| Author |
Comment/Response |
Rhys Ulerich
|
 06/15/10 12:11pm
Dataset /u in the attached HDF5 file contains NaN values. This can be confirmed using 'h5dump bar.h5 | grep -i nan'.
However, using Mathematica 7's Import does not cause either NaN or Missing[] values in the input data. Observe
u = Import["bar.h5", {"Datasets", "/u"} ];
u[[5, 47 ;; 55, 1]]
which shows a range of NaN and non-NaN data. The NaN's appear, on my system, as "-1024.". These "-1024." values do not behave like either normal floating point numbers or NaNs:
(#1 < 0) & /@ h[[5, 47 ;; 55, 1]]
(#1 >= 0) & /@ h[[5, 47 ;; 55, 1]]
(#1 == #1) & /@ h[[5, 47 ;; 55, 1]]
However, InputForm shows no other information to indicate there's anything special about their representation. Because of this, I cannot seem to filter the data to replace these obnoxious things with Missing[] or something else that's suitable.
The documentation, from what I can tell, says nothing about what happens when importing NaN values.
Questions, with any answers appreciated:
1) What gives?
2) Can I get them to import as NaNs or Missing[] directly?
3) If not, can I filter them out in some way? I cannot seem to write a predicate to detect them since they behave so funkily.
Thanks in advance,
Rhys
- Rhys
Attachment: bar.h5, URL: , |
|
