group2NA_Benchmark_Option1

gistfile1.Rmd

## Setup

```r
library(knitr)
knitr::opts_chunk$set(echo = TRUE)
## Installing dev versions of plotly with new data.table group2NA()

library(devtools)
library(withr)
library(ggplot2,lib.loc = "/home/Matthew14786/dev/")
# withr::with_libpaths(new = "/home/Matthew14786/dev/",
#                      install_local("/home/Matthew14786/plotly/",repos = NULL,subdir = NULL, force = TRUE))
withr::with_libpaths(new = "/home/Matthew14786/dev/",
                     install_github("msummersgill/plotly",ref = "data.table", force = TRUE))
library(plotly,lib.loc = "/home/Matthew14786/dev/")
library(data.table)
library(stringi)
library(microbenchmark)

## Function to generate sample data

exampleData <- function(RowCount = 10,
                        nGroupsA = 3,
                        nGroupsB = 3,
                        nGroupsC = 3,
                        nIntCols = 3){
  set.seed(1)
  data.frame(
    GroupA = replicate(1,sample(stri_rand_strings(nGroupsA,5),RowCount,rep=TRUE)),
    GroupB = replicate(1,sample(stri_rand_strings(nGroupsB,5),RowCount,rep=TRUE)),
    GroupC = replicate(1,sample(stri_rand_strings(nGroupsC,5),RowCount,rep=TRUE)),
    replicate(nIntCols,sample(0:1e9,RowCount,rep=TRUE))
  )
}

```

## Benchmark Steps

```r

Metrics <- NULL
ptm <- proc.time()

RowPowerStart <- 2
GroupPowerStart <-1

for (nRows in RowPowerStart:7){
  for (nGroups in 1:6){
    if(nRows > nGroups){
      print(cat(paste0("Starting 1e", nRows," rows with \t1e", nGroups, " groups\n")))
      Data <- exampleData(RowCount = 1*10^nRows, nGroupsA = 1*10^nGroups, nGroupsB = 1*10^nGroups, nGroupsC = 1*10^nGroups, nIntCols = 3)
      New <- microbenchmark(
        plotly:::group2NA(Data,"GroupA", retrace.first = TRUE),
        plotly:::group2NA(Data,"GroupA", "GroupB",retrace.first = TRUE),
        plotly:::group2NA(Data,"GroupA", "GroupB","GroupC",retrace.first = TRUE),
        plotly:::group2NA(Data,"GroupA", "GroupB","X1",retrace.first = TRUE),
        plotly:::group2NA(Data,"GroupA", "GroupB",c("X1","X2"),retrace.first = TRUE),
        plotly:::group2NA(Data,"GroupA", "GroupB",c("X1","X2","X3"),retrace.first = TRUE),
        times = 5,
        unit = "s"
      )
      New$expr <- as.character(New$expr)
      New$NumberRows <- 1*10^nRows
      New$NumberGroups <- 1*10^nGroups
      Metrics <- rbind(Metrics,New)
      rm(Data)
      gc()
    }
  }
}

TotalTime <- proc.time() - ptm

MetricsDT <- copy(Metrics)
setDT(MetricsDT)

ComplexityOrder <- c("plotly:::group2NA(Data, \"GroupA\", retrace.first = TRUE)",
                     "plotly:::group2NA(Data, \"GroupA\", \"GroupB\", retrace.first = TRUE)",
                     "plotly:::group2NA(Data, \"GroupA\", \"GroupB\", \"GroupC\", retrace.first = TRUE)",
                     "plotly:::group2NA(Data, \"GroupA\", \"GroupB\", \"X1\", retrace.first = TRUE)",
                     "plotly:::group2NA(Data, \"GroupA\", \"GroupB\", c(\"X1\", \"X2\"), retrace.first = TRUE)",
                     "plotly:::group2NA(Data, \"GroupA\", \"GroupB\", c(\"X1\", \"X2\", \"X3\"),      retrace.first = TRUE)")

MetricsDT[,seconds := time/1e8]
MetricsDT[,expr := ordered(expr, levels = ComplexityOrder)]
MetricsDT[,Sorted := as.factor(ifelse(!grepl("(A|B)\",\\s+retrace", expr),"Sorted","Not Sorted"))]

```

## Results

```r
MetricsDT %>%
  plot_ly %>% 
  add_trace(x = ~jitter(as.integer(as.factor(NumberRows))+RowPowerStart-1),
            y = ~jitter(as.integer(as.factor(NumberGroups))+GroupPowerStart-1),
            z = ~seconds,
            color = ~expr,
            text = ~paste0("Runtime: ", round(seconds,4),"<br>",
                           expr,"<br>",
                           NumberRows, " rows<br>",
                           NumberGroups, " groups<br>"),
            hoverinfo = "text",
            type = "scatter3d",
            mode = "markers",
            marker = list(size = 4,
                          opacity = 0.6)) %>%
  layout(scene = list(
    xaxis = list(
      showgrid = TRUE,
      showspikes = FALSE,
      gridcolor = "rgba(255,255,255,1)",
      showbackground = TRUE,
      backgroundcolor = 'rgba(235,235,235,1)',
      gridwith = 2,
      title = "# Rows",
      tickprefix = "1e+0",
      tickangle = 0,
      tickfont = list(size = 12)
    ),
    yaxis = list(
      showgrid = TRUE,
      showspikes = FALSE,
      showbackground = TRUE,
      backgroundcolor = 'rgba(235,235,235,1)',
      gridcolor = "rgba(255,255,255,1)",
      gridwith = 2,
      title = "# Groups",
      tickprefix = "1e+0"
    ),
    zaxis = list(
      showgrid = TRUE,
      showspikes = FALSE,
      showbackground = TRUE,
      backgroundcolor = 'rgba(235,235,235,1)',
      gridcolor = "rgba(255,255,255,1)",
      gridwith = 2,
      title = "Time",
      ticksuffix = "s"
    )),
    paper_bgcolor = 'rgba(235,235,235,0)',
    plot_bgcolor = 'rgba(235,235,235,0)') 
```