Skip to content

Latest commit

 

History

History
164 lines (118 loc) · 4.98 KB

README.md

File metadata and controls

164 lines (118 loc) · 4.98 KB

fread-benchmarks-rsuite

Introduction

This is the rsuite version of the original tests published by Jozef Hajnala in Gitlab.

It is conceptually the same scripts but adding these changes:

  1. a new R script using awk: 06_fread_awk.R
  2. modification of the project structure using the rsuite paradigm
  3. addition of comments on scripts
  4. generation of results dataframe: show_results.R
  5. plots for memory and average time per operation
  6. packages are all installed exclusively for this project, under the folder packages, insulated from the global environment

Data

The data is the same used in the original project (airlines), CSV files located here. The files are downloaded compressed as bunzip, then expanded as csv.

Running the benchmarks

For base R

bash bench.sh R/01_base.R &> results/out_base.txt

For data.table::fread

bash bench.sh R/02_fread.R &> results/out_fread.txt

For readr::read_csv

bash bench.sh R/03_readr.R &> results/out_readr.txt

For data.table::fread with grep

bash bench.sh R/04_fread_grep.R &> results/out_fread_grep.txt

For readr::read_csv with grep

bash bench.sh R/05_readr_grep.R &> results/out_readr_grep.txt

For data.table::fread with awk

bash bench.sh R/06_fread_awk.R &> results/out_fread_awk.txt

Results

Tables

library(dplyr)

source("R/show_results.R")

df <- get_results_df() %>% 
    arrange(-mem_gb) %>% 
    select(result_file, rscript, mem_gb, avg_secs) %>% 
    mutate_if(is.numeric, round, digits = 2) %>% 
    print()
#>          result_file            rscript mem_gb avg_secs
#> 1      out_readr.txt      R/03_readr.R   27.01   121.57
#> 2       out_base.txt       R/01_base.R   21.93   293.52
#> 3      out_fread.txt      R/02_fread.R   15.24    26.86
#> 4 out_readr_grep.txt R/05_readr_grep.R    1.64    28.19
#> 5 out_fread_grep.txt R/04_fread_grep.R    1.62    10.01
#> 6  out_fread_awk.txt  R/06_fread_awk.R    1.47    25.60

This is the same table but including the original description for the scripts.

#>                               description            rscript mem_gb
#> 1     `readr::read_csv` + `purrr:map_dfr`      R/03_readr.R   27.01
#> 2       `utils::read.csv` + `base::rbind`       R/01_base.R   21.93
#> 3       `data.table::fread` + `rbindlist`      R/02_fread.R   15.24
#> 4 `readr::read_csv`+ `pipe()` from `grep` R/05_readr_grep.R    1.64
#> 5         `data.table::fread` from `grep` R/04_fread_grep.R    1.62
#> 6          `data.table::fread` from `awk`  R/06_fread_awk.R    1.47
#>   avg_secs
#> 1   121.57
#> 2   293.52
#> 3    26.86
#> 4    28.19
#> 5    10.01
#> 6    25.60
  • mem_gb = Maximum resident set size, gigabytes
  • avg_secs = Average of real time and user time as measured by time, seconds

Memory required (GB)

library(ggplot2)
library(scales)

df$rscript <- reorder(df$rscript, df$mem_gb)
ggplot(df, aes(x = rscript, y = mem_gb)) +
    geom_col() +
    scale_y_continuous(breaks=pretty_breaks(n=20)) +
    coord_flip()

Average Time (seconds)

df$rscript <- reorder(df$rscript, df$avg_secs)
ggplot(df, aes(x = rscript, y = avg_secs)) +
    geom_col() +
    scale_y_continuous(breaks=pretty_breaks(n=20)) +
    coord_flip()

How to reproduce this project yourself

  1. Download and install the RSuite client. Available for Linux, Mac and Windows.
  2. Install the rsuite package with rsuite install
  3. Clone or download this repository.
  4. Change to this repo folder and install the dependencies on its own isolated reproducible environment. Use rsuite proj depsinst
  5. Build the project with rsuite proj build
  6. Download the data running this from the console Rscript R/data_prep.R
  7. Run each of the tests. Example: bash bench.sh R/01_base.R &> results/out_base.txt. See above for the rest.
  8. Generate a comparative table with the results running Rscript R/show_results.R

References