Difficult to define Encoders/Decoders for custom types which 'match' a common type

[ChrisPenner]

Hello again; I'm working through more of the upgrade to the latest version, I think enforcing OID matching makes sense and I've run into errors from mismatches in previous versions, so I'm thankful to get errors.

In service of that, I've got to write Encoders/Decoders for the common citext type. Previously I was easily able to just use the text decoders/encoders and everything just worked; but with OID checks it now crashes at runtime.

I see the custom decoder functionality, which is nice, and I was able to wire it up by copying bits of the internals of the text encoder/decoders; but I was wondering about the viability of a combinator which would allow using one Value decoder with a different OID;

E.g. I'd love to have been able to use a variant of refine which takes a different type-name so I could:

citextDecoder = refineWithCustomType "citext" CaseInsensitive.mk

Or something like it :)

Also, if you'd be amenable to the addition of including citext encoders/decoders in the library itself that'd be helpful and I'd be happy to contribute a patch.

Cheers!

[shinzui]

I'm also having trouble with the same thing.

[nikita-volkov]

Have you guys tried casting to text in SQL?

You can do that for both the results and params.

select my_citext::text from ... where my_citext = $1::text::ci_text

Then you should be able to use the text codec and rely on the Postgres type system to ensure that such "refinements" are compatible.

I was wondering about the viability of a combinator which would allow using one Value decoder with a different OID;

E.g. I'd love to have been able to use a variant of refine which takes a different type-name so I could:

citextDecoder = refineWithCustomType "citext" CaseInsensitive.mk

The problem of missing codecs is best addressed by just adding the missing ones. Do you have any other use cases for this combinator in mind? I suppose it might be useful for domain types. So let's think about it.

Also, if you'd be amenable to the addition of including citext encoders/decoders in the library itself that'd be helpful and I'd be happy to contribute a patch.

Of course!

Also theres's an option to add it to "postgresql-types". It'll be easy to reuse another codec in the implementation there. It is compatible with both "hasql" and "postgresql-simple".

[nikita-volkov]

citext codecs released