filter_obs.R

#' @title Filter observation list to exclude situations, variables or dates
#'
#' @inheritParams estim_param
#' @param var (optional, if not given all variables will be kept) Vector containing the names of the variables to include or exclude
#' @param situation (optional, if not given all situations will be kept) Vector containing the names of the situations to include or exclude
#' @param dates (optional, if not given all dates will be kept) Vector containing the dates (POSIXct format) to include or exclude
#' @param include (optional, FALSE by default) Flag indicating if the variables / situations / dates listed in inputs must be included (TRUE) or not (FALSE) in the resulting list
#' @param var_names `r lifecycle::badge("deprecated")` `var_names` is no
#'   longer supported, use `var` instead.
#' @param sit_names `r lifecycle::badge("deprecated")` `sit_names` is no
#'   longer supported, use `situation` instead.
#'
#' @return obs_list List of filtered observed values (same format as `obs_list` input argument)
#'
#' @seealso For more detail and examples, see the different vignettes in
#' [CroptimizR website](https://sticsrpacks.github.io/CroptimizR/)
#'
#' @export
#'
#' @importFrom rlang .data
#'
#' @examples
#'
#' obs_list <- list(
#'   sit1 = data.frame(
#'     Date = as.POSIXct(c("2009-11-30", "2009-12-10")),
#'     var1 = c(1.1, 1.5), var2 = c(NA, 2.1)
#'   ),
#'   sit2 = data.frame(
#'     Date = as.POSIXct(c("2009-11-30", "2009-12-5")),
#'     var1 = c(1.3, 2)
#'   )
#' )
#'
#' # Keep only var1
#' filter_obs(obs_list, var = c("var1"), include = TRUE)
#'
#' # Exclude observations at date "2009-11-30"
#' filter_obs(obs_list, dates = as.POSIXct(c("2009-11-30")))
#'
filter_obs <- function(obs_list, var = NULL, situation = NULL, dates = NULL,
                       include = FALSE, var_names = lifecycle::deprecated(),
                       sit_names = lifecycle::deprecated()) {

  # Managing parameter names changes between versions:
  if (lifecycle::is_present(sit_names)) {
    lifecycle::deprecate_warn("0.5.0", "filter_obs(sit_names)", "filter_obs(situation)")
  } else {
    sit_names <- situation # to remove when we update inside the function
  }
  if (lifecycle::is_present(var_names)) {
    lifecycle::deprecate_warn("0.5.0", "filter_obs(var_names)", "filter_obs(var)")
  } else {
    var_names <- var # to remove when we update inside the function
  }

  # Check obs_list format
  if (is.null(obs_list)) {
    stop("obs_list is NULL.")
  }
  if (!is.obs(obs_list)) {
    stop("Incorrect format for argument obs_list.")
  }

  # Filter Situations
  ## check that sit_names are in obs_list
  if (!is.null(sit_names)) {
    tmp <- intersect(sit_names, names(obs_list))
    if (is.null(tmp) || !setequal(tmp, sit_names)) {
      warning("Argument sit_names contains situations that are not included in obs_list. \n obs_list contains: ", paste(names(obs_list), collapse = " "))
      sit_names <- tmp
    }
    ## Filter
    if (include) {
      obs_list <- obs_list[sit_names]
    } else {
      obs_list[sit_names] <- NULL
      if (length(obs_list) == 0) {
        warning("All situations have been excluded from the list")
        return(NULL)
      }
    }
  }


  # Transform obs_list in a data.frame for easier filtering of var and dates
  df <- dplyr::bind_rows(obs_list, .id = "id")

  # Filter Variables
  ## check that var_names are in obs_list
  if (!is.null(var_names)) {
    tmp <- intersect(var_names, names(df))
    if (is.null(tmp) || !setequal(tmp, var_names)) {
      warning("Argument var_names contains variables that are not included in obs_list. \n obs_list contains: ", paste(colnames(df), collapse = " "))
      if (length(tmp) == 0) {
        return(list())
      } # If variable does not exist at all, return empty df
      var_names <- tmp
    }
    ## Filter
    if (include) {
      keep <- c("id", "Date", intersect("Plant", names(df)))
      df <- df[, c(keep, var_names)]
    } else {
      df[var_names] <- NULL
      if (ncol(df) == 2) {
        warning("All variables have been excluded from the list")
        return(NULL)
      }
    }
  }

  # Filter Dates
  ## check that dates are in obs_list
  if (!is.null(dates)) {
    included <- sapply(dates, function(x) any(df$Date == x))
    if (!all(c = included)) {
      warning("Argument dates contains dates that are not included in obs_list: ", paste(dates[!included], collapse = " "))
      dates <- dates[included]
    }
    ## Filter
    if (include) {
      df <- dplyr::filter(df, .data$Date == dates)
    } else {
      df <- dplyr::filter(df, .data$Date != dates)
      if (nrow(df) == 0) {
        warning("All dates have been excluded from the list.")
        return(NULL)
      }
    }
  }

  if ("Plant" %in% colnames(df)) {
    skeepcols <- 3
  } else {
    skeepcols <- 2
  }

  # Remove rows with only NAs:
  df <- df[rowSums(is.na(df[, (skeepcols + 1):ncol(df), drop = FALSE])) != (ncol(df) - skeepcols), ]

  if (!all(names(obs_list) %in% unique(df$id))) {
    warning(
      "No observations found in situation(s) ",
      paste(names(obs_list)[!(names(obs_list) %in% unique(df$id))], collapse = ", ")
    )
  }

  # Re-transform the df into a list
  obs_list <- split(df, df$id)

  # Remove column "id" and remove columns with only NAs:
  obs_list <- lapply(
    obs_list,
    function(x) {
      select(x, !.data$id & where(~ !all(is.na(.x))))
    }
  )

  return(obs_list)
}

# Remove when tidyselect exports where() (very soon),
# see https://github.com/r-lib/tidyselect/issues/244
utils::globalVariables("where")