RFC to replace H5Fis_hdf5 with a new routine

Would have been nice to have this a few years ago when developing a custom VFD for Silo.

That said, I am not sure how much more you are offering here if caller has to tell HDF5 lib which VFD to do. I mean, if Silo can produce its files with sec2, family, multi-part, or custom VFD, thats 4 different vfds I have to try.

Is there much value added in this routine over simply attempting H5Fopen using different fapls and taking first that succeeds?

Mark

···

From: Mohamad Chaarawi <chaarawi@hdfgroup.org<mailto:chaarawi@hdfgroup.org>>
Reply-To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Date: Friday, August 8, 2014 1:52 PM
To: "hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>" <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Subject: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Hi All,

I attached a RFC to deprecate H5Fis_HDF5() and add a new routine to replace it named H5Fis_accessible.
Please let me know if you have any comments/questions/opinions.

Thanks,
Mohamad

Hi Mark,

···

From: Hdf-forum [mailto:hdf-forum-bounces@lists.hdfgroup.org] On Behalf Of Miller, Mark C.
Sent: Friday, August 08, 2014 4:00 PM
To: HDF Users Discussion List
Subject: Re: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Would have been nice to have this a few years ago when developing a custom VFD for Silo.

That said, I am not sure how much more you are offering here if caller has to tell HDF5 lib which VFD to do. I mean, if Silo can produce its files with sec2, family, multi-part, or custom VFD, thats 4 different vfds I have to try.

I do agree that it is not very convenient/useful to have the application specify the access properties (VFDs, VOLs, and other properties). This routine is more of a bug fix, rather than a super new functionality offering :).
It might be easy to cycle through VFDs like the sec2 driver, but others would not be straightforward, especially if it is a custom VFD, without knowing some driver specific information/parameters used to create the file. The VOL layer is another thing to take into account. So I do not see a straightforward way to do what you are asking for.

Is there much value added in this routine over simply attempting H5Fopen using different fapls and taking first that succeeds?

Not much, but the internal implementation can do something smarter than going through the entire H5Fopen()/H5Fclose() stack, for example (in the native implementation) call directly the driver open routine.

Thanks,
Mohamad

Mark

From: Mohamad Chaarawi <chaarawi@hdfgroup.org<mailto:chaarawi@hdfgroup.org>>
Reply-To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Date: Friday, August 8, 2014 1:52 PM
To: "hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>" <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Subject: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Hi All,

I attached a RFC to deprecate H5Fis_HDF5() and add a new routine to replace it named H5Fis_accessible.
Please let me know if you have any comments/questions/opinions.

Thanks,
Mohamad

Ok, that makes sense to me. Thanks for the quick response.

Mark

···

From: Mohamad Chaarawi <chaarawi@hdfgroup.org<mailto:chaarawi@hdfgroup.org>>
Reply-To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Date: Friday, August 8, 2014 2:32 PM
To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Subject: Re: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Hi Mark,

From: Hdf-forum [mailto:hdf-forum-bounces@lists.hdfgroup.org] On Behalf Of Miller, Mark C.
Sent: Friday, August 08, 2014 4:00 PM
To: HDF Users Discussion List
Subject: Re: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Would have been nice to have this a few years ago when developing a custom VFD for Silo.

That said, I am not sure how much more you are offering here if caller has to tell HDF5 lib which VFD to do. I mean, if Silo can produce its files with sec2, family, multi-part, or custom VFD, thats 4 different vfds I have to try.

I do agree that it is not very convenient/useful to have the application specify the access properties (VFDs, VOLs, and other properties). This routine is more of a bug fix, rather than a super new functionality offering :).
It might be easy to cycle through VFDs like the sec2 driver, but others would not be straightforward, especially if it is a custom VFD, without knowing some driver specific information/parameters used to create the file. The VOL layer is another thing to take into account. So I do not see a straightforward way to do what you are asking for.

Is there much value added in this routine over simply attempting H5Fopen using different fapls and taking first that succeeds?

Not much, but the internal implementation can do something smarter than going through the entire H5Fopen()/H5Fclose() stack, for example (in the native implementation) call directly the driver open routine.

Thanks,
Mohamad

Mark

From: Mohamad Chaarawi <chaarawi@hdfgroup.org<mailto:chaarawi@hdfgroup.org>>
Reply-To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Date: Friday, August 8, 2014 1:52 PM
To: "hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>" <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Subject: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Hi All,

I attached a RFC to deprecate H5Fis_HDF5() and add a new routine to replace it named H5Fis_accessible.
Please let me know if you have any comments/questions/opinions.

Thanks,
Mohamad

Also, we are planning to add VFD autodetection soon so that setting the VFD in the FAPL will be unnecessary also.

···

--
Quincey Koziol, The HDF Group
Full Contact Info: http://bitly.com/QuinceyKoziol

On Aug 9, 2014, at 12:39 AM, Miller, Mark C. <miller86@llnl.gov<mailto:miller86@llnl.gov>> wrote:
Ok, that makes sense to me. Thanks for the quick response.

