[Rd] astonishing memory phenomenon

Kurt Hornik Kurt.Hornik@ci.tuwien.ac.at
Sat, 12 Feb 2000 15:22:27 +0100 (CET)

>>>>> =?iso-8859-1?Q?Jens Oehlschl=E4gel-Akiyoshi?= writes:

> Thanks for the explanation Brian!

>> You only have one character data object, of length 0, 4096, 8192,>and lots
> of references to it.
>> You don't need multiple copies of the same data.

> This (definitely clever) behaviour of R makes it more difficult to decide,
> HOW to import large amounts of data, if the data type is DATE. Dates could
> be imported as CHARACTER or as DOUBLE (chron-objects).

> Memory requirements of chron would be approx   n*8 Bytes
> Memory requirements of character dates like "01/01/2000" would be n*4 + u*10
> Memory requirements of character times like "23:03:20" would be n*4 + u*8
> Memory requirements of character date-times like "01/01/2000 23:03:20" would
> be n*4 + u*19
> where n=no of cases and u=no of unique values

> Concerning date-times from

>   n*8 < n*4 + u*19

> we get

>   n/u < 4.75

> as a condition for the chron object beeing smaller than a character
> representation, thus the answer is: IT DEPENDS.

> BTW: Is there any specific reason that currently we cannot have chron
> objects as columns of data.frames?

Currently, chron objects are stored as numeric vectors with certain
attributes (such as `origin').  I don't see how you could keep the
attributes ...

If I ever get to it, there may be a future version of chron based on a
struct tm like representation.

r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch