[erlang-questions] ux — a Unicode library

[Michael Uvarov]

Hello,

About deps.
I use module reloader from mochiweb. These dependencies can be
removed, which I did on github.

About unpredictability.
It's very interesting.

The first version of the library converted data to erlang code and
loaded it. It was fast, but I had the huge module.

Then, there are two versions of the DUCET file. It stores sorting
settings. The first version is standard, the second is for the CLDR
(although it is not yet implemented). That's why the library needs two
copies (or more) of the data of the same type.

In new realization, different processes are able to use different settings.

Disadvantages of loading on the fly:
eunit needs the timeout directive. Also, I don't know how to say to
eunit to load ux before running tests. So, when  TESTS is defined,
library behaves a little differently, because ux_unidata_filelist is
not running.

Does anybody khow  how to ask Eunit run application ux before the tests?

There is no serialization: data stores in the ETS, the parser returns
a function for each data type that knows how to get data and what to
do if the ETS table was killed.

In order to avoid unnecessary requests to other processes, these
functions are cached in the dictionary process.