AMR/R/eucast_rules.R

864 lines
41 KiB
R
Raw Normal View History

2018-02-21 11:52:31 +01:00
# ==================================================================== #
# TITLE #
# Antimicrobial Resistance (AMR) Analysis #
# #
2019-01-02 23:24:07 +01:00
# SOURCE #
# https://gitlab.com/msberends/AMR #
2018-02-21 11:52:31 +01:00
# #
# 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) #
2018-02-21 11:52:31 +01:00
# #
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. #
2018-02-21 11:52:31 +01:00
# ==================================================================== #
2019-04-05 18:47:39 +02:00
# global variables
EUCAST_VERSION_BREAKPOINTS <- "9.0, 2019"
EUCAST_VERSION_EXPERT_RULES <- "3.1, 2016"
2018-10-17 17:32:34 +02:00
#' EUCAST rules
2018-02-21 11:52:31 +01:00
#'
2018-10-17 17:32:34 +02:00
#' Apply susceptibility rules as defined by the European Committee on Antimicrobial Susceptibility Testing (EUCAST, \url{http://eucast.org}), see \emph{Source}. This includes (1) expert rules, (2) intrinsic resistance and (3) inferred resistance as defined in their breakpoint tables.
2019-05-10 16:44:59 +02:00
#' @param x data with antibiotic columns, like e.g. \code{AMX} and \code{AMC}
2018-02-21 11:52:31 +01:00
#' @param info print progress
2018-10-18 12:10:10 +02:00
#' @param rules a character vector that specifies which rules should be applied - one or more of \code{c("breakpoints", "expert", "other", "all")}
#' @param verbose a logical to turn Verbose mode on and off (default is off). In Verbose mode, the function does not apply rules to the data, but instead returns a data set in logbook form with extensive info about which rows and columns would be effected and in which way.
2019-05-10 16:44:59 +02:00
#' @param ... column name of an antibiotic, see section Antibiotics
2018-11-01 20:50:10 +01:00
#' @inheritParams first_isolate
2019-04-05 18:47:39 +02:00
#' @details
2019-05-10 16:44:59 +02:00
#' \strong{Note:} This function does not translate MIC values to RSI values. Use \code{\link{as.rsi}} for that. \cr
#' \strong{Note:} When ampicillin (AMP, J01CA01) is not available but amoxicillin (AMX, J01CA04) is, the latter will be used for all rules where there is a dependency on ampicillin. These drugs are interchangeable when it comes to expression of antimicrobial resistance.
2019-04-09 14:59:17 +02:00
#'
2019-06-01 20:40:49 +02:00
#' The file containing all EUCAST rules is located here: \url{https://gitlab.com/msberends/AMR/blob/master/data-raw/eucast_rules.tsv}.
2019-04-05 18:47:39 +02:00
#'
2018-08-31 13:36:19 +02:00
#' @section Antibiotics:
2019-05-10 16:44:59 +02:00
#' To define antibiotics column names, leave as it is to determine it automatically with \code{\link{guess_ab_col}} or input a text (case-insensitive), or use \code{NULL} to skip a column (e.g. \code{TIC = NULL} to skip ticarcillin). Manually defined but non-existing columns will be skipped with a warning.
2018-12-07 12:04:55 +01:00
#'
#' The following antibiotics are used for the functions \code{\link{eucast_rules}} and \code{\link{mdro}}. These are shown below in the format '\strong{antimicrobial ID}: name (\href{https://www.whocc.no/atc/structure_and_principles/}{ATC code})', sorted by name:
2018-07-26 16:30:42 +02:00
#'
2019-05-10 16:44:59 +02:00
#' \strong{AMK}: amikacin (\href{https://www.whocc.no/atc_ddd_index/?code=J01GB06}{J01GB06}),
#' \strong{AMX}: amoxicillin (\href{https://www.whocc.no/atc_ddd_index/?code=J01CA04}{J01CA04}),
2019-05-29 19:56:17 +02:00
#' \strong{AMC}: amoxicillin/clavulanic acid (\href{https://www.whocc.no/atc_ddd_index/?code=J01CR02}{J01CR02}),
2019-05-10 16:44:59 +02:00
#' \strong{AMP}: ampicillin (\href{https://www.whocc.no/atc_ddd_index/?code=J01CA01}{J01CA01}),
#' \strong{SAM}: ampicillin/sulbactam (\href{https://www.whocc.no/atc_ddd_index/?code=J01CR01}{J01CR01}),
2019-05-10 16:44:59 +02:00
#' \strong{AZM}: azithromycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01FA10}{J01FA10}),
#' \strong{AZL}: azlocillin (\href{https://www.whocc.no/atc_ddd_index/?code=J01CA09}{J01CA09}),
#' \strong{ATM}: aztreonam (\href{https://www.whocc.no/atc_ddd_index/?code=J01DF01}{J01DF01}),
2019-05-29 19:56:17 +02:00
#' \strong{CAP}: capreomycin (\href{https://www.whocc.no/atc_ddd_index/?code=J04AB30}{J04AB30}),
2019-05-10 16:44:59 +02:00
#' \strong{RID}: cefaloridine (\href{https://www.whocc.no/atc_ddd_index/?code=J01DB02}{J01DB02}),
2019-05-29 19:56:17 +02:00
#' \strong{CZO}: cefazolin (\href{https://www.whocc.no/atc_ddd_index/?code=J01DB04}{J01DB04}),
2019-05-10 16:44:59 +02:00
#' \strong{FEP}: cefepime (\href{https://www.whocc.no/atc_ddd_index/?code=J01DE01}{J01DE01}),
#' \strong{CTX}: cefotaxime (\href{https://www.whocc.no/atc_ddd_index/?code=J01DD01}{J01DD01}),
#' \strong{CTT}: cefotetan (\href{https://www.whocc.no/atc_ddd_index/?code=J01DC05}{J01DC05}),
2019-05-10 16:44:59 +02:00
#' \strong{FOX}: cefoxitin (\href{https://www.whocc.no/atc_ddd_index/?code=J01DC01}{J01DC01}),
#' \strong{CPT}: ceftaroline (\href{https://www.whocc.no/atc_ddd_index/?code=J01DI02}{J01DI02}),
2019-05-10 16:44:59 +02:00
#' \strong{CAZ}: ceftazidime (\href{https://www.whocc.no/atc_ddd_index/?code=J01DD02}{J01DD02}),
#' \strong{CRO}: ceftriaxone (\href{https://www.whocc.no/atc_ddd_index/?code=J01DD04}{J01DD04}),
