benchmarks/benchmark.R

In jackwasey/icd: Comorbidity Calculations and Tools for ICD-9 and ICD-10 Codes

# Copyright (C) 2014 - 2018  Jack O. Wasey
#
# This file is part of icd.
#
# icd is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# icd is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with icd. If not, see <http:#www.gnu.org/licenses/>.

# nocov start

slow_tests <- function(...) {
  r <- as.data.frame(res <- devtools::test(reporter = testthat::ListReporter))
  r[order(r$real, decreasing = T), c(1, 3, 11)] %>% head(10)
  invisible(res)
}

my_check <- function(values) {
  sapply(values, function(x) message("dims: ", nrow(x), " by ", ncol(x)))
  sapply(values, function(x) message("digest: ", digest::digest(x)))
  sapply(values, function(x) {
    print(head(x))
    print(tail(x))
  })
  all(sapply(values[-1], function(x) identical(values[[1]], x)))
}

#' benchmark and profile major functions with larger data sets
#'
#' \code{icd9} package is intended to be used with large data sets, with
#' millions or rows. Performance of core functions is therefore of some
#' importance, after correctness. R package test code is for correctness,
#' whereas this script stresses the core functions, and looks for bottlenecks.
#' @keywords internal
icd9_benchmark <- function() {
  # generate large data set: this is copied from test-ICD9.R for now...
  set.seed(1441)
  n <- 1E7 # 10 million rows

  rpts <- generate_random_pts(n)

  tmp <- tempfile(fileext = ".Rprof")
  utils::Rprof(filename = tmp, line.profiling = TRUE, memory.profiling = TRUE)
  utils::capture.output(icd9_comorbid_ahrq(rpts, short_code = TRUE))
  utils::Rprof(NULL)
  utils::summaryRprof(filename = tmp, memory = "both", lines = "show")

  # see how we do scaling up:
  set.seed(1441)
  microbenchmark::microbenchmark(
    comorbid_ahrq.icd9(generate_random_pts(1), short_code = TRUE),
    comorbid_ahrq.icd9(generate_random_pts(10), short_code = TRUE),
    comorbid_ahrq.icd9(generate_random_pts(100), short_code = TRUE),
    comorbid_ahrq.icd9(generate_random_pts(1000), short_code = TRUE),
    # argh, we fall off a cliff between 1000 and 10000 and get much slower.
    comorbid_ahrq.icd9(generate_random_pts(10000), short_code = TRUE),
    comorbid_ahrq.icd9(generate_random_pts(100000), short_code = TRUE),
    times = 5
  )

  tmp <- tempfile(fileext = ".Rprof")
  utils::Rprof(filename = tmp, line.profiling = TRUE, memory.profiling = FALSE)
  utils::capture.output(children.icd9("300" %i9s% "450", short_code = TRUE))
  utils::Rprof(NULL)
  utils::summaryRprof(filename = tmp, lines = "show")

  mydf <- data.frame(
    visitId = c("a", "b", "c"),
    icd9 = c("441", "412.93", "004.0"),
    stringsAsFactors = TRUE
  )
  prof_charl <- profr::profr(charlson(mydf,
    return.df = TRUE,
    stringsAsFactors = TRUE,
    short_code = FALSE
  ))
  ggplot2::ggplot(prof_charl, minlabel = 0.04)

  rng <- "300" %i9s% "450"
  prof_child <- profr::profr(children.icd9(rng, short_code = TRUE))
  ggplot2::ggplot(prof_child, minlabel = 0.001)
  ggplot2::ggsave("tmpggplot.jpg", width = 250, height = 5, dpi = 200, limitsize = FALSE)

  microbenchmark::microbenchmark(
    times = 500, # initial about 2ms
    icd9AddLeadingZeroesMajor(c(1 %i9mj% 999, paste("V", 1:9, sep = "")))
  )

  microbenchmark::microbenchmark(
    times = 500, # initial about 2ms
    icd9AddLeadingZeroesMajor(c(1 %i9mj% 999, paste("V", 1:9, sep = "")))
  )

  # sprintf wins
  microbenchmark::microbenchmark(times = 500000, sprintf("%s%s", "410", "01"))
  microbenchmark::microbenchmark(times = 500000, paste("410", "01", sep = ""))
  rsicd9 <- generate_random_short_icd9()
  microbenchmark::microbenchmark(times = 50, trim(rsicd9))
  microbenchmark::microbenchmark(times = 50, strip(rsicd9))

  # initializing empty data frame
  microbenchmark::microbenchmark(data.frame(matrix(ncol = 2, nrow = 100000)))
  microbenchmark::microbenchmark(data.frame(
    major = character(100000),
    minor = character(100000)
  ))
  # regex is a little faster than fixed
  icd9 <- rep(times = 500, c("1", "not", "V10.0", " E950", ""))
  microbenchmark::microbenchmark(
    times = 3,
    grepl(pattern = "E", icd9, fixed = TRUE) |
      grepl(pattern = "e", icd9, fixed = TRUE) |
      grepl(pattern = "V", icd9, fixed = TRUE) |
      grepl(pattern = "v", icd9, fixed = TRUE)
  )

  microbenchmark::microbenchmark(times = 3, grepl(pattern = "[EeVv]", rpts))
}

# nocov end