h5md-user
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [h5md-user] Mandate Variable-length string datatype


From: Felix Höfling
Subject: Re: [h5md-user] Mandate Variable-length string datatype
Date: Wed, 25 Sep 2013 09:31:59 +0200
User-agent: Opera Mail/12.15 (Linux)

Am 24.09.2013, 17:54 Uhr, schrieb Peter Colberg <address@hidden>:

On Tue, Sep 24, 2013 at 05:29:42PM +0200, Felix Höfling wrote:
Now (since Peter made the vlen string mandatory), the H5MD user reading
the spec has to bother about the details of HDF5 String. Before (when all
kinds of HDF5 Strings were allowed), the wording of the spec was simpler
and the user didn't even have to know that there are different strings [1].

[1] Actually I didn't either until recently due to HALMD and h5py using
different string types.

Yes, but don't you see, this is what an explicit choice solves.

Without the explicit specification of string types, a developer will
probably be ignorant; the result is that the user of the developer’s
software will be annnoyed due to the incompatibility of H5MD-using
softwares. (I was annoyed when a highly accelerated large-scale
molecular dynamics package in an earlier version did not read the
H5MD files written by my software ;-).)

Peter


Then, it boils down to an implementation hint (or warning) for developers of H5MD libraries, but we don't need it in the specification.

BTW, I find the asymmetric resource management of the HDF5 library pretty weird (and error prone): the library allocates memory, but the application has to free it.

Felix



reply via email to

[Prev in Thread] Current Thread [Next in Thread]