#' \strong{CXM}: cefuroxime (\href{https://www.whocc.no/atc_ddd_index/?code=J01DC02}{J01DC02}),
#' \strong{CED}: cephradine (\href{https://www.whocc.no/atc_ddd_index/?code=J01DB09}{J01DB09}),
2019-05-10 16:44:59 +02:00
#' \strong{CHL}: chloramphenicol (\href{https://www.whocc.no/atc_ddd_index/?code=J01BA01}{J01BA01}),
#' \strong{CIP}: ciprofloxacin (\href{https://www.whocc.no/atc_ddd_index/?code=J01MA02}{J01MA02}),
#' \strong{CLR}: clarithromycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01FA09}{J01FA09}),
#' \strong{CLI}: clindamycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01FF01}{J01FF01}),
#' \strong{COL}: colistin (\href{https://www.whocc.no/atc_ddd_index/?code=J01XB01}{J01XB01}),
#' \strong{DAP}: daptomycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01XX09}{J01XX09}),
#' \strong{DOR}: doripenem (\href{https://www.whocc.no/atc_ddd_index/?code=J01DH04}{J01DH04}),
2019-05-10 16:44:59 +02:00
#' \strong{DOX}: doxycycline (\href{https://www.whocc.no/atc_ddd_index/?code=J01AA02}{J01AA02}),
#' \strong{ETP}: ertapenem (\href{https://www.whocc.no/atc_ddd_index/?code=J01DH03}{J01DH03}),
#' \strong{ERY}: erythromycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01FA01}{J01FA01}),
2019-05-29 19:56:17 +02:00
#' \strong{ETH}: ethambutol (\href{https://www.whocc.no/atc_ddd_index/?code=J04AK02}{J04AK02}),
#' \strong{FLC}: flucloxacillin (\href{https://www.whocc.no/atc_ddd_index/?code=J01CF05}{J01CF05}),
2019-05-10 16:44:59 +02:00
#' \strong{FOS}: fosfomycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01XX01}{J01XX01}),
#' \strong{FUS}: fusidic acid (\href{https://www.whocc.no/atc_ddd_index/?code=J01XC01}{J01XC01}),
2019-05-29 19:56:17 +02:00
#' \strong{GAT}: gatifloxacin (\href{https://www.whocc.no/atc_ddd_index/?code=J01MA16}{J01MA16}),
2019-05-10 16:44:59 +02:00
#' \strong{GEN}: gentamicin (\href{https://www.whocc.no/atc_ddd_index/?code=J01GB03}{J01GB03}),
#' \strong{GEH}: gentamicin-high (no ATC code),
2019-05-10 16:44:59 +02:00
#' \strong{IPM}: imipenem (\href{https://www.whocc.no/atc_ddd_index/?code=J01DH51}{J01DH51}),
2019-05-29 19:56:17 +02:00
#' \strong{INH}: isoniazid (\href{https://www.whocc.no/atc_ddd_index/?code=J04AC01}{J04AC01}),
2019-05-10 16:44:59 +02:00
#' \strong{KAN}: kanamycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01GB04}{J01GB04}),
#' \strong{LVX}: levofloxacin (\href{https://www.whocc.no/atc_ddd_index/?code=J01MA12}{J01MA12}),
#' \strong{LIN}: lincomycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01FF02}{J01FF02}),
#' \strong{LNZ}: linezolid (\href{https://www.whocc.no/atc_ddd_index/?code=J01XX08}{J01XX08}),
#' \strong{MEM}: meropenem (\href{https://www.whocc.no/atc_ddd_index/?code=J01DH02}{J01DH02}),
2019-05-29 19:56:17 +02:00
#' \strong{MTR}: metronidazole (\href{https://www.whocc.no/atc_ddd_index/?code=J01XD01}{J01XD01}),
2019-05-10 16:44:59 +02:00
#' \strong{MEZ}: mezlocillin (\href{https://www.whocc.no/atc_ddd_index/?code=J01CA10}{J01CA10}),
#' \strong{MNO}: minocycline (\href{https://www.whocc.no/atc_ddd_index/?code=J01AA08}{J01AA08}),
#' \strong{MFX}: moxifloxacin (\href{https://www.whocc.no/atc_ddd_index/?code=J01MA14}{J01MA14}),
#' \strong{NAL}: nalidixic acid (\href{https://www.whocc.no/atc_ddd_index/?code=J01MB02}{J01MB02}),
#' \strong{NEO}: neomycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01GB05}{J01GB05}),
#' \strong{NET}: netilmicin (\href{https://www.whocc.no/atc_ddd_index/?code=J01GB07}{J01GB07}),
#' \strong{NIT}: nitrofurantoin (\href{https://www.whocc.no/atc_ddd_index/?code=J01XE01}{J01XE01}),
#' \strong{NOR}: norfloxacin (\href{https://www.whocc.no/atc_ddd_index/?code=J01MA06}{J01MA06}),
#' \strong{NOV}: novobiocin (\href{https://www.whocc.no/atc_ddd_index/?code=QJ01XX95}{QJ01XX95}),
2019-05-10 16:44:59 +02:00
#' \strong{OFX}: ofloxacin (\href{https://www.whocc.no/atc_ddd_index/?code=J01MA01}{J01MA01}),
2019-05-29 19:56:17 +02:00
#' \strong{OXA}: oxacillin (\href{https://www.whocc.no/atc_ddd_index/?code=J01CF04}{J01CF04}),
2019-10-27 00:31:59 +02:00
#' \strong{PEN}: penicillin G (\href{https://www.whocc.no/atc_ddd_index/?code=J01CE01}{J01CE01}),
2019-05-10 16:44:59 +02:00
#' \strong{PIP}: piperacillin (\href{https://www.whocc.no/atc_ddd_index/?code=J01CA12}{J01CA12}),
#' \strong{TZP}: piperacillin/tazobactam (\href{https://www.whocc.no/atc_ddd_index/?code=J01CR05}{J01CR05}),
#' \strong{PLB}: polymyxin B (\href{https://www.whocc.no/atc_ddd_index/?code=J01XB02}{J01XB02}),
#' \strong{PRI}: pristinamycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01FG01}{J01FG01}),
2019-05-29 19:56:17 +02:00
#' \strong{PZA}: pyrazinamide (\href{https://www.whocc.no/atc_ddd_index/?code=J04AK01}{J04AK01}),
2019-05-10 16:44:59 +02:00
#' \strong{QDA}: quinupristin/dalfopristin (\href{https://www.whocc.no/atc_ddd_index/?code=J01FG02}{J01FG02}),
2019-05-29 19:56:17 +02:00
#' \strong{RIB}: rifabutin (\href{https://www.whocc.no/atc_ddd_index/?code=J04AB04}{J04AB04}),
2019-05-10 16:44:59 +02:00
#' \strong{RIF}: rifampicin (\href{https://www.whocc.no/atc_ddd_index/?code=J04AB02}{J04AB02}),
2019-05-29 19:56:17 +02:00
#' \strong{RFP}: rifapentine (\href{https://www.whocc.no/atc_ddd_index/?code=J04AB05}{J04AB05}),
2019-10-27 00:31:59 +02:00
