David Carlisle skrev 2014-05-05 12.40:
On 05/05/2014 10:28, Lars Hellström wrote:
1. More IEEE float formats.
OM currently only supports the 64-bit double format for floats, but
the current version of the source IEEE standard defines floats of
arbitrarily high precision (every multiple of 32 bits, if I recall
correctly; there's a formula for how many bits go into the exponent
and how many go into the mantissa). The OM standard could be extended
to allow OMF with hex attribute values any multiple of 8 hexdigits
long. (https://trac.mathweb.org/OM3/ticket/137)
A natural translation of some
<OMF hex="0123456789ABCDEF0123456789ABCDEF"/>
(128 bits) to strict OM2 would be as some
<OMA> <OMS cd="IEEEfloat" name="fromOMB"/>
<OMB>ASNFZ4mrze8BI0VniavN7w==</OMB> </OMA>
I much prefer the existing OMS notation for that. There always were
different formats for floating point numbers so could have used a CD for
all of them, OMF special cased Double as a convenience as that was (and
is) overwhelmingly the most common type for floating point literals
in programming languages. I don't think the extended precision types
change that reasoning any more than IEEE single. This is a cosmetic
change that breaks applications unless they are updated or the code is
pre-processed back to OM2.
A quite reasonable position. But then the standard should explicitly state
that OMF is provided as a convenience, and suggest the constructor symbol
solution for people who want other types of floats.
2. OMOBJs as mathematical objects
I thought we had a meta CD for that, but apparently not (or at least we
don't seem to have one now?) meta CD allows the quoting of CDs (only)
Yes, and I would argue that without a quoting mechanism, the meta CD gets it
wrong (embedding naked omels for the OMOBJs in the .ocd file, rather than
quoted omels).
Lars Hellström
I'm sure at one time we had a CD with symbols OMOBJ OMA etc.
David
_______________________________________________
Om3 mailing list
[email protected]
http://openmath.org/mailman/listinfo/om3
_______________________________________________
Om3 mailing list
[email protected]
http://openmath.org/mailman/listinfo/om3
