Re: CIF-JSON draft 2017-05-08

> As far as numbers go, it is clear that representation of numbers as strings
> should be allowed in order to support translation from CIF files.

Translation from CIF is *easier* when the numbers are written as
strings, because one doesn't even need to parse numbers. But,
comparing with complexity of parsing the CIF format, parsing numbs and
writing them as possibly two separate numbers is not that difficult.
The downside of the quoted representation in JSON is of course that
the recipient of such JSON file, after presumably using a third-party
JSON parser, needs to finish the parsing himself.

John reasonably argued that a single representation is better than
two. After thinking about it I'd agree. But I'd not agree with the
choice. Parsing numbers on the reading side should be done entirely by
a JSON parser. Usually there are more consumers of file formats than
producers, and the extra complexity is preferable in the CIF->JSON
step rather than when working with JSON.

If anyone thinks that parsing numbs is trivial and no extra complexity
is involved, I propose that someone familiar with C or C++ writes here
a (thread-safe) function that can parse the numb format. As a hint:
functions to parse numbers in a locale-independent way are available
only in C++17 which is not widely adopted yet.

Marcin

_______________________________________________
cif-developers mailing list
cif-developers@iucr.org
http://mailman.iucr.org/cgi-bin/mailman/listinfo/cif-developers