplyr)
#' NGA_CMR_PR <- getPR(country = c("Nigeria", "Cameroon"), species = "Pf")
#'
#' # Filter the data frame then readd pr.points class so that autoplot can be used.
#' NGA_CMR_PR %>%
#'   filter(year_start > 2010) %>%
#'   as.pr.points %>%
#'   autoplot
#'
#' }

as.pr.points <- function(x) {
  expected_col_names <-
    c(
      "dhs_id",
      "site_id",
      "site_name",
      "latitude",
      "longitude",
      "rural_urban",
      "country",
      "country_id",
      "continent_id",
      "month_start",
      "year_start",
      "month_end",
      "year_end",
      "lower_age",
      "upper_age",
      "examined",
      "positive",
      "pr",
      "species",
      "method",
      "rdt_type",
      "pcr_type",
      "malaria_metrics_available",
      "location_available",
      "permissions_info",
      "citation1",
      "citation2",
      "citation3"
    )
  
  missing_columns <-
    expected_col_names[!(expected_col_names %in% names(x))]
  
  stopifnot(inherits(x, 'data.frame'))
  
  if (length(missing_columns < 0)) {
    warning('Creating columns of NAs: ',
            paste0(missing_columns, collapse = ', '))
    newcols <-
      data.frame(matrix(NA, ncol = length(missing_columns), nrow = nrow(x)))
    names(newcols) <- missing_columns
    x <- cbind(x, newcols)
  }
  
  class(x) <- c('pr.points', class(x))
  return(x)
}