#' \strong{RXT}: roxithromycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01FA06}{J01FA06}),
2019-05-10 16:44:59 +02:00
#' \strong{SIS}: sisomicin (\href{https://www.whocc.no/atc_ddd_index/?code=J01GB08}{J01GB08}),
#' \strong{STH}: streptomycin-high (no ATC code),
2019-05-10 16:44:59 +02:00
#' \strong{TEC}: teicoplanin (\href{https://www.whocc.no/atc_ddd_index/?code=J01XA02}{J01XA02}),
#' \strong{TLV}: telavancin (\href{https://www.whocc.no/atc_ddd_index/?code=J01XA03}{J01XA03}),
2019-05-10 16:44:59 +02:00
#' \strong{TCY}: tetracycline (\href{https://www.whocc.no/atc_ddd_index/?code=J01AA07}{J01AA07}),
#' \strong{TIC}: ticarcillin (\href{https://www.whocc.no/atc_ddd_index/?code=J01CA13}{J01CA13}),
#' \strong{TCC}: ticarcillin/clavulanic acid (\href{https://www.whocc.no/atc_ddd_index/?code=J01CR03}{J01CR03}),
2019-05-10 16:44:59 +02:00
#' \strong{TGC}: tigecycline (\href{https://www.whocc.no/atc_ddd_index/?code=J01AA12}{J01AA12}),
#' \strong{TOB}: tobramycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01GB01}{J01GB01}),
#' \strong{TMP}: trimethoprim (\href{https://www.whocc.no/atc_ddd_index/?code=J01EA01}{J01EA01}),
#' \strong{SXT}: trimethoprim/sulfamethoxazole (\href{https://www.whocc.no/atc_ddd_index/?code=J01EE01}{J01EE01}),
#' \strong{VAN}: vancomycin (\href{https://www.whocc.no/atc_ddd_index/?code=J01XA01}{J01XA01}).
2018-07-26 16:30:42 +02:00
#' @keywords interpretive eucast reading resistance
2018-11-16 20:50:50 +01:00
#' @rdname eucast_rules
2018-02-21 11:52:31 +01:00
#' @export
2019-04-05 18:47:39 +02:00
#' @importFrom dplyr %>% select pull mutate_at vars group_by summarise n
2019-08-20 11:40:54 +02:00
#' @importFrom crayon bold bgGreen bgYellow bgRed black green blue italic strip_style white red make_style
2019-08-06 14:46:30 +02:00
#' @importFrom utils menu
2019-05-23 16:58:59 +02:00
#' @return The input of \code{x}, possibly with edited values of antibiotics. Or, if \code{verbose = TRUE}, a \code{data.frame} with all original and new values of the affected bug-drug combinations.
2018-02-21 11:52:31 +01:00
#' @source
2018-10-17 17:32:34 +02:00
#' \itemize{
#' \item{
#' EUCAST Expert Rules. Version 2.0, 2012. \cr
#' Leclercq et al. \strong{EUCAST expert rules in antimicrobial susceptibility testing.} \emph{Clin Microbiol Infect.} 2013;19(2):141-60. \cr
#' \url{https://doi.org/10.1111/j.1469-0691.2011.03703.x}
#' }
#' \item{
#' EUCAST Expert Rules, Intrinsic Resistance and Exceptional Phenotypes Tables. Version 3.1, 2016. \cr
#' \url{http://www.eucast.org/fileadmin/src/media/PDFs/EUCAST_files/Expert_Rules/Expert_rules_intrinsic_exceptional_V3.1.pdf}
#' }
#' \item{
#' EUCAST Breakpoint tables for interpretation of MICs and zone diameters. Version 9.0, 2019. \cr
#' \url{http://www.eucast.org/fileadmin/src/media/PDFs/EUCAST_files/Breakpoint_tables/v_9.0_Breakpoint_Tables.xlsx}
2018-10-17 17:32:34 +02:00
#' }
#' }
2019-01-02 23:24:07 +01:00
#' @inheritSection AMR Read more on our website!
2018-02-21 11:52:31 +01:00
#' @examples
#' a <- data.frame(mo = c("Staphylococcus aureus",
#' "Enterococcus faecalis",
#' "Escherichia coli",
#' "Klebsiella pneumoniae",
#' "Pseudomonas aeruginosa"),
2019-05-10 16:44:59 +02:00
#' VAN = "-", # Vancomycin
#' AMX = "-", # Amoxicillin
#' COL = "-", # Colistin
#' CAZ = "-", # Ceftazidime
#' CXM = "-", # Cefuroxime
#' PEN = "S", # Penicillin G
#' FOX = "S", # Cefoxitin
2018-02-22 21:37:10 +01:00
#' stringsAsFactors = FALSE)
2018-10-18 12:10:10 +02:00
#'
2018-02-22 21:37:10 +01:00
#' a
2019-05-10 16:44:59 +02:00
#' # mo VAN AMX COL CAZ CXM PEN FOX
2018-10-18 12:10:10 +02:00
#' # 1 Staphylococcus aureus - - - - - S S
#' # 2 Enterococcus faecalis - - - - - S S
#' # 3 Escherichia coli - - - - - S S
#' # 4 Klebsiella pneumoniae - - - - - S S
#' # 5 Pseudomonas aeruginosa - - - - - S S
#'
2019-02-08 16:06:54 +01:00
#'
#' # apply EUCAST rules: 18 results are forced as R or S
#' b <- eucast_rules(a)
2018-04-02 16:05:09 +02:00
#'
2018-02-22 21:37:10 +01:00
#' b
2019-05-10 16:44:59 +02:00
#' # mo VAN AMX COL CAZ CXM PEN FOX
2018-10-18 12:10:10 +02:00
#' # 1 Staphylococcus aureus - S R R S S S
#' # 2 Enterococcus faecalis - - R R R S R
#' # 3 Escherichia coli R - - - - R S
#' # 4 Klebsiella pneumoniae R R - - - R S
#' # 5 Pseudomonas aeruginosa R R - - R R R
2019-02-08 16:06:54 +01:00
#'
#'
#' \donttest{
2019-07-30 13:12:40 +02:00
#' # do not apply EUCAST rules, but rather get a data.frame
2019-02-08 16:06:54 +01:00
#' # with 18 rows, containing all details about the transformations:
#' c <- eucast_rules(a, verbose = TRUE)
#' }
2019-04-05 18:47:39 +02:00
eucast_rules <- function(x,
2018-11-01 20:23:33 +01:00
col_mo = NULL,
2018-02-21 11:52:31 +01:00
info = TRUE,
2018-10-18 12:10:10 +02:00
rules = c("breakpoints", "expert", "other", "all"),
2018-11-01 20:23:33 +01:00
verbose = FALSE,
2019-04-05 18:47:39 +02:00
...) {
if (verbose == TRUE & interactive()) {
txt <- paste0("WARNING: In Verbose mode, the eucast_rules() function does not apply rules to the data, but instead returns a data set in logbook form with extensive info about which rows and columns would be effected and in which way.",
"\n\nThis may overwrite your existing data if you use e.g.:",
"\ndata <- eucast_rules(data, verbose = TRUE)\n\nDo you want to continue?")
