NUMERIC, Scientific, NaN

[jchia]

A Postgresql NUMERIC value can be NaN, but a Haskell Scientific value cannot.

However, numeric links NUMERIC to Scientific. What can I do if I want to use NUMERIC columns with NaN values?

The next best thing I can do is to make the column nullable, but I would rather simply use NaN since Postgresql allows it.

decimal does what I want except that I am trying to use decimal fixed-point numbers, which Scientific affords, not binary floating-point numbers.

If instead of:

numeric :: BeamSqlBackend be => Maybe (Word, Maybe Word) -> DataType be Scientific

there is:

numeric :: BeamSqlBackend be => Maybe (Word, Maybe Word) -> DataType be (Maybe Scientific)

Then NaN could be represented as Nothing and my immediate problem would be fixed, though +/-Inf would still be problematic.

[LaurentRDC]

Low-level mapping between basic Haskell types and PostgreSQL types is handled by the postgresql-simple library. Since Scientific is such a foundational type, I am weary of messing with it in beam