On Thu, 24 Nov 2005, Mark Taylor wrote:
>> JNIHDS looks a longer job (is Mark volunteering?).
>
> Hi chaps. I've only been keeping half an eye on this topic, can someone
> summarise for me (or point me to a description of) what are the changes
> in HDS are which need to be tracked?
Basically just switch to calling C HDS routines rather than Fortran ones,
the only complicating factor is that locators are now an opaque type
(HDSLoc), rather than the old character strings.
Another minor issue is that DAT_WHERE didn't make it into the C interface,
so support for that would need to be deprecated (quick grep of the source
tree doesn't show any actual use, so it could, in fact, just throw a
RuntimeException to discourage any use, or even disappear).
> Also, what are the consequences of not doing it - is the old FORTRAN
> interface going to disappear, or just be deprecated?
Neither, it will continue as it is. So this work is just a tidying up
exercise. I've already built JNIHDS against this version of HDS.
> I think(?) the only user packages which make use of JNIHDS are Treeview
> and SoG; there's no funding for support of Treeview and I presume the
> same is the case for SoG? If it's going to continue to work and only
> unsupported packages depend on it I'm probably inclined to leave it be,
> but if either of those is false I'd at least reconsider. But i'm open
> to arguments in any case.
Actually SPLAT uses this too (under Windows and via webstart). It would be
really nice if HDS could be compiled just using a C compiler as that would
make creating the Windows DLL much easier. If you don't want to do this I
could have a look sometime, but fiddling with all those macros doesn't
appeal!
Cheers,
Peter.
|