Mark

From: Mohamad Chaarawi <chaarawi@hdfgroup.org<mailto:chaarawi@hdfgroup.org>>
Reply-To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Date: Friday, August 8, 2014 2:32 PM
To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Subject: Re: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Hi Mark,

From: Hdf-forum [mailto:hdf-forum-bounces@lists.hdfgroup.org] On Behalf Of Miller, Mark C.
Sent: Friday, August 08, 2014 4:00 PM
To: HDF Users Discussion List
Subject: Re: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Would have been nice to have this a few years ago when developing a custom VFD for Silo.

That said, I am not sure how much more you are offering here if caller has to tell HDF5 lib which VFD to do. I mean, if Silo can produce its files with sec2, family, multi-part, or custom VFD, thats 4 different vfds I have to try.

I do agree that it is not very convenient/useful to have the application specify the access properties (VFDs, VOLs, and other properties). This routine is more of a bug fix, rather than a super new functionality offering :).
It might be easy to cycle through VFDs like the sec2 driver, but others would not be straightforward, especially if it is a custom VFD, without knowing some driver specific information/parameters used to create the file. The VOL layer is another thing to take into account. So I do not see a straightforward way to do what you are asking for.

Is there much value added in this routine over simply attempting H5Fopen using different fapls and taking first that succeeds?

Not much, but the internal implementation can do something smarter than going through the entire H5Fopen()/H5Fclose() stack, for example (in the native implementation) call directly the driver open routine.

Thanks,
Mohamad

Mark

From: Mohamad Chaarawi <chaarawi@hdfgroup.org<mailto:chaarawi@hdfgroup.org>>
Reply-To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Date: Friday, August 8, 2014 1:52 PM
To: "hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>" <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Subject: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Hi All,

I attached a RFC to deprecate H5Fis_HDF5() and add a new routine to replace it named H5Fis_accessible.
Please let me know if you have any comments/questions/opinions.

Thanks,
Mohamad

_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@lists.hdfgroup.org<mailto:Hdf-forum@lists.hdfgroup.org>
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5

Oh, well, even better :wink:

Out of curiosity, will that entail some easy-to-discover identifier in the boot block (similar to how custom data pipeline filters are handled) which explicitly indicates the VFD that should be used?

Also, how would you address the case that multiple different VFD's may be suitable to open the file but that one may be 'better' than another in certain circumstances? For example, with the Silo block-based VFD, I designed things so that other HDF5 built-in VFDs could still read the file (sec2, stdio, etc.). However, I would favor the SIlo block-based VFD over those on a BG/Q or BG/P system.

Mark

···

From: Quincey Koziol <koziol@hdfgroup.org<mailto:koziol@hdfgroup.org>>
Reply-To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Date: Monday, August 11, 2014 7:40 AM
To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Subject: Re: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Also, we are planning to add VFD autodetection soon so that setting the VFD in the FAPL will be unnecessary also.
--
Quincey Koziol, The HDF Group
Full Contact Info: http://bitly.com/QuinceyKoziol

On Aug 9, 2014, at 12:39 AM, Miller, Mark C. <miller86@llnl.gov<mailto:miller86@llnl.gov>> wrote:
Ok, that makes sense to me. Thanks for the quick response.

Mark

From: Mohamad Chaarawi <chaarawi@hdfgroup.org<mailto:chaarawi@hdfgroup.org>>
Reply-To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Date: Friday, August 8, 2014 2:32 PM
To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Subject: Re: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Hi Mark,

From: Hdf-forum [mailto:hdf-forum-bounces@lists.hdfgroup.org] On Behalf Of Miller, Mark C.
Sent: Friday, August 08, 2014 4:00 PM
To: HDF Users Discussion List
Subject: Re: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Would have been nice to have this a few years ago when developing a custom VFD for Silo.

That said, I am not sure how much more you are offering here if caller has to tell HDF5 lib which VFD to do. I mean, if Silo can produce its files with sec2, family, multi-part, or custom VFD, thats 4 different vfds I have to try.

I do agree that it is not very convenient/useful to have the application specify the access properties (VFDs, VOLs, and other properties). This routine is more of a bug fix, rather than a super new functionality offering :).
It might be easy to cycle through VFDs like the sec2 driver, but others would not be straightforward, especially if it is a custom VFD, without knowing some driver specific information/parameters used to create the file. The VOL layer is another thing to take into account. So I do not see a straightforward way to do what you are asking for.

Is there much value added in this routine over simply attempting H5Fopen using different fapls and taking first that succeeds?

Not much, but the internal implementation can do something smarter than going through the entire H5Fopen()/H5Fclose() stack, for example (in the native implementation) call directly the driver open routine.

Thanks,
Mohamad

Mark

From: Mohamad Chaarawi <chaarawi@hdfgroup.org<mailto:chaarawi@hdfgroup.org>>
Reply-To: HDF Users Discussion List <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Date: Friday, August 8, 2014 1:52 PM
To: "hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>" <hdf-forum@lists.hdfgroup.org<mailto:hdf-forum@lists.hdfgroup.org>>
Subject: [Hdf-forum] RFC to replace H5Fis_hdf5 with a new routine

Hi All,

I attached a RFC to deprecate H5Fis_HDF5() and add a new routine to replace it named H5Fis_accessible.
Please let me know if you have any comments/questions/opinions.

Thanks,
Mohamad

_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@lists.hdfgroup.org<mailto:Hdf-forum@lists.hdfgroup.org>
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5