Hi John,
Thanks for the reply and making the feature request for HDFView. You and
Werner are correct that the dump/ls tools do produce information about
committed types - and this is the debug information I was after - unless
they are part of a H5T_COMPOUND.

In the case of a compound, only raw information about the datatype is
presented. It would be good if it were able to present the location of
committed types in the same way, e.g.

   DATASET "test" {
>       DATATYPE  H5T_COMPOUND {
>          H5T_IEEE_F64LE "a";
>          H5T_IEEE_F64LE "b";
>          H5T_IEEE_F64LE "c";
>       }
>       ...


becomes

   DATASET "test" {
>       DATATYPE  H5T_COMPOUND {
>          "/unit1" "a";
>          "/unit2" "b";
>          "/unit3" "c";
>       }
>       ...


or similar.

Cheers,
Martijn

On 9 December 2014 at 18:13, John Readey <[email protected]> wrote:

>  Hi Martijn,
>   In h5dump, I see that datasets that use a committed types have the path
> to the type identified.  E.g.:
>
>  DATASET "dset1" {
>
>       DATATYPE  "/dtype1"
>
>       DATASPACE  SIMPLE { ( 10 ) / ( 10 ) }
>
>       DATA {
>
>       (0): 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
>
>       }
>
> }
>
> DATATYPE "dtype1" H5T_IEEE_F32LE;
>
>
>
>  Is that not sufficient to see how the committed type is used?
>
>
>  In HDFView, the committed types are shown in the Tree View, but
> dataset’s that reference them don’t indicate this.  So that would be a good
> feature to add.  I’ve created an issue to track this request.
>
>
>  Regards,
>
> John Readey
>
> The HDF Group
>
>
>
>
>   From: Werner Benger <[email protected]>
> Reply-To: "[email protected]" <[email protected]>
> Date: Sunday, November 16, 2014 at 6:41 AM
> To: "[email protected]" <[email protected]>
> Subject: Re: [Hdf-forum] Committed datatypes in HDFView/h5dump
>
>   Hi Martijn,
>
>  if you use h5ls, then it will indicate named data types as being "shared"
> such as:
>
>
>  h5ls -rv
> LOD.f5/t=000000000.0000000000/Triangulation/Faces/Points/Positions
> Opened "LOD.f5" with sec2 driver.
> t=000000000.0000000000/Triangulation/Faces/Points/Positions Dataset
> {94525/94525}
>     Location:  1:1442029
>     Links:     1
>     Modified:  2014-09-17 01:34:18 Central Daylight Time
>     Storage:   1134300 logical bytes, 1134300 allocated bytes, 100.00%
> utilization
>     Type:      shared-1:1440965 struct {
>                    "i"                +0    native 32-bit field
>                    "j"                +4    native 32-bit field
>                    "k"                +8    native 32-bit field
>                } 12 bytes
>
>
> If you do an h5ls on the type itself, it will show its attributes and the
> shared-identifier as well:
>
>  h5ls -rv LOD.f5/Charts/triangular/SinglePrecision/Point
> Opened "LOD.f5" with sec2 driver.
> Charts/triangular/SinglePrecision/Point Type
>     Attribute: ChartDomain scalar
>         Type:      11-byte null-terminated ASCII string
>         Data:  "triangular"
>     Attribute: F5::CellDimensionality scalar
>         Type:      native int
>         Data:  2
>     Attribute: MemoryOrder {3}
>         Type:      native int
>         Data:  1, 0, -1
>     Location:  1:1440965
>     Links:     2
>     Modified:  2014-09-17 01:34:18 Central Daylight Time
>     Type:      shared-1:1440965 struct {
>                    "i"                +0    native 32-bit field
>                    "j"                +4    native 32-bit field
>                    "k"                +8    native 32-bit field
>                } 12 bytes
>
>
> Cheers,
>
>       Werner
>
>
> On 16.11.2014 02:12, Martijn Jasperse wrote:
>
> Hi all,
> I'd like to be able to tell in HDFView/h5dump when a dataset has a
> committed datatype. Ideally I'd like a path to the committed type be
> provided (a la H5Iget_name), but it would be sufficient to identify it as
> being a committed type. It would be very useful to me for the debugging
> tools to display this and not have to use my tools to check their files.
>
>  I'm using committed types to implement a type of units handling, and it
> works great, except that the standard tools are not very helpful in working
> with them. It would be excellent to add this feature to HDFView and/or
> h5dump, but I would also be interested in other approaches and/or
> workarounds.
>
>  Cheers,
> Martijn
>
>
> _______________________________________________
> Hdf-forum is for HDF software users discussion.
> [email protected]http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
> Twitter: https://twitter.com/hdf5
>
>
> --
> ___________________________________________________________________________
> Dr. Werner Benger                Visualization Research
> Center for Computation & Technology at Louisiana State University (CCT/LSU)
> 2019  Digital Media Center, Baton Rouge, Louisiana 70803
> Tel.: +1 225 578 4809                        Fax.: +1 225 578-5362
>
>
> _______________________________________________
> Hdf-forum is for HDF software users discussion.
> [email protected]
>
> http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
> Twitter: https://twitter.com/hdf5
>
_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5

Reply via email to