[Rd] pmax and long vector

Martin Maechler m@ech|er @end|ng |rom @t@t@m@th@ethz@ch
Tue Jan 22 18:56:56 CET 2019


>>>>> Kasper Daniel Hansen 
>>>>>     on Mon, 21 Jan 2019 21:51:55 -0500 writes:

    > Gabe, I don't (yet) know much about long vectors at the C level. So feel
    > free to address this.

    > Duncan, I'll see what I can do regarding systematically compiling a list of
    > functions without long vector support. These days I frequently work with
    > big enough matrices that I need it.

Thank you, Kasper, Duncan and Gabriel!

I agree with Duncan about  "probably be useful to R Core".

Best,
Martin 


    > On Mon, Jan 21, 2019 at 3:09 PM Gabriel Becker <gabembecker using gmail.com>
    > wrote:

    >> Kasper,
    >> 
    >> If you're not interested or dont have time to create said patch yourself
    >> let me know and i can do it.
    >> 
    >> Best,
    >> ~G
    >> 
    >> On Mon, Jan 21, 2019, 11:36 AM Duncan Murdoch <murdoch.duncan using gmail.com
    >> wrote:
    >> 
    >>> On 21/01/2019 12:35 p.m., Kasper Daniel Hansen wrote:
    >>> > I see that base::pmax() does not support long vectors.
    >>> >
    >>> > Is R-devel interested in reports like this; ie. is there a goal of full
    >>> > support for long vectors in "basic" functions, something I at least
    >>> would
    >>> > greatly appreciate?
    >>> >
    >>> > MRE:
    >>> >
    >>> >> pmax(rep(1L, 3*10^9), 0)
    >>> >
    >>> > Error in pmax(rep(1L, 3 * 10^9), 0) :
    >>> >    long vectors not supported yet:
    >>> > ../../../R-devel-src/src/include/Rinlinedfuns.h:522
    >>> 
    >>> 
    >>> I think a carefully tested patch that fixes pmax (it would need to
    >>> change this call from length() to xlength(), and make some other
    >>> necessary changes that follow from this), would probably be useful to R
    >>> Core, and could be posted to bugs.r-project.org.
    >>> 
    >>> It might also be useful on R-devel to post a list of all known commonly
    >>> used functions that don't support long vectors; this could be updated on
    >>> a regular basis.  This might encourage people to produce patches as above.
    >>> 
    >>> I'm not so sure a report about a single function won't just get lost.
    >>> 
    >>> Duncan Murdoch
    >>> 
    >>> ______________________________________________
    >>> R-devel using r-project.org mailing list
    >>> https://stat.ethz.ch/mailman/listinfo/r-devel
    >>> 
    >> 

    > [[alternative HTML version deleted]]

    > ______________________________________________
    > R-devel using r-project.org mailing list
    > https://stat.ethz.ch/mailman/listinfo/r-devel



More information about the R-devel mailing list