While the interactive exploratory aspects of the pandas are
attractive, in my case the interaction has just been a crutch to
discover how to correctly use their api.
Once through that api learning curve, I'd mainly be interested in
repeating the operations that worked correctly. The execution
speed would be more important to me at that point.
In the recent pandas documents, they describe some speed
improvements available from using eval(expression_string) calls
that get executed by a numexpr app. Their testing shows it only
improves execution time when table sizes go beyond about 10k
rows. Seems like this puts the improvements beyond the reach of
my particular app.
ok, thanks. I'll have to dig into it some more.
