Re: [h5md-user] Offset

[Pierre de Buyl]

On Thu, Sep 26, 2013 at 09:05:56AM +0200, Konrad Hinsen wrote:
> Pierre de Buyl writes:
> 
>  > Even if there is no guarantee of folding, offset still serves a purpose to
>  > define the box. At the very least, as a convenience for checkpointing.
> 
> Could you explain that a bit more? I see checkpointing as a program
> storing information for its own later use. I wouldn't put that into
> any standardized data format, i.e. I'd put it outside of the datasets
> and groups specified by H5MD.

Well, my box is defined by xmin,xmax (repeat for y, z).

Without the offset, you have to assume "xmin=0" or "xmin=-edges[0]/2".

So, yes, offset is part of the box to me. If we think also (for the future, I am
not suggesting anything for now) of continuum methods, there is no "obvious"
choice for xmin.

P