[Rd] gc(reset=TRUE) reset timing

From: Stavros Macrakis <macrakis_at_alum.mit.edu>
Date: Wed, 11 Feb 2009 11:53:50 -0500

The man page for gc reads:

     The final two columns show the maximum space used since the last
     call to 'gc(reset=TRUE)' (or since R started).

The word 'last' here is ambiguous: does it include the *current* call to gc? When I first read this, I assumed that it did not; indeed, I only realized that there was an ambiguity after trying to use reset=TRUE for some measurements and discovering that the implementation corresponds to a reading where it *does* include the current call to gc. (see below for transcript)

The interpretation that 'last' does *not* include the current call seems more useful than the implemented behavior, since with the implemented behavior, to report maximum memory usage for each of a series of operations, you need to make *2* calls to GC between operations, one with reset=FALSE (to return the maximum space used) and one with reset=TRUE (to perform the reset for the next measurement).

So I suggest that the reset happen only *after* the "max used" calculation. Alternatively, the documentation could be clarified.


{gc(reset=T); numeric(10^7); gc(reset=FALSE)["Vcells","max used"]}

[1] 10353316                        <<< shows max used by numeric

> {gc(reset=T); numeric(10^7); gc(reset=TRUE)["Vcells","max used"]}
[1] 353329 <<< does not show max used by numeric ______________________________________________
R-devel_at_r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel Received on Wed 11 Feb 2009 - 16:14:45 GMT

Archive maintained by Robert King, hosted by the discipline of statistics at the University of Newcastle, Australia.
Archive generated by hypermail 2.2.0, at Wed 11 Feb 2009 - 16:30:21 GMT.

Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-devel. Please read the posting guide before posting to the list.

list of date sections of archive