v0.22
This release has improved performance for large files.
CSVRead
execution time is reduced by approximately 35% for large files in this release. This results from switching from using Scripting.TextStream
to using ADODB.Stream
for the low-level task of reading data from disk into a string variable (and working around the performance shortcoming of the latter). The relevant method is ReadAllFromStream
.
The largest file in RDatasets is 76Mb in size and has 1.4 million rows and seven columns. Time to execute CSVRead("C:\Projects\RDatasets\csv\openintro\military.csv")
falls from 22 seconds using v0.4 to 14 seconds using v0.22.