Consider implementing an OID-based decoder compatibility check

[nikita-volkov]

No description provided.

[ocharles]

postgresql-simple does this, which you might want to use for reference.

[nikita-volkov]

The only usefulness of that would be in catching only a portion of possible errors and doing so in runtime. To me the benefit of that seems arguable.

[ocharles]

Personally I take the philosophy that most bugs lie at the boundaries of systems, and hasql is precisely there to interface between two systems. This is where a lot of error checking should be done, so I'd be in favour of doing this check.

[nikita-volkov]

I agree. However implementation-wise, it's gonna take some effort. I'll schedule doing this in the libpq-free version

[bergey]

Do you think checking the type info is easier if you bypass libpq? Or is your schedule just about prioritization?

[nikita-volkov]

Do you think checking the type info is easier if you bypass libpq?

No. It's perfectly possible to get this data thru libpq. And I would accept a PR implementing this.

Or is your schedule just about prioritization?

It is.