A safe, extensible ORM and Query Builder for Rust – For feature requests and longer questions also see https://github.com/diesel-rs/diesel/discussions/categories/q-a
But, it doesn't work. I need to make a struct for this. Is there a way to not need this?
@BlackDex
That's because we can check at compile time that the later variant returns exactly one column/field, while we don't know that for the first one.
sql_query requires always a custom struct that implements QueryableByName. Instead of using sql_query you could define your version function via sql_function! and use the typed query builder (so diesel::select(version()) which allows you to select directly into a String.That's because we can check at compile time that the later variant returns exactly one column/field, while we don't know that for the first one.
Thx for the explanation. I will try the ::select()
@NoraCodes That's something that is not supported yet on any released diesel version. The next feature release will contain larger changes to the migration system, that allow configuring if a migration should be run in a transaction or not. (And may other things. See diesel-rs/diesel#2657 for details)
Okay. We're using Diesel in production, and as a work around we've created the index on the production database manually; should we wait for that issue to be resolved to bring the migration state up to the real state, or is there a workaround to record that we've run the migration manually? (other than just modifying the migration table ourselves)
Whats the case for creating an index not as a migration? In lemmy we do have a case where we want to re-index, so we run those in a job scheduler in our code using
sql_query.
Ideally we wanted to have the database state fully captured by migrations; in fact this would be only the third migration of this service ever in 3 years of operating. It's not that we wanted to rebuild the index, but actually create a new one while the service is up, without disruption.
That's then not a unique or exclusion constraint, but a foreign key constrain. See the postgres documentation for details about the different constrain types.
There is really not much to add here otherwise. I do not see what else you expect to hear from me other than: That's how postgres decided to implement this stuff, which is something that is completely out of diesels control.
Sorry for bothering you again, @weiznich.
I'm wondering if what I'm trying to do here is even possible.
I've just migrated my table from having one long JSON string to splitting that up into multiple tables, and for now I need some backwards compatibility, so I wrote a MySQL query to join the tables back into a JSON string.
So my understanding was that I could write an arbitrary query like this, bind a parameter, so that I can choose which user's data to get, and load the string into a struct deriving from
QueryableByName. That seems so be what's in the example, at least.
the trait bound `query_builder::sql_query::UncheckedBind<SqlQuery, u32, diesel::sql_types::Unsigned<diesel::sql_types::Integer>>: LoadQuery<_, Exhibitions>` is not satisfied
the following implementations were found:
<query_builder::sql_query::UncheckedBind<Query, Value, ST> as LoadQuery<Conn, T>>#[derive(QueryableByName)]
struct Exhibitions {
#[sql_type = "String"]
exhibitions: String,
}
#[post("/sqlconnect/readExhibitions.php", data = "<form>")]
pub fn get_handler(
form: Form<GetHandlerForm>,
conn: Database,
) -> Result<JsonValue, Custom<String>> {
Ok(
match sql_query(include_str!("select_exhibitions.sql"))
.bind::<Unsigned<Integer>, _>(form.cid)
.load::<Exhibitions>(&*conn)
.map_err(|e| Custom(Status::InternalServerError, e.to_string()))?
.pop()
.map(|w| w.exhibitions)
{
Some(json) => serde_json::to_value(json)
.map_err(|e| {
Custom(
Status::InternalServerError,
format!("Failed to serialize Exhibitions: {}", e.to_string()),
)
})?
.into(),
None => json!({
"exhibitions": []
}),
},
)
}load. Since Exhibitions derives from QueryableByName, I thought I could load it from the query, but something is wrong.
Also, the
exhibitions member of the struct refers to the exhibitions column in the table returned by my custom query. And that's a big JSON_OBJECT.
Is it possible to conditionally perform update on conflicts with multiple values? (I'm using postrgres)
Something like this:
#[derive(Insertable)]
struct InsertPost {
id: i64,
date: NaiveDate,
}
let insert_vec = vec![Test{id: 1, date: "2014.05.05"}, Test{id: 2, date: "2013.03.05"}];
diesel::insert_into(tests)
.values(&insert_vec)
.on_conflict(id)
.do_update()
.filter(date.ge(excluded(date)))
.set(date.eq(excluded(date)));I want to insert multiple values and only update the rows that meet a certain criteria, is there any way to acheive this?
The example above does not compile as the type IncompleteDoUpdate type does not contain a filter method.