[seqfan] Re: qlog in R

[David Wilson]

OK, so I typed up my program and was all ready to send it to you when lo 
and behold
it worked. I hate those Schroedinger bugs (bugs that manifest/go away 
when the code
is observed). Now I feel a bit like Emily Litella ("Oh. Never mind").

I was able to make the qlog plot work in for A122803 (powers of -2). The 
negative values
in that sequence are large enough to uncover a bug in a simplistic asinh 
implementation,
but I don't see any problems, so I must have been doing something else 
wrong in my
testing, what I don't know. I was fiddling with some function 
definitions, and may have
corrupted some variables (R doesn't have local variables), yeah, that's it.

So attached please find the program with a qlog plot in place of the log 
plots. I know
that Russ requested that the qlog plot be added below the log plot, but 
I couldn't figure
out how to make that happen.

On 10/14/2011 5:12 PM, David Applegate wrote:
> David,
>
> Could you give me an example of a large negative number for which R's
> asinh failed?  I did some quick testing and it seems to work for anything
> that would pass the is.finite() test at the start (that is, anything
> that was a representable double-precision floating-point number).
>
> If there is a reasonably fixable failure case, I might be able to talk
> to one of the R maintainers about fixing it.
>
> -Dave
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: prog.r
URL: <http://list.seqfan.eu/pipermail/seqfan/attachments/20111015/b6845347/attachment.ksh>