if ("rstudioapi" %in% rownames(utils::installed.packages())) {
q_continue <- rstudioapi::showQuestion("Using verbose = TRUE with eucast_rules()", txt)
} else {
q_continue <- menu(choices = c("OK", "Cancel"), graphics = TRUE, title = txt)
}
if (q_continue %in% c(FALSE, 2)) {
2019-08-20 11:40:54 +02:00
message("Cancelled, returning original data")
return(x)
}
}
2019-05-23 16:58:59 +02:00
if (!is.data.frame(x)) {
2019-05-10 16:44:59 +02:00
stop("`x` must be a data frame.", call. = FALSE)
2018-11-01 20:23:33 +01:00
}
2018-11-01 20:23:33 +01:00
# try to find columns based on type
# -- mo
2019-01-15 12:45:24 +01:00
if (is.null(col_mo)) {
2019-05-23 16:58:59 +02:00
col_mo <- search_type_in_df(x = x, type = "mo")
2018-12-22 22:39:34 +01:00
}
if (is.null(col_mo)) {
stop("`col_mo` must be set.", call. = FALSE)
2018-02-21 11:52:31 +01:00
}
2018-10-18 12:10:10 +02:00
if (!all(rules %in% c("breakpoints", "expert", "other", "all"))) {
2019-04-05 18:47:39 +02:00
stop("`rules` must be one or more of: 'breakpoints', 'expert', 'other', 'all'.")
2018-10-18 12:10:10 +02:00
}
2018-11-01 20:23:33 +01:00
if (is.null(col_mo)) {
2019-04-05 18:47:39 +02:00
stop("`col_mo` must be set")
2018-11-01 20:23:33 +01:00
}
decimal.mark <- getOption("OutDec")
big.mark <- ifelse(decimal.mark != ",", ",", ".")
formatnr <- function(x) {
trimws(format(x, big.mark = big.mark, decimal.mark = decimal.mark))
}
2019-08-20 11:40:54 +02:00
grey <- make_style("grey")
2018-10-17 17:32:34 +02:00
warned <- FALSE
2019-10-11 17:21:02 +02:00
txt_error <- function() {
cat("", bgRed(white(" ERROR ")), "\n\n")
}
txt_warning <- function() {
if (warned == FALSE) {
cat("", bgYellow(black(" WARNING ")))
}
warned <<- TRUE
}
txt_ok <- function(no_added, no_changed) {
2018-10-17 17:32:34 +02:00
if (warned == FALSE) {
if (no_added + no_changed == 0) {
2019-08-20 11:40:54 +02:00
cat(pillar::style_subtle(" (no changes)\n"))
2018-10-17 17:32:34 +02:00
} else {
# opening
2019-08-20 11:40:54 +02:00
cat(grey(" ("))
# additions
if (no_added > 0) {
if (no_added == 1) {
2019-08-20 11:40:54 +02:00
cat(green("1 value added"))
} else {
2019-08-20 11:40:54 +02:00
cat(green(formatnr(no_added), "values added"))
}
}
# separator
if (no_added > 0 & no_changed > 0) {
2019-08-20 11:40:54 +02:00
cat(grey(", "))
}
# changes
if (no_changed > 0) {
if (no_changed == 1) {
cat(blue("1 value changed"))
} else {
cat(blue(formatnr(no_changed), "values changed"))
}
}
# closing
2019-08-20 11:40:54 +02:00
cat(grey(")\n"))
2018-10-17 17:32:34 +02:00
}
warned <<- FALSE
}
}
cols_ab <- get_column_abx(x = x,
2019-05-10 16:44:59 +02:00
soft_dependencies = c("AMC",
"AMK",
"AMX",
"AMP",
"AZM",
"AZL",
"ATM",
"RID",
"FEP",
"CTX",
"FOX",
"CED",
"CAZ",
"CRO",
"CXM",
"CHL",
"CIP",
"CLR",
"CLI",
"FLC",
"COL",
"CZO",
"DAP",
"DOX",
"ETP",
"ERY",
"FOS",
"FUS",
"GEN",
"IPM",
"KAN",
"LVX",
"LIN",
"LNZ",
"MEM",
"MEZ",
"MNO",
"MFX",
"NAL",
"NEO",
"NET",
"NIT",
"NOR",
"NOV",
"OFX",
"OXA",
"PEN",
"PIP",
"TZP",
"PLB",
"PRI",
"QDA",
"RIF",
"RXT",
"SIS",
"TEC",
"TCY",
"TIC",
"TGC",
"TOB",
"TMP",
"SXT",
"VAN"),
hard_dependencies = NULL,
2019-05-20 19:12:41 +02:00
verbose = verbose,
...)
