AMR/R/join_microorganisms.R

159 lines
6.3 KiB
R
Raw Normal View History

# ==================================================================== #
# TITLE #
# Antimicrobial Resistance (AMR) Analysis #
# #
2019-01-02 23:24:07 +01:00
# SOURCE #
# https://gitlab.com/msberends/AMR #
# #
# LICENCE #
2019-01-02 23:24:07 +01:00
# (c) 2019 Berends MS (m.s.berends@umcg.nl), Luz CF (c.f.luz@umcg.nl) #
# #
2019-01-02 23:24:07 +01:00
# This R package is free software; you can freely use and distribute #
# it for both personal and commercial purposes under the terms of the #
# GNU General Public License version 2.0 (GNU GPL-2), as published by #
# the Free Software Foundation. #
# #
# This R package was created for academic research and was publicly #
# released in the hope that it will be useful, but it comes WITHOUT #
# ANY WARRANTY OR LIABILITY. #
2019-04-05 18:47:39 +02:00
# Visit our website for more info: https://msberends.gitlab.io/AMR. #
# ==================================================================== #
#' Join a table with \code{microorganisms}
2018-02-21 11:52:31 +01:00
#'
#' Join the dataset \code{\link{microorganisms}} easily to an existing table or character vector.
2018-02-21 11:52:31 +01:00
#' @rdname join
#' @name join
#' @aliases join inner_join
2018-10-12 16:35:18 +02:00
#' @param x existing table to join, or character vector
#' @param by a variable to join by - if left empty will search for a column with class \code{mo} (created with \code{\link{as.mo}}) or will be \code{"mo"} if that column name exists in \code{x}, could otherwise be a column name of \code{x} with values that exist in \code{microorganisms$mo} (like \code{by = "bacteria_id"}), or another column in \code{\link{microorganisms}} (but then it should be named, like \code{by = c("my_genus_species" = "fullname")})
#' @param suffix if there are non-joined duplicate variables in \code{x} and \code{y}, these suffixes will be added to the output to disambiguate them. Should be a character vector of length 2.
2018-02-27 20:01:02 +01:00
#' @param ... other parameters to pass on to \code{dplyr::\link[dplyr]{join}}.
2018-10-12 16:35:18 +02:00
#' @details \strong{Note:} As opposed to the \code{\link[dplyr]{join}} functions of \code{dplyr}, characters vectors are supported and at default existing columns will get a suffix \code{"2"} and the newly joined columns will not get a suffix. See \code{\link[dplyr]{join}} for more information.
2019-01-02 23:24:07 +01:00
#' @inheritSection AMR Read more on our website!
2018-02-21 11:52:31 +01:00
#' @export
2018-04-02 11:11:21 +02:00
#' @examples
2018-10-12 16:35:18 +02:00
#' left_join_microorganisms(as.mo("K. pneumoniae"))
#' left_join_microorganisms("B_KLBSL_PNE")
2018-04-02 11:11:21 +02:00
#'
#' library(dplyr)
#' septic_patients %>% left_join_microorganisms()
2018-04-02 11:11:21 +02:00
#'
2018-02-22 20:48:48 +01:00
#' df <- data.frame(date = seq(from = as.Date("2018-01-01"),
#' to = as.Date("2018-01-07"),
#' by = 1),
2018-10-12 16:35:18 +02:00
#' bacteria = as.mo(c("S. aureus", "MRSA", "MSSA", "STAAUR",
#' "E. coli", "E. coli", "E. coli")),
2018-02-22 20:48:48 +01:00
#' stringsAsFactors = FALSE)
#' colnames(df)
2018-10-12 16:35:18 +02:00
#' df_joined <- left_join_microorganisms(df, "bacteria")
#' colnames(df_joined)
inner_join_microorganisms <- function(x, by = NULL, suffix = c("2", ""), ...) {
checked <- joins_check_df(x, by)
x <- checked$x
by <- checked$by
2018-07-23 14:14:03 +02:00
join <- suppressWarnings(
2018-10-12 16:35:18 +02:00
dplyr::inner_join(x = x, y = AMR::microorganisms, by = by, suffix = suffix, ...)
2018-07-23 14:14:03 +02:00
)
2018-02-21 11:52:31 +01:00
if (nrow(join) > nrow(x)) {
2018-10-12 16:35:18 +02:00
warning('The newly joined tbl contains ', nrow(join) - nrow(x), ' rows more that its original.')
