h5dump: what am I doing wrong?

Hdf-forum archives
1

I'm once again puzzled by the behavior of h5dump. In this case I have the following dataset:
    DATASPACE SIMPLE { ( 288, 2, 32, 1 ) / ( 288, H5S_UNLIMITED, H5S_UNLIMITED, H5S_UNLIMITED ) }

and I was trying to dump the first "row" (i.e. all elements with the 1st dimension = 0). I tried
h5dump -d dataset-name -s "0,0,0,0" -k "1,2,32,1" somefile.h5

but the results were not at all what I expected. Here's a brief listing of just the coordinates h5dump prints out:
       (0,0,0,0): {
       (0,0,1,0): {
       (0,0,2,0): {
       (0,0,3,0): {
...
       (0,0,31,0): {
       (0,1,0,0): {

what's more, the elements at each coordinate are completely different. The first 0,0,1,0 matches the fill data, while the second 0,0,1,0 has data in it.

Now, I was doing this in an attempt to verify what hdfdump (2.7) was telling me, which was that 0,0,0,0 was exactly the same as 0,1,0,0 which shouldn't be the case unless I have a bug in my software. h5dump doesn't really confirm this but both hdfview and h5dump appear to be displaying things incorrectly, certainly inconsistently.

I'm using a modified (for iRODS support) HDF5 1.8.8 snap 6.

2

Ok so after I posted that I remembered having gone through this once already. Using -c instead of -k works, and verifies that what hdfview was displaying was incorrect (which has also been double-checked using my own code).

···

On 02/03/12 09:44, John Knutson wrote:

I'm once again puzzled by the behavior of h5dump. In this case I have
the following dataset:
   DATASPACE SIMPLE { ( 288, 2, 32, 1 ) / ( 288, H5S_UNLIMITED,
H5S_UNLIMITED, H5S_UNLIMITED ) }

and I was trying to dump the first "row" (i.e. all elements with the
1st dimension = 0). I tried
h5dump -d dataset-name -s "0,0,0,0" -k "1,2,32,1" somefile.h5

but the results were not at all what I expected. Here's a brief
listing of just the coordinates h5dump prints out:
      (0,0,0,0): {
      (0,0,1,0): {
      (0,0,2,0): {
      (0,0,3,0): {
...
      (0,0,31,0): {
      (0,1,0,0): {

what's more, the elements at each coordinate are completely
different. The first 0,0,1,0 matches the fill data, while the second
0,0,1,0 has data in it.

Now, I was doing this in an attempt to verify what hdfdump (2.7) was
telling me, which was that 0,0,0,0 was exactly the same as 0,1,0,0
which shouldn't be the case unless I have a bug in my software.
h5dump doesn't really confirm this but both hdfview and h5dump appear
to be displaying things incorrectly, certainly inconsistently.

I'm using a modified (for iRODS support) HDF5 1.8.8 snap 6.

_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@hdfgroup.org
http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org

3

John,

There is a bug in HDFView version 1.8 which causes incorrect values to be displayed in ImageView using Show Value. You don't indicate if that is your failure mode, but a patch for this was released and is available from the hdfgroups anonymous FTP server, which can also be accessed from their web site:

  http://www.hdfgroup.org/ftp/HDF5/hdf-java/patches/p1/

--dan

John Knutson wrote:

···

Ok so after I posted that I remembered having gone through this once already. Using -c instead of -k works, and verifies that what hdfview was displaying was incorrect (which has also been double-checked using my own code).

On 02/03/12 09:44, John Knutson wrote:

I'm once again puzzled by the behavior of h5dump. In this case I have
the following dataset:
   DATASPACE SIMPLE { ( 288, 2, 32, 1 ) / ( 288, H5S_UNLIMITED,
H5S_UNLIMITED, H5S_UNLIMITED ) }

and I was trying to dump the first "row" (i.e. all elements with the
1st dimension = 0). I tried
h5dump -d dataset-name -s "0,0,0,0" -k "1,2,32,1" somefile.h5

but the results were not at all what I expected. Here's a brief
listing of just the coordinates h5dump prints out:
      (0,0,0,0): {
      (0,0,1,0): {
      (0,0,2,0): {
      (0,0,3,0): {
...
      (0,0,31,0): {
      (0,1,0,0): {

what's more, the elements at each coordinate are completely
different. The first 0,0,1,0 matches the fill data, while the second
0,0,1,0 has data in it.

Now, I was doing this in an attempt to verify what hdfdump (2.7) was
telling me, which was that 0,0,0,0 was exactly the same as 0,1,0,0
which shouldn't be the case unless I have a bug in my software.
h5dump doesn't really confirm this but both hdfview and h5dump appear
to be displaying things incorrectly, certainly inconsistently.

I'm using a modified (for iRODS support) HDF5 1.8.8 snap 6.

_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@hdfgroup.org
http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org

_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@hdfgroup.org
http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org

--
Daniel Kahn
Science Systems and Applications Inc.
301-867-2162