2019-10-11 17:21:02 +02:00
AMC <- cols_ab["AMC"]
AMK <- cols_ab["AMK"]
AMP <- cols_ab["AMP"]
AMX <- cols_ab["AMX"]
ATM <- cols_ab["ATM"]
AZL <- cols_ab["AZL"]
AZM <- cols_ab["AZM"]
CAZ <- cols_ab["CAZ"]
CED <- cols_ab["CED"]
CHL <- cols_ab["CHL"]
CIP <- cols_ab["CIP"]
CLI <- cols_ab["CLI"]
CLR <- cols_ab["CLR"]
COL <- cols_ab["COL"]
CRO <- cols_ab["CRO"]
CTX <- cols_ab["CTX"]
CXM <- cols_ab["CXM"]
CZO <- cols_ab["CZO"]
DAP <- cols_ab["DAP"]
DOX <- cols_ab["DOX"]
ERY <- cols_ab["ERY"]
ETP <- cols_ab["ETP"]
FEP <- cols_ab["FEP"]
FLC <- cols_ab["FLC"]
FOS <- cols_ab["FOS"]
FOX <- cols_ab["FOX"]
FUS <- cols_ab["FUS"]
GEN <- cols_ab["GEN"]
IPM <- cols_ab["IPM"]
KAN <- cols_ab["KAN"]
LIN <- cols_ab["LIN"]
LNZ <- cols_ab["LNZ"]
LVX <- cols_ab["LVX"]
MEM <- cols_ab["MEM"]
MEZ <- cols_ab["MEZ"]
MFX <- cols_ab["MFX"]
MNO <- cols_ab["MNO"]
NAL <- cols_ab["NAL"]
NEO <- cols_ab["NEO"]
NET <- cols_ab["NET"]
NIT <- cols_ab["NIT"]
NOR <- cols_ab["NOR"]
NOV <- cols_ab["NOV"]
OFX <- cols_ab["OFX"]
OXA <- cols_ab["OXA"]
PEN <- cols_ab["PEN"]
PIP <- cols_ab["PIP"]
PLB <- cols_ab["PLB"]
PRI <- cols_ab["PRI"]
QDA <- cols_ab["QDA"]
RID <- cols_ab["RID"]
RIF <- cols_ab["RIF"]
RXT <- cols_ab["RXT"]
SIS <- cols_ab["SIS"]
SXT <- cols_ab["SXT"]
TCY <- cols_ab["TCY"]
TEC <- cols_ab["TEC"]
TGC <- cols_ab["TGC"]
TIC <- cols_ab["TIC"]
TMP <- cols_ab["TMP"]
TOB <- cols_ab["TOB"]
TZP <- cols_ab["TZP"]
VAN <- cols_ab["VAN"]
2019-04-05 18:47:39 +02:00
ab_missing <- function(ab) {
all(ab %in% c(NULL, NA))
}
2019-02-08 16:06:54 +01:00
verbose_info <- data.frame(row = integer(0),
col = character(0),
mo_fullname = character(0),
old = as.rsi(character(0)),
new = as.rsi(character(0)),
2019-04-05 18:47:39 +02:00
rule = character(0),
2019-02-08 16:06:54 +01:00
rule_group = character(0),
2019-04-05 18:47:39 +02:00
rule_name = character(0),
2018-10-19 00:17:03 +02:00
stringsAsFactors = FALSE)
# helper function for editing the table
2018-10-17 17:32:34 +02:00
edit_rsi <- function(to, rule, rows, cols) {
2019-01-03 23:56:19 +01:00
cols <- unique(cols[!is.na(cols) & !is.null(cols)])
2018-02-21 11:52:31 +01:00
if (length(rows) > 0 & length(cols) > 0) {
2019-05-23 16:58:59 +02:00
before_df <- x_original
2018-10-17 17:32:34 +02:00
tryCatch(
# insert into original table
2019-05-23 16:58:59 +02:00
x_original[rows, cols] <<- to,
2018-10-17 17:32:34 +02:00
warning = function(w) {
2019-10-11 17:21:02 +02:00
if (w$message %like% "invalid factor level") {
2019-08-07 15:37:39 +02:00
x_original <<- x_original %>% mutate_at(vars(cols), ~factor(x = as.character(.), levels = c(to, levels(.))))
x <<- x %>% mutate_at(vars(cols), ~factor(x = as.character(.), levels = c(to, levels(.))))
x_original[rows, cols] <<- to
2019-10-11 17:21:02 +02:00
warning('Value "', to, '" added to the factor levels of column(s) `', paste(cols, collapse = "`, `"), "` because this value was not an existing factor level.\nA better way is to use as.rsi() on beforehand on antimicrobial columns to guarantee the right structure.", call. = FALSE)
2019-08-07 15:37:39 +02:00
txt_warning()
warned <<- FALSE
2018-10-17 17:32:34 +02:00
} else {
warning(w$message, call. = FALSE)
2019-08-07 15:37:39 +02:00
txt_warning()
cat("\n") # txt_warning() does not append a "\n" on itself
2018-10-17 17:32:34 +02:00
}
},
error = function(e) {
txt_error()
2019-08-08 15:52:07 +02:00
stop(paste0("In row(s) ", paste(rows[1:min(length(rows), 10)], collapse = ","),
ifelse(length(rows) > 10, "...", ""),
2019-10-11 17:21:02 +02:00
" while writing value '", to,
"' to column(s) `", paste(cols, collapse = "`, `"),
"`:\n", e$message),
2019-08-08 15:52:07 +02:00
call. = FALSE)
2018-10-17 17:32:34 +02:00
}
)
2019-08-07 15:37:39 +02:00
tryCatch(
x[rows, cols] <<- x_original[rows, cols],
error = function(e) {
stop(paste0("In row(s) ", paste(rows[1:min(length(rows), 10)], collapse = ","),
2019-10-11 17:21:02 +02:00
"... while writing value '", to,
"' to column(s) `", paste(cols, collapse = "`, `"),
"`:\n", e$message), call. = FALSE)
2019-08-07 15:37:39 +02:00
}
)
# before_df might not be a data.frame, but a tibble or data.table instead
2019-10-11 17:21:02 +02:00
old <- as.data.frame(before_df, stringsAsFactors = FALSE)[rows, ]
track_changes <- list(added = 0,
changed = 0)
2019-10-11 17:21:02 +02:00
for (i in seq_len(length(cols))) {
2019-04-05 18:47:39 +02:00
verbose_new <- data.frame(row = rows,
col = cols[i],
2019-05-23 16:58:59 +02:00
mo_fullname = x[rows, "fullname"],
old = as.rsi(as.character(old[, cols[i]]), warn = FALSE),
new = as.rsi(as.character(x[rows, cols[i]])),
2019-04-05 18:47:39 +02:00
rule = strip_style(rule[1]),
rule_group = strip_style(rule[2]),
rule_name = strip_style(rule[3]),
stringsAsFactors = FALSE)
colnames(verbose_new) <- c("row", "col", "mo_fullname", "old", "new", "rule", "rule_group", "rule_name")
verbose_new <- verbose_new %>% filter(old != new | is.na(old))
# save changes to data set 'verbose_info'
2019-04-05 18:47:39 +02:00
verbose_info <<- rbind(verbose_info, verbose_new)
# count adds and changes
track_changes$added <- track_changes$added + verbose_new %>% filter(is.na(old)) %>% nrow()
track_changes$changed <- track_changes$changed + verbose_new %>% filter(!is.na(old)) %>% nrow()
2018-10-17 17:32:34 +02:00
}
# after the applied changes: return list with counts of added and changed
return(track_changes)
