Support schema in custom postgres type names, generate PgHasArrayType impls for enums #2791
Conversation
|
Adding a generated impl to the derive is unfortunately a breaking change as it conflicts with any manually added impl. We've learned this the hard way. |
|
should we add a separate derive then? and should we split off the first commit into a separate pr? what's the best path forward you think? |
This comment was marked as spam.
This comment was marked as spam.
|
It would be great to have this, maybe the |
| @@ -1152,7 +1152,10 @@ impl PartialEq<PgType> for PgType { | |||
| true | |||
| } else { | |||
| // Otherwise, perform a match on the name | |||
| self.name().eq_ignore_ascii_case(other.name()) | |||
| // If either name has no schema qualification, match only on the name part. | |||
There was a problem hiding this comment.
I guess it is not possible with the current info (due to search_path) to create a fully correct eq method using pg_ type names. How did you decide on this matching?
| @@ -84,7 +84,7 @@ async fn it_describes_composite() -> anyhow::Result<()> { | |||
|
|
|||
| let ty = d.columns()[0].type_info(); | |||
|
|
|||
| assert_eq!(ty.name(), "inventory_item"); | |||
| assert_eq!(ty.name(), "public.inventory_item"); | |||
There was a problem hiding this comment.
There should be tests involving cases where the database will forcibly escape names, like "my::schema"."TypeName".
| // If either name has no schema qualification, match only on the name part. | ||
| // If both names have a schema qualification, match both the schema and name parts. | ||
| std::iter::zip(self.name().rsplitn(2, '.'), other.name().rsplitn(2, '.')) | ||
| .all(|(part, other_part)| part.eq_ignore_ascii_case(other_part)) |
There was a problem hiding this comment.
PostgreSQL doesn't use case-insensitive matching for escaped names. public."TypeName", for example, won't match public.typename (but public.TYPENAME, without escaping, will). I suggest we simply treat escaped and non-escaped names on an entirely separate branch, using case-insensitive ASCII comparison for non-quoted names, and using strict utf-8 equality otherwise.
This comment was marked as spam.
This comment was marked as spam.
This comment was marked as spam.
This comment was marked as spam.
1 similar comment
This comment was marked as spam.
This comment was marked as spam.
Fixes #2595, #2262, and #1576.
replaces #2641