[datatable-help] About adding fastmatch and fasttime to data.table

stat quant statquant at outlook.com
Mon Feb 25 19:40:35 CET 2013


Hello list,

Looking at fastmatch and fasttime, I realized that those package consists
solely in 1 C file (each).
We spoke about the possibility to add those to data.table, I tried to
contact S.Urbanek without any success so I do not have feedback from his
side.
Using fastPOSIXct provide a huge gain when one have to load files with
datetime, on my laptop using data.table:::fread, I realized that most of
the time is spent casting datetimes to POSIXct (I have several columns).

Looking at fasttime, you can see pretty good improvement (factor 15)

R) ts <- as.character(.POSIXct(runif(1e6) * unclass(Sys.time())))
R)   system.time(a <- as.POSIXct(ts, "GMT"))
utilisateur     système      écoulé
       6.49        0.04        6.57
R)   system.time(b <- fastPOSIXct(ts, "GMT"))
utilisateur     système      écoulé
       0.40        0.00        0.41

When colClasses will be implemented in fread, can I suggest to allow using
fasttime as an option ?
Concerning fastmatch, the vignette already shows some nice benchmarks, I
tend to do a lot of selects based on string columns, not sure if this is
the case for most of us.

My 0.002 cent
Cheers
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/datatable-help/attachments/20130225/f45e5d57/attachment.html>


More information about the datatable-help mailing list