Jack Andrews wrote:
> i've got many data. how can i deal with a readcsv call that consumes
> all available memory?
> how can i use j to analyse more data than comfortably fits?
> i guess i can partition the data in different ways.
> i could also use another language to parse the csv, row-by-row, and
> invert the table.

You could process the CSV in chunks and dump the parsed data into J
binary data files (one per column) using 3!:4 and 3!:5 (for ints and
floats, strings and other more complicated types would be a little bit
more work) and fappend from 'files', and then map those into memory
using explicit map_jmf_ .

Robby

----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to