Difficulty employing Duration to PostgreSQL :interval with solution

First of all I am happy that you found the solution, but this is not an actual issue of Ecto, but a fact of life that you have used postgres specific functionality. A PR to improve awareness of this to newcomers would be more welcome!

Per-documentation this is supported out of the box, more info you can find in add/3.

That’s absolutely not true, ecto is a generic library that supports many database engines, there is no reason ecto would support postgres features exclusively.

Ecto schemas and migrations are not linked in any way, please read the disclaimer in the link above. There are reasons why it was implemented like this and one of those is flexibility (note in documentation how you can define random custom database types in migrations).

This was very helpful, thanks! The docs do indeed show snippets of the required configuration, but don’t give you enough information to piece it together. I’d definitely recommend submitting a PR to improve this.

Specifically, I was missing the need to slot Ecto.Adapters.Postgres.extensions() into the call to Postgrex.Types.define/3, which I’m pretty sure is not mentioned anywhere. A PR could add an ex_doc cheatsheet/guide to this effect, or add to the Postgrex.Types moduledoc.

I don’t believe the author was citing this as an issue with Ecto; rather, Postgrex documentation.

The author is talking about Postgrex here, not Ecto. Postgrex is the (main) library that extends generic Ecto with postgres features.

This is a problem the author understands and is sharing their solution for.

• Ecto SQL supports intervals in migrations with :interval, but not in schemas.
• Postgrex has support for intervals in schemas with :duration, using a Postgrex.Duration polyfill in case your Elixir predates the stdlib Duration.
• It also has documentation on how to configure it to use the stdlib Duration, but not really enough guidance on how to do this without consulting source code, hence this helpful post.

Good points!

I still think ecto should be responsible for covering this kind of information, otherwise it’s really hard to find the documentation, as migrations even though are not linked in any way to schemas, they are still part of ecto functionality.

I disagree, per your own point that

I’d say this extends to documentation on postgre(s|x) exclusive features: that information should live in adapter-specific docs, which could be improved for this specific stlib Duration feature of Postgrex the author (and later, I) was struggling with.

Or are you simply referring to Ecto documenting somewhere the information that schema and migration DSLs are not guaranteed to align? I agree that might be helpful to newcomers to Ecto! Just pointing out that that concept is unrelated to this post.

I think you might be confused about what exactly the author was struggling with: the correct incantation of Postgrex.Types.define/3 to get schemas to hydrate :duration fields into stdlib Duration structs instead of Postgrex.Duration ones. It’s a feature that Postgrex documents exists, but does not provide enough information to leverage. It’s unrelated to schema/migration DSL differences or any confusion around that.

OK, this makes more sense now. I always thought that postgrex was strictly doing the plumbing between ecto ↔ postgres, this is the first time I hear about it offering extensions for postgres specific types that can be used in ecto, even though logically that makes perfect sense.

I apologize for not making my original post clear enough to highlight the main point, and thank @christhekeele for explaining it with such precision. While I agree the Postgrex documentation should be improved, I don’t expect to do it myself, so the task is up for grabs for anyone who wants it.