Re: [Psidev-ms-dev] ASCII data representation (was: PSI-MSS WG Tuesday call reminder)

[Mike C. <tu...@gm...>]

On Tue, Jun 30, 2009 at 6:11 PM, Matthew
Chambers<mat...@va...> wrote:
> Hi Mike,
>
> Are you using long doubles in greylag? The reasonable fix if more than
> 15 digits are truly needed is to use a bigger data type, although a
> standard and portable long double does not exist AFAIK.

No, long doubles seem like overkill, at least at present.  I gave the
example only for informational purposes.  It does appear to me that
single precision floats are too small for some of the calculations
required for recent instruments.  Maybe they're enough for the peak
representation itself--I'm not sure.

> If one wanted to write trivial code to read XML...

For the languages I use, access to XML parsing, base64 coding, and
libz aren't really serious issues, but it is a little more involved
than what I do with our current format (old ms2, which is basically
one peak per line, represented as two floats in ASCII format), which
makes it simple to do several basic transformations using standard
Unix command-line tools.

It's tempting to concentrate on the size of spectrum files as a
metric, but the amount of programmer time it takes to do things
probably matters more at my shop.