On Mon, 20 Oct 2014 19:33:25 +0200, Luc Maisonobe wrote:
Le 20/10/2014 16:00, Hank Grabowski a écrit :
I have some time this week to try to get these changes made to the
interpolators. I don't want to do anything without consensus
however. So
to try to incorporate the discussion above plus a concerned raised
over the
weekend on the JIRA thread I propose:
1. Adding the original functionality back so that the original
classes are
there, but adding deprecation and very explicit warnings to the
documentation that they should not be used. One release ahead of
this I
think they should be removed however.
2. Move my code into a new PiecewiseBicubicSpline classes.
The changes I discussed in the e-mail that Giles referred to above I
will
be making to these new PiecewiseBicubicSpline classes, and the
TricubicSpline i will do the same Piecewise surname rather than
replacing
the classes.
I do not want to do this with backing outt the existing pull
request. I
will simply make the changes and then initiate a follow-up one.
How does that sound to everyone? We need to get accurate
interpolators in
the next release, so the sooner I can squeeze this into my schedule
to get
this committed the better.
It sounds good to me, but I prefer to wait for Gilles analysis as he
is
more aware than me of any problems on this topic.
I'm fine with the plan, as long as we've agreed that whatever is
committed is an "experimental" version, still subject to change
until everybody is satisfied with the new design.
[I'm not referring to the main functionality; I of course agree
that accurate interpolation should be made available. But from
what I've seen, trivial modifications are required to set the
code on a par with what we had usually committed as new
contributions.]
Regards,
Gilles
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
For additional commands, e-mail: dev-h...@commons.apache.org
