fix: re-introspection with @@map and relationMode
#3335
Conversation
…ionMode/referentialIntegrity and with/without '@@Map'
|
Note: |
introspection-engine/introspection-engine-tests/tests/re_introspection/relation_mode/sqlite.rs
Outdated
Show resolved
Hide resolved
| let old_model_name_to_final_database_name: HashMap<String, String> = old_data_model | ||
| .models() | ||
| .map(|m| (m.name.clone(), String::from(m.final_database_name()))) | ||
| .collect(); |
There was a problem hiding this comment.
Any suggestion on how to avoid these clones while still being able to use the function in the following closures is well accepted.
There was a problem hiding this comment.
I wouldn't worry about it here, but you can change line 147 to .map(|m| (m.name.as_str(), m.final_database_name())) and the type ascription to HashMap<&str, &str>. To optimize further, you could make the HashMap potentially a lot smaller if you .filter() for only the models with a mapped name first — then a model missing in the map would mean the model is not remapped, which would be handled with the existing code lower down.
But we shouldn't worry too much here, we're in the process of rewriting this file and the string comparisons will go away, we'll have a big map of integer identifiers for these lookups, since they happen repeatedly during introspection and reintrospection.
There was a problem hiding this comment.
Thanks. I have simply got rid of .clone() in favor of .as_str(). I will skip other optimizations for now, as we expect this file to change significantly already
Yes it's probably a whitespace issue — I've never encountered it though so I'm a bit puzzled. Maybe it's due to the extra indentation from the inline modules ( |
| let old_model_name_to_final_database_name: HashMap<String, String> = old_data_model | ||
| .models() | ||
| .map(|m| (m.name.clone(), String::from(m.final_database_name()))) | ||
| .collect(); |
There was a problem hiding this comment.
I wouldn't worry about it here, but you can change line 147 to .map(|m| (m.name.as_str(), m.final_database_name())) and the type ascription to HashMap<&str, &str>. To optimize further, you could make the HashMap potentially a lot smaller if you .filter() for only the models with a mapped name first — then a model missing in the map would mean the model is not remapped, which would be handled with the existing code lower down.
But we shouldn't worry too much here, we're in the process of rewriting this file and the string comparisons will go away, we'll have a big map of integer identifiers for these lookups, since they happen repeatedly during introspection and reintrospection.
introspection-engine/introspection-engine-tests/tests/re_introspection/relation_mode/mssql.rs
Outdated
Show resolved
Hide resolved
| Ok(()) | ||
| } | ||
|
|
||
| // referentialIntegrity = "foreignKeys" with @@map loses track of the relation policy ("foreignKeys"), but preserves @relations, which are moved to the bottom. |
There was a problem hiding this comment.
I'm wondering whether we should be testing relationIntegrity = "foreignKeys" more than once, since that's the default everywhere. Let's keep these tests since they're already there.
There was a problem hiding this comment.
How do you mean, "more than once"? I'm not sure I'm following
There was a problem hiding this comment.
once would be on just sqlite, or just postgres, for example
introspection-engine/introspection-engine-tests/tests/re_introspection/relation_mode/mssql.rs
Outdated
Show resolved
Hide resolved
Currently, re-introspection with
relationMode = "prisma"orreferentialIntegrity = "prisma"loses track of@relations in the schema. This PR fixes that.Fixes prisma/prisma#11022.