I cannot reproduce the problem with cygwin PDL-2.4.10 but the win32 PDL-2.4.10 give NaNs for the extrapolated values.
InSuk, please update the bug ticket on sf.net with your perl -V and perldl -V information. Thanks, Chris On Fri, Feb 10, 2012 at 9:26 AM, Chris Marshall <[email protected]> wrote: > I ran a C version of the example and it worked > and gave -1 for the value. I'll cook up a simple > test case to add to our test suite for use in > debugging the problem. If you see what is wrong > with the current gsl_interp.pd code, patches are > definitely welcome---and get fixes released > sooner. > > Thanks for the report, > Chris > > > On Thu, Feb 9, 2012 at 8:16 PM, chm <[email protected]> wrote: >> Hmm. Ok. So much for the quick fix. >> >> I did not think that a cspline fit was >> valid for the GSL interp eval routine. >> I guess someone who can check against the >> actual GSL interp routine results can show >> what the correct value should be. >> >> Sorry I couldn't be more helpful. >> >> --Chris >> >> >> On 2/9/2012 7:57 PM, InSuk Joung wrote: >>> >>> Hi, >>> Thank you for the reply. Anyway, it should not be an exception. The >>> correct >>> answer of the test code should be -1 not nan. At least, that's the number >>> I >>> get from v2.4.6. >>> >>> On Thu, Feb 9, 2012 at 7:37 PM, chm<[email protected]> wrote: >>> >>>> On 2/9/2012 6:01 PM, InSuk Joung wrote: >>>> >>>>> use strict; >>>>> use warnings; >>>>> use PDL; >>>>> use PDL::GSL::INTERP; >>>>> >>>>> my $x = sequence(10); >>>>> my $y = $x**2; >>>>> >>>>> my $spl = PDL::GSL::INTERP->init( 'cspline', $x, $y ); >>>>> print $spl->eval( -1, {Extrapolate=>1} ), "\n"; >>>>> >>>> >>>> The problem with the exception handler appears >>>> to have been fixed since the PDL-2.4.7 release >>>> (according to the git log). To avoid the problem >>>> you'll need to avoid out-of-domain values for the >>>> input parameter being evaluated or switch to PDL >>>> version 2.4.8 or newer (the PDL-2.4.10 release >>>> this week has many bugs fixed and new features!). >>>> I ran your test code on 2.4.10 and the result >>>> was nan with no error as expected. >>>> >>>> --Chris >>>> >>> >>> >>> >> _______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
