-
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
important numerical issue with (at least) SQRT() & ALOG() for negative inputs (due to change in IDL 8.5.1 in NaN for MIN/MAX) #1353
Comments
SORRY for the very dramatic way of my message, I was in a middle of processing of some urgent data from JWST, tired, ... And I did not take time to check whether it was a IDL or a GDL problem !!! (furthermore I install a new OS and loss my local old IDL versions) This is a serious concern in IDL : on the cluster here, default is IDL 8.5 :
The trick is here : https://www.l3harrisgeospatial.com/docs/max.html (what a stupid idea to change such an important behavior ! 😢) What to do ? (I see serious side effect in larges codes ...) |
Actually IMHO (and, should I say, in my educated opinion) NaNs ARE missing values, so it is ample time that IDL treats them as such. See how missing values are tagged in FITS. Menanig that IDL did take the good decision. Not that other equivalent software do better, btw. Most choke (or chocked, last time I tried them) on NaNs in a way or another. I personnaly would love to see GDL ignore all the /NAN options IDL added along the ages and just treat every NAN as 'missing', not just when the NAN option is present (and that would greatly simplify the code). Anybody wants to comment on that? In the present case, I suppose looking a documentation differences between 8.4 (the max version I have acces to) and current version will show the places where NaN should be taken into account without /NAN, I would argue that the relevant code is already present in GDL, just not triggered. |
I completely agree with you!
I never understood why we have to write max(array,/nan) if we want to know the maximum value of an array that might have NAN inside.
The max function should return the maximum value of the array and nothing else. If someone wants to know if there are NAN inside the array he can use the finite function.
… On 6. Aug 2022, at 08:47, Giloo ***@***.***> wrote:
Actually IMHO (and, should I say, in my educated opinion) NaNs ARE missing values, so it is ample time that IDL treats them as such. See how missing values are tagged in FITS.
Menanig that IDL did take the good decision.
Not that other equivalent software do better, btw. Most choke (or chocked, last time I tried them) on NaNs in a way or another.
I personnaly would love to see GDL ignore all the /NAN options IDL added along the ages and just treat every NAN as 'missing', not just when the NAN option is present (and that would greatly simplify the code). Anybody wants to comment on that?
In the present case, I suppose looking a documentation differences between 8.4 (the max version I have acces to) and current version will show the places where NaN should be taken into account without /NAN, I would argue that the relevant code is already present in GDL, just not triggered.
—
Reply to this email directly, view it on GitHub <#1353 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOC5K6DI7X642TJBFT4LFA3VXZGFXANCNFSM55XTKLXQ>.
You are receiving this because you are subscribed to this thread.
|
note that GDL has option |
A few meandering thoughts here. Also while it is likely true that most of the time we do want Also what happens with things like |
And going back in the past all GDL versions tested since 1.0.0 in august 2021 are buggy
but
The text was updated successfully, but these errors were encountered: