Prisma generating a NUMERIC(65, 30) for PostgreSQL, doesn't match docs

Prisma 2 migrate autogenerated a column of Prisma’s Float type as a PostgreSQL Numeric with a precision of 65 and a scale of 30.

1. The documentation says that this should be a real type in PostgreSQL, not a numeric
2. Why is Prisma making such a precise column? This will result in massively more data consumption and will make calculations wildly slow compared to any other number type (see PostgreSQL documentation 8.1.2. Arbitrary Precision Numbers)

Edit: The generated migration Markdown file says it was attempting to create a Decimal(65,30), which I believe is just an alias for the Numeric type. Perhaps look through source code for that string?

[tomhoule]

@molomby this is what we are planning. The decimal situation is an old workaround for client tests that will be cleaned up.

[albertoperdomo]

The idea that a migration could be created when the schema hasn’t changed, and that precision in stored data could be lost seems like a Bad Idea.

This is going to be indeed a breaking change we will have to carefully communicate to our users but we believe it’s for the best, because in the future we will be providing a better default. The way around the precision loss is indeed to introspect, which could be considered to some extent a codemod.