Mathieu, (simple) HDF5 datasets are logically dense arrays.
With chunking you can mimic sparse storage, but you loose
track of where the non-zero entries are.
If you need that kind of information, you have to keep track
of that with, for example, a region reference or emulate
Compressed Row Storage or other sparse format directly
with more than one dataset.

Best, G.



-----Original Message-----
From: [email protected] [mailto:[email protected]]
On Behalf Of [email protected]
Sent: Friday, May 11, 2012 5:24 AM
To: [email protected]
Subject: Re: [Hdf-forum] Read All Available Chunk in a Dataset.

Thank you ! So as i understand there is no way to automatically select
availlable data without knowing where it is.

WHen i read the data, i normally dont know wich part of the dataset is full
of data and wich part is not, so i will have to store an hyperslab wich
describe where the available data is.

Thanks for your help

Mathieu

Quoting Gerd Heber <[email protected]>:

> That's why I asked you to define the smallest hyperslab or union of 
> hyperslabs, or point selections that cover you chunks and read just 
> those. The library will locate just those chunks (tree traversal, not 
> linear) and deal with them.
> Have a look at section 7.4. Dataspaces and Data Transfer in the user 
> guide!
>
> Best, G.
>
>
> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]]
> On Behalf Of Mathieu westphal
> Sent: Thursday, May 10, 2012 8:15 AM
> To: HDF Users Discussion List
> Subject: Re: [Hdf-forum] Read All Available Chunk in a Dataset.
>
>
>
> On 05/10/2012 02:01 PM, Gerd Heber wrote:
>> Mathieu, don't worry about the chunks. Just read/write what you need 
>> in terms of the dataspace (hyperslab, point selection...) and the 
>> library will determine for you which chunks are touched 
>> (read/written) by the operation.
>> It'll just return fill values for un-allocated/partially filled chunks.
> Yep you're right! But the time to read the data is linear with the 
> size of all the Dataset, not only of the size of available data !
> My real dataset is really sparse ! i need to read all available value 
> without taking time to read unavailable value.
> Is there anyway to select only real value?
>
> Anyway how to know how many value are really read from dataset?
>
> Thanks
>
> Mathieu
>
>>
>> Best, G.
>>
>> -----Original Message-----
>> From: [email protected]
>> [mailto:[email protected]]
>> On Behalf Of Mathieu westphal
>> Sent: Thursday, May 10, 2012 6:46 AM
>> To: [email protected]
>> Subject: [Hdf-forum] Read All Available Chunk in a Dataset.
>>
>> I have a dataset of size 16*16
>> I write Chunk of Data of size 4*4.
>>
>> Some data aren't available so i wrote only 3 chunk on my dataset.
>>
>> How to read only the 3 available chunk?
>>
>> Thanks
>>
>> Mathieu
>>
>> _______________________________________________
>> Hdf-forum is for HDF software users discussion.
>> [email protected]
>> http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org
>>
>>
>>
>> _______________________________________________
>> Hdf-forum is for HDF software users discussion.
>> [email protected]
>> http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org
>
>
> _______________________________________________
> Hdf-forum is for HDF software users discussion.
> [email protected]
> http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org
>
>
>
> _______________________________________________
> Hdf-forum is for HDF software users discussion.
> [email protected]
> http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org
>




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



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

Reply via email to