2018-02-21 11:52:31 +01:00
}
join
}
#' @rdname join
#' @export
2018-10-12 16:35:18 +02:00
left_join_microorganisms <- function(x, by = NULL, suffix = c("2", ""), ...) {
checked <- joins_check_df(x, by)
x <- checked$x
by <- checked$by
2018-07-23 14:14:03 +02:00
join <- suppressWarnings(
2018-10-12 16:35:18 +02:00
dplyr::left_join(x = x, y = AMR::microorganisms, by = by, suffix = suffix, ...)
2018-07-23 14:14:03 +02:00
)
2018-02-21 11:52:31 +01:00
if (nrow(join) > nrow(x)) {
2018-10-12 16:35:18 +02:00
warning('The newly joined tbl contains ', nrow(join) - nrow(x), ' rows more that its original.')
2018-02-21 11:52:31 +01:00
}
join
}
#' @rdname join
#' @export
2018-10-12 16:35:18 +02:00
right_join_microorganisms <- function(x, by = NULL, suffix = c("2", ""), ...) {
checked <- joins_check_df(x, by)
x <- checked$x
by <- checked$by
2018-07-23 14:14:03 +02:00
join <- suppressWarnings(
2018-10-12 16:35:18 +02:00
dplyr::right_join(x = x, y = AMR::microorganisms, by = by, suffix = suffix, ...)
2018-07-23 14:14:03 +02:00
)
2018-02-21 11:52:31 +01:00
if (nrow(join) > nrow(x)) {
2018-10-12 16:35:18 +02:00
warning('The newly joined tbl contains ', nrow(join) - nrow(x), ' rows more that its original.')
2018-02-21 11:52:31 +01:00
}
join
}
#' @rdname join
#' @export
2018-10-12 16:35:18 +02:00
full_join_microorganisms <- function(x, by = NULL, suffix = c("2", ""), ...) {
checked <- joins_check_df(x, by)
x <- checked$x
by <- checked$by
2018-07-23 14:14:03 +02:00
join <- suppressWarnings(
2018-10-12 16:35:18 +02:00
dplyr::full_join(x = x, y = AMR::microorganisms, by = by, suffix = suffix, ...)
2018-07-23 14:14:03 +02:00
)
2018-03-27 17:43:42 +02:00
if (nrow(join) > nrow(x)) {
2018-10-12 16:35:18 +02:00
warning('The newly joined tbl contains ', nrow(join) - nrow(x), ' rows more that its original.')
2018-03-27 17:43:42 +02:00
}
join
2018-02-21 11:52:31 +01:00
}
#' @rdname join
#' @export
2018-10-12 16:35:18 +02:00
semi_join_microorganisms <- function(x, by = NULL, ...) {
checked <- joins_check_df(x, by)
x <- checked$x
by <- checked$by
2018-07-23 14:14:03 +02:00
suppressWarnings(
2018-10-12 16:35:18 +02:00
dplyr::semi_join(x = x, y = AMR::microorganisms, by = by, ...)
2018-07-23 14:14:03 +02:00
)
2018-02-21 11:52:31 +01:00
}
#' @rdname join
#' @export
2018-10-12 16:35:18 +02:00
anti_join_microorganisms <- function(x, by = NULL, ...) {
checked <- joins_check_df(x, by)
x <- checked$x
by <- checked$by
suppressWarnings(
dplyr::anti_join(x = x, y = AMR::microorganisms, by = by, ...)
)
}
joins_check_df <- function(x, by) {
if (!any(class(x) %in% c("data.frame", "matrix"))) {
2018-08-31 13:36:19 +02:00
x <- data.frame(mo = as.character(x), stringsAsFactors = FALSE)
2018-10-12 16:35:18 +02:00
if (is.null(by)) {
by <- "mo"
}
}
if (is.null(by)) {
# search for column with class `mo` and return first one found
by <- colnames(x)[lapply(x, is.mo) == TRUE][1]
if (is.na(by)) {
if ("mo" %in% colnames(x)) {
by <- "mo"
} else {
stop("Cannot join - no column found with name or class `mo`.", call. = FALSE)
}
}
message('Joining, by = "', by, '"') # message same as dplyr::join functions
}
2018-02-21 11:52:31 +01:00
if (is.null(names(by))) {
joinby <- colnames(AMR::microorganisms)[1]
2018-02-21 11:52:31 +01:00
names(joinby) <- by
} else {
joinby <- by
}
2018-10-12 16:35:18 +02:00
list(x = x,
by = joinby)
2018-02-21 11:52:31 +01:00
}