2018-10-17 17:32:34 +02:00
}
# no changes were applied: return number of (new) changes: none.
return(list(added = 0,
changed = 0))
2018-02-21 11:52:31 +01:00
}
2018-10-18 12:10:10 +02:00
# save original table
2019-05-23 16:58:59 +02:00
x_original <- x
2018-10-18 12:10:10 +02:00
# join to microorganisms data set
2019-04-05 18:47:39 +02:00
suppressWarnings(
2019-05-23 16:58:59 +02:00
x <- x %>%
2019-04-05 18:47:39 +02:00
mutate_at(vars(col_mo), as.mo) %>%
left_join_microorganisms(by = col_mo, suffix = c("_oldcols", "")) %>%
mutate(gramstain = mo_gramstain(pull(., col_mo), language = "en"),
genus_species = paste(genus, species)) %>%
as.data.frame(stringsAsFactors = FALSE)
)
2018-10-17 17:32:34 +02:00
if (info == TRUE) {
2019-04-05 18:47:39 +02:00
cat(paste0(
"\nRules by the ", bold("European Committee on Antimicrobial Susceptibility Testing (EUCAST)"),
"\n", blue("http://eucast.org/"), "\n"))
2018-10-17 17:32:34 +02:00
}
# since ampicillin ^= amoxicillin, get the first from the latter (not in original EUCAST table)
2019-05-10 16:44:59 +02:00
if (!ab_missing(AMP) & !ab_missing(AMX)) {
2018-10-19 00:57:10 +02:00
if (verbose == TRUE) {
2019-02-08 16:06:54 +01:00
cat("\n VERBOSE: transforming",
2019-05-23 16:58:59 +02:00
length(which(x[, AMX] == "S" & !x[, AMP] %in% c("S", "I", "R"))),
2019-02-08 16:06:54 +01:00
"empty ampicillin fields to 'S' based on amoxicillin. ")
cat("\n VERBOSE: transforming",
2019-05-23 16:58:59 +02:00
length(which(x[, AMX] == "I" & !x[, AMP] %in% c("S", "I", "R"))),
2019-02-08 16:06:54 +01:00
"empty ampicillin fields to 'I' based on amoxicillin. ")
cat("\n VERBOSE: transforming",
2019-05-23 16:58:59 +02:00
length(which(x[, AMX] == "R" & !x[, AMP] %in% c("S", "I", "R"))),
2019-02-08 16:06:54 +01:00
"empty ampicillin fields to 'R' based on amoxicillin. \n")
2018-10-19 00:57:10 +02:00
}
2019-05-23 16:58:59 +02:00
x[which(x[, AMX] == "S" & !x[, AMP] %in% c("S", "I", "R")), AMP] <- "S"
x[which(x[, AMX] == "I" & !x[, AMP] %in% c("S", "I", "R")), AMP] <- "I"
x[which(x[, AMX] == "R" & !x[, AMP] %in% c("S", "I", "R")), AMP] <- "R"
2019-05-10 16:44:59 +02:00
} else if (ab_missing(AMP) & !ab_missing(AMX)) {
# ampicillin column is missing, but amoxicillin is available
2019-05-10 16:44:59 +02:00
message(blue(paste0("NOTE: Using column `", bold(AMX), "` as input for ampicillin (J01CA01) since many EUCAST rules depend on it.")))
AMP <- AMX
2018-10-17 17:32:34 +02:00
}
2019-10-11 17:21:02 +02:00
# nolint start
# antibiotic classes
2019-05-10 16:44:59 +02:00
aminoglycosides <- c(TOB, GEN, KAN, NEO, NET, SIS)
tetracyclines <- c(DOX, MNO, TCY) # since EUCAST v3.1 tigecycline (TGC) is set apart
polymyxins <- c(PLB, COL)
macrolides <- c(ERY, AZM, RXT, CLR) # since EUCAST v3.1 clinda is set apart
glycopeptides <- c(VAN, TEC)
streptogramins <- c(QDA, PRI) # should officially also be quinupristin/dalfopristin
aminopenicillins <- c(AMP, AMX)
cephalosporins <- c(FEP, CTX, FOX, CED, CAZ, CRO, CXM, CZO)
cephalosporins_without_CAZ <- cephalosporins[cephalosporins != ifelse(is.null(CAZ), "", CAZ)]
carbapenems <- c(ETP, IPM, MEM)
ureidopenicillins <- c(PIP, TZP, AZL, MEZ)
all_betalactams <- c(aminopenicillins, cephalosporins, carbapenems, ureidopenicillins, AMC, OXA, FLC, PEN)
fluoroquinolones <- c(OFX, CIP, NOR, LVX, MFX)
2019-10-11 17:21:02 +02:00
# nolint end
2019-04-05 18:47:39 +02:00
# Help function to get available antibiotic column names ------------------
get_antibiotic_columns <- function(x, df) {
x <- trimws(unlist(strsplit(x, ",", fixed = TRUE)))
y <- character(0)
2019-10-11 17:21:02 +02:00
for (i in seq_len(length(x))) {
2019-08-08 15:52:07 +02:00
if (is.function(get(x[i]))) {
stop("Column ", x[i], " is also a function. Please create an issue on github.com/msberends/AMR/issues.")
