Less precise event_value and cumulative_probability data types

[nhoening]

We now use Double Precision data types in the event_value and cumulative_probability columns.

This issue proposes to move to Real types (read more here), so both columns use less space.

The precision loss is moving from 15 to 6 decimal digits. On the plus side, we'd use 8 bytes less per belief, which is more than 10% of all data (or almost 20% if we reduced the space of the belief column).

I researched a bit how to change that value in our ORM (we use Float, see here), but I didn't see yet what change to make here.

I changed the value to Real or Double but the generated migration did not pick this up.

• SQLAlchemy people are also not quite straightforward: switch float to cast to Double , not Float sqlalchemy/sqlalchemy#10300
• An idea how to directly change the column: https://stackoverflow.com/a/25552445 (in an empty migration file) - if this is the only way we see working, we can document this in the Readme for users to consider)