jimhester / hashsize Goto Github PK

library(hashsize)

# Load the tidyverse to populate the symbol table
suppressPackageStartupMessages(library(tidyverse))

buckets <- .Call(hashsize:::hashsize_symbols, 4119L)
lens <- lengths(buckets)
total <- sum(lens)
total
#> [1] 17666
summary(lens)
#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>   0.000   3.000   4.000   4.289   6.000  13.000
hist(lens)

elapsed_time <- function(expr) system.time(expr)[[3]]

time_large_bucket <- function(buckets, n, median, times = 1e8) {
  b <- buckets[lengths(buckets) >= n]
  b <- b[[sample.int(length(b), 1)]]

  front <- as.character(b[[1]])
  back <- as.character(b[[length(b)]])
  med <- as.character(b[[median]])
  list(
    bucket_length = length(b),
    front_name = front,
    front = elapsed_time(.Call(hashsize:::hashsize_lookup_bench, front, times)),
    back_name = back,
    back = elapsed_time(.Call(hashsize:::hashsize_lookup_bench, back, times)),
    med_name = med,
    med = elapsed_time(.Call(hashsize:::hashsize_lookup_bench, med, times)))
}

to_df <- function(x) {
  df <- do.call(rbind.data.frame, x)

  percent <- function(x) sprintf("%.2f%%", x * 100)

  df$total <- percent((df$back - df$front) / df$front)
  df$median <- percent((df$med - df$front) / df$front)
  df
}

knitr::kable(
  to_df(replicate(50, time_large_bucket(buckets, 10, 5, 1e6), simplify = FALSE)))