}
2019-04-05 18:47:39 +02:00
y <- c(y, tryCatch(get(x[i]), error = function(e) ""))
2018-10-18 12:10:10 +02:00
}
2019-04-05 18:47:39 +02:00
y[y != "" & y %in% colnames(df)]
}
get_antibiotic_names <- function(x) {
x %>%
strsplit(",") %>%
unlist() %>%
trimws() %>%
2019-10-11 17:21:02 +02:00
sapply(function(x) if (x %in% AMR::antibiotics$ab) ab_name(x, language = NULL, tolower = TRUE) else x) %>%
sort() %>%
paste(collapse = ", ")
}
format_antibiotic_names <- function(ab_names, ab_results) {
ab_names <- trimws(unlist(strsplit(ab_names, ",")))
ab_results <- trimws(unlist(strsplit(ab_results, ",")))
if (length(ab_results) == 1) {
if (length(ab_names) == 1) {
# like FOX S
x <- paste(ab_names, "is")
} else if (length(ab_names) == 2) {
# like PEN,FOX S
x <- paste(paste0(ab_names, collapse = " and "), "are both")
} else {
# like PEN,FOX,GEN S (although dependency on > 2 ABx does not exist at the moment)
x <- paste(paste0(ab_names, collapse = " and "), "are all")
}
return(paste0(x, " '", ab_results, "'"))
} else {
if (length(ab_names) == 2) {
# like PEN,FOX S,R
paste0(ab_names[1], " is '", ab_results[1], "' and ",
ab_names[2], " is '", ab_results[2], "'")
} else {
# like PEN,FOX,GEN S,R,R (although dependency on > 2 ABx does not exist at the moment)
paste0(ab_names[1], " is '", ab_results[1], "' and ",
ab_names[2], " is '", ab_results[2], "' and ",
ab_names[3], " is '", ab_results[3], "'")
}
}
}
2019-06-01 20:40:49 +02:00
eucast_rules_df <- eucast_rules_file # internal data file
no_added <- 0
no_changed <- 0
2019-10-11 17:21:02 +02:00
for (i in seq_len(nrow(eucast_rules_df))) {
2019-04-05 18:47:39 +02:00
rule_previous <- eucast_rules_df[max(1, i - 1), "reference.rule"]
rule_current <- eucast_rules_df[i, "reference.rule"]
rule_next <- eucast_rules_df[min(nrow(eucast_rules_df), i + 1), "reference.rule"]
rule_group_previous <- eucast_rules_df[max(1, i - 1), "reference.rule_group"]
rule_group_current <- eucast_rules_df[i, "reference.rule_group"]
if (is.na(eucast_rules_df[i, 4])) {
rule_text <- paste0("always report as '", eucast_rules_df[i, 7], "': ", get_antibiotic_names(eucast_rules_df[i, 6]))
2019-04-05 18:47:39 +02:00
} else {
rule_text <- paste0("report as '", eucast_rules_df[i, 7], "' when ",
format_antibiotic_names(ab_names = get_antibiotic_names(eucast_rules_df[i, 4]),
ab_results = eucast_rules_df[i, 5]), ": ",
get_antibiotic_names(eucast_rules_df[i, 6]))
2018-10-18 12:10:10 +02:00
}
2019-04-05 18:47:39 +02:00
if (i == 1) {
rule_previous <- ""
rule_group_previous <- ""
2018-10-18 12:10:10 +02:00
}
2019-04-05 18:47:39 +02:00
if (i == nrow(eucast_rules_df)) {
rule_next <- ""
2018-10-18 12:10:10 +02:00
}
2019-04-05 18:47:39 +02:00
# don't apply rules if user doesn't want to apply them
if (rule_group_current %like% "breakpoint" & !any(c("all", "breakpoints") %in% rules)) {
next
2018-10-18 12:10:10 +02:00
}
2019-04-05 18:47:39 +02:00
if (rule_group_current %like% "expert" & !any(c("all", "expert") %in% rules)) {
next
2018-10-18 12:10:10 +02:00
}
2019-04-05 18:47:39 +02:00
if (rule_group_current %like% "other" & !any(c("all", "other") %in% rules)) {
next
2018-10-18 12:10:10 +02:00
}
2018-10-18 12:10:10 +02:00
if (info == TRUE) {
2019-04-05 18:47:39 +02:00
# Print rule (group) ------------------------------------------------------
if (rule_group_current != rule_group_previous) {
# is new rule group, one of Breakpoints, Expert Rules and Other
cat(bold(
case_when(
rule_group_current %like% "breakpoint" ~
paste0("\nEUCAST Clinical Breakpoints (v", EUCAST_VERSION_BREAKPOINTS, ")\n"),
rule_group_current %like% "expert" ~
paste0("\nEUCAST Expert Rules, Intrinsic Resistance and Exceptional Phenotypes (v", EUCAST_VERSION_EXPERT_RULES, ")\n"),
TRUE ~
"\nOther rules\n"
)
))
}
# Print rule -------------------------------------------------------------
if (rule_current != rule_previous) {
# is new rule within group, print its name
if (rule_current %in% c(AMR::microorganisms$family,
AMR::microorganisms$fullname)) {
cat(italic(rule_current))
} else {
cat(rule_current)
}
warned <- FALSE
}
2018-10-18 12:10:10 +02:00
}
2019-04-05 18:47:39 +02:00
# Get rule from file ------------------------------------------------------
col_mo_property <- eucast_rules_df[i, 1]
like_is_one_of <- eucast_rules_df[i, 2]
2019-04-05 18:47:39 +02:00
# be sure to comprise all coagulase-negative/-positive Staphylococci when they are mentioned
if (eucast_rules_df[i, 3] %like% "coagulase-") {
suppressWarnings(
all_staph <- AMR::microorganisms %>%
filter(genus == "Staphylococcus") %>%
mutate(CNS_CPS = mo_name(mo, Becker = "all"))
2019-04-05 18:47:39 +02:00
)
if (eucast_rules_df[i, 3] %like% "coagulase-") {
eucast_rules_df[i, 3] <- paste0("^(",
paste0(all_staph %>%
filter(CNS_CPS %like% "coagulase-negative") %>%
pull(fullname),
collapse = "|"),
")$")
} else {
eucast_rules_df[i, 3] <- paste0("^(",
paste0(all_staph %>%
filter(CNS_CPS %like% "coagulase-positive") %>%
pull(fullname),
collapse = "|"),
")$")
}
like_is_one_of <- "like"
}
2019-04-05 18:47:39 +02:00
if (like_is_one_of == "is") {
mo_value <- paste0("^", eucast_rules_df[i, 3], "$")
} else if (like_is_one_of == "one_of") {
2019-10-11 17:21:02 +02:00
# so 'Clostridium, Actinomyces, ...' will turn into '^(Clostridium|Actinomyces|...)$'
2019-04-05 18:47:39 +02:00
mo_value <- paste0("^(",
paste(trimws(unlist(strsplit(eucast_rules_df[i, 3], ",", fixed = TRUE))),
collapse = "|"),
")$")
} else if (like_is_one_of == "like") {
mo_value <- eucast_rules_df[i, 3]
} else {
stop("invalid like_is_one_of", call. = FALSE)
2018-10-18 12:10:10 +02:00
}
2019-04-05 18:47:39 +02:00
source_antibiotics <- eucast_rules_df[i, 4]
source_value <- trimws(unlist(strsplit(eucast_rules_df[i, 5], ",", fixed = TRUE)))
target_antibiotics <- eucast_rules_df[i, 6]
