-
Notifications
You must be signed in to change notification settings - Fork 61
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
avg.pro with GDL 1.0.3 #1650
Comments
@AndyAstronomer indeed this was due to an unfortunate optimization. PR #1651 should solve this. |
Hi @AndyAstronomer . Is someone in you team registered to the GDL announces mailing list ? Thanks for the feedback, this change of dims. is not so easy to understand and test in GDL & IDL. Any idea to write some tests will be welcome ! Since AVG is now mentioned as deprecated, I tested whether MEAN is really equivalent I found a strange behavior in IDL, then I will send them a bug report :
@GillesDuvert @pjb7687 I vote for a 1.0.4 ASAP since we need to have feedback on fast GAUSSFIT from end-users |
Hi @alaingdl , I don't think there's a problem there. The DIMENSION key starts from 1, so DIMENSION=0 is equivalent to not setting the dimension. Hence you are trying to take the mean over a dimension of a scalar so I'd expect the result to be undefined. P.S. At least GDL gives an error for taking a dimension=1 mean of a scalar. |
But I can legitimately call MEAN() with dim=0 ?!
(as you know dist() gives a 2D array) |
My initial comment was wrong, and I corrected it. It's (e.g.)
that gives a NaN in IDL and an error in GDL. |
I already submitted a bug report to IDL NV, we will see. I found it is not "consistent". I found the message in GDL way better 😄 |
Hi @alaingdl, no one in the group is on the official list, however, some of my colleagues are based at Observatoire de Paris. If memory serves, there is a GDL announcement email list there, correct? |
@AndyAstronomer in fact it is a public diffusion list with very few messages, anybody even outside Obs can registered |
Well, since we started a discussion 😄 may I ask @AndyAstronomer what difficulties they found (find?) in porting their code from IDL to GDL. |
Hi @GillesDuvert. So @alaingdl may remember helping us some years ago on this problem in Paris. We had quite some difficulty getting the call_external function to properly handle the Fortran parts of the code, especially when they were compiled using Intel Fortran. I cannot fully remember the reasons behind the difficulties; something to do with array stacking optimisation. This was solved by compiling with GNU Fortran. Other than that, it took some time as we found that the syntax is stricter in GDL, which is no bad thing So we spent some time editing dummy arguments and calls for a number of the subroutines that make up our code. It mostly works now, except for this new issue. |
reverted ominous oops and close #1650
Just merged my patch and closed this issue. Thanks for the report and sorry for the inconvenience! |
My working group have discovered a small issue with GDL 1.0.3 and avg.pro in the NASA astronomy library. It no longer reforms a 3D array into a 1D array when the first two dimensions contain a single element. Here is a simple example:
In these examples, the first two dimensions should have been averaged out, and only the third dimension should be returned, like in the third example. This was properly addressed in version 1.0.0:
This leads to many problems with many of our codes as we continue to port from IDL to GDL.
The text was updated successfully, but these errors were encountered: