Preliminary Benchmarking

[haberdashPI]

Establishes some initial benchmarks for stable_hash speeds. It compares calls of the form sha256(data) to stable_hash(data, alg=sha256).

Results run on my local machine shown here:

 Row │ benchmark   hash       base        trait       ratio      
     │ SubStrin…   SubStrin…  String      String      Float64    
─────┼───────────────────────────────────────────────────────────
   1 │ structs     crc        78.958 μs   98.674 ms   1249.7
   2 │ dataframes  crc        78.959 μs   4.290 ms      54.3383
   3 │ numbers     crc        39.708 μs   2.097 ms      52.8115
   4 │ symbols     crc        597.791 μs  15.910 ms     26.6143
   5 │ strings     crc        597.583 μs  10.400 ms     17.4036
   6 │ tuples      crc        4.204 ms    24.588 ms      5.84912
   7 │ structs     sha256     545.458 μs  168.409 ms   308.747
   8 │ dataframes  sha256     544.750 μs  9.425 ms      17.3024
   9 │ numbers     sha256     271.500 μs  4.652 ms      17.1356
  10 │ symbols     sha256     4.083 ms    28.409 ms      6.9572
  11 │ strings     sha256     4.084 ms    19.677 ms      4.81845
  12 │ tuples      sha256     9.307 ms    41.904 ms      4.50222

You can find a much faster version of the code in the dfl/optimize-1 branch, which I plan to break into a series of more digestable PRs.

[codecov[bot]]

Codecov Report

Merging #28 (0de33e7) into main (ebb9289) will not change coverage. The diff coverage is n/a.

@@           Coverage Diff           @@
##             main      #28   +/-   ##
=======================================
  Coverage   97.26%   97.26%           
=======================================
  Files           1        1           
  Lines         146      146           
=======================================
  Hits          142      142           
  Misses          4        4           

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more