target_value <- eucast_rules_df[i, 7]
2019-04-05 18:47:39 +02:00
if (is.na(source_antibiotics)) {
2019-05-23 16:58:59 +02:00
rows <- tryCatch(which(x[, col_mo_property] %like% mo_value),
2019-04-05 18:47:39 +02:00
error = function(e) integer(0))
} else {
2019-05-23 16:58:59 +02:00
source_antibiotics <- get_antibiotic_columns(source_antibiotics, x)
2019-04-05 18:47:39 +02:00
if (length(source_value) == 1 & length(source_antibiotics) > 1) {
source_value <- rep(source_value, length(source_antibiotics))
}
if (length(source_antibiotics) == 0) {
rows <- integer(0)
} else if (length(source_antibiotics) == 1) {
2019-05-23 16:58:59 +02:00
rows <- tryCatch(which(x[, col_mo_property] %like% mo_value
& x[, source_antibiotics[1L]] == source_value[1L]),
2019-04-05 18:47:39 +02:00
error = function(e) integer(0))
} else if (length(source_antibiotics) == 2) {
2019-05-23 16:58:59 +02:00
rows <- tryCatch(which(x[, col_mo_property] %like% mo_value
& x[, source_antibiotics[1L]] == source_value[1L]
& x[, source_antibiotics[2L]] == source_value[2L]),
2019-04-05 18:47:39 +02:00
error = function(e) integer(0))
} else if (length(source_antibiotics) == 3) {
2019-05-23 16:58:59 +02:00
rows <- tryCatch(which(x[, col_mo_property] %like% mo_value
& x[, source_antibiotics[1L]] == source_value[1L]
& x[, source_antibiotics[2L]] == source_value[2L]
& x[, source_antibiotics[3L]] == source_value[3L]),
2019-04-05 18:47:39 +02:00
error = function(e) integer(0))
} else {
stop("only 3 antibiotics supported for source_antibiotics ", call. = FALSE)
}
2018-10-18 12:10:10 +02:00
}
2019-05-23 16:58:59 +02:00
cols <- get_antibiotic_columns(target_antibiotics, x)
2019-04-05 18:47:39 +02:00
# Apply rule on data ------------------------------------------------------
# this will return the unique number of changes
run_changes <- edit_rsi(to = target_value,
rule = c(rule_text, rule_group_current, rule_current),
rows = rows,
cols = cols)
no_added <- no_added + run_changes$added
no_changed <- no_changed + run_changes$changed
2019-04-05 18:47:39 +02:00
# Print number of new changes ---------------------------------------------
if (info == TRUE & rule_next != rule_current) {
# print only on last one of rules in this group
txt_ok(no_added = no_added, no_changed = no_changed)
# and reset counters
no_added <- 0
no_changed <- 0
2018-11-01 17:06:08 +01:00
}
2019-04-05 18:47:39 +02:00
}
2019-04-05 18:47:39 +02:00
# Print overview ----------------------------------------------------------
2018-02-21 11:52:31 +01:00
if (info == TRUE) {
2018-10-19 00:57:10 +02:00
if (verbose == TRUE) {
wouldve <- "would have "
} else {
wouldve <- ""
}
2019-04-05 18:47:39 +02:00
verbose_info <- verbose_info %>%
arrange(row, rule_group, rule_name, col)
2019-08-20 11:40:54 +02:00
cat(paste0("\n", grey(strrep("-", options()$width - 1)), "\n"))
2019-10-11 17:21:02 +02:00
cat(bold(paste("EUCAST rules", paste0(wouldve, "affected"),
2019-04-05 18:47:39 +02:00
formatnr(n_distinct(verbose_info$row)),
2019-10-11 17:21:02 +02:00
"out of", formatnr(nrow(x_original)),
"rows, making a total of", formatnr(nrow(verbose_info)), "edits\n")))
n_added <- verbose_info %>% filter(is.na(old)) %>% nrow()
n_changed <- verbose_info %>% filter(!is.na(old)) %>% nrow()
2019-04-05 18:47:39 +02:00
# print added values ----
if (n_added == 0) {
2019-04-05 18:47:39 +02:00
colour <- cat # is function
} else {
2019-08-20 11:40:54 +02:00
colour <- green # is function
2019-02-08 16:06:54 +01:00
}
2019-04-05 18:47:39 +02:00
cat(colour(paste0("=> ", wouldve, "added ",
bold(formatnr(verbose_info %>%
filter(is.na(old)) %>%
nrow()), "test results"),
2019-02-08 16:06:54 +01:00
"\n")))
if (n_added > 0) {
2019-04-05 18:47:39 +02:00
verbose_info %>%
filter(is.na(old)) %>%
group_by(new) %>%
summarise(n = n()) %>%
mutate(plural = ifelse(n > 1, "s", ""),
txt = paste0(formatnr(n), " test result", plural, " added as ", new)) %>%
pull(txt) %>%
2019-08-08 15:52:07 +02:00
paste(" -", ., collapse = "\n") %>%
2019-04-05 18:47:39 +02:00
cat()
}
2019-04-05 18:47:39 +02:00
# print changed values ----
if (n_changed == 0) {
2019-04-05 18:47:39 +02:00
colour <- cat # is function
} else {
colour <- blue # is function
}
if (n_added + n_changed > 0) {
cat("\n")
}
cat(colour(paste0("=> ", wouldve, "changed ",
2019-04-05 18:47:39 +02:00
bold(formatnr(verbose_info %>%
filter(!is.na(old)) %>%
nrow()), "test results"),
2019-02-08 16:06:54 +01:00
"\n")))
if (n_changed > 0) {
2019-04-05 18:47:39 +02:00
verbose_info %>%
filter(!is.na(old)) %>%
group_by(old, new) %>%
summarise(n = n()) %>%
mutate(plural = ifelse(n > 1, "s", ""),
txt = paste0(formatnr(n), " test result", plural, " changed from ", old, " to ", new)) %>%
pull(txt) %>%
2019-08-08 15:52:07 +02:00
paste(" -", ., collapse = "\n") %>%
2019-04-05 18:47:39 +02:00
cat()
cat("\n")
}
2019-08-20 11:40:54 +02:00
cat(paste0(grey(strrep("-", options()$width - 1)), "\n"))
2019-04-05 18:47:39 +02:00
if (verbose == FALSE & nrow(verbose_info) > 0) {
2019-08-07 15:37:39 +02:00
cat(paste("\nUse", bold("eucast_rules(..., verbose = TRUE)"), "(on your original data) to get a data.frame with all specified edits instead.\n\n"))
} else if (verbose == TRUE) {
2019-08-08 15:52:07 +02:00
cat(paste0("\nUsed 'Verbose mode' (", bold("verbose = TRUE"), "), which returns a data.frame with all specified edits.\nUse ", bold("verbose = FALSE"), " to apply the rules on your data.\n\n"))
2019-03-28 21:33:28 +01:00
}
2018-10-17 17:32:34 +02:00
}
2019-04-05 18:47:39 +02:00
# Return data set ---------------------------------------------------------
2018-10-19 00:17:03 +02:00
if (verbose == TRUE) {
2019-04-05 18:47:39 +02:00
verbose_info
} else {
2019-05-23 16:58:59 +02:00
x_original
2018-02-21 11:52:31 +01:00
}
}