Dynamic Table Name， a lot of tables with the same structure, it is not appropriate to define many structures with macros

[langzime]

Motivation

Proposed Solutions

Additional Information

[billy1624]

Hey @langzime, is this a follow up question of

• custom table names like user1, user2,user... ， They all have same columns, but table name #782

[billy1624]

Yeah, this is the tradeoff. Defining the repeated structs over and over again OR defining repeated struct with macro. Define with macro will improve the readability and avoid repeated code but it's harder to develop upon as not all IDEs are able to expand the macros.

[langzime]

Yes, i tried it. A lot of macros cause ide very laggy and a lot of extra code

[billy1624]

From my experience, it's just a one-time long compilation as 200+ entities were added at once. Once it's compiled it will never be recompiled on the same machine given the fact that entity file is unchanged.

[langzime]

I have 500+ ...

[langzime]

In cases like this, I currently use both mysql_async

[billy1624]

I think there are three options:

1. Fork SeaORM and change this part, making it to be provided in runtime:

   sea-orm/src/tests_cfg/cake_expanded.rs

   Lines 7 to 11 in 1ea8b45

   	impl EntityName for Entity {
   	fn table_name(&self) -> &str {
   	"cake"
   	}
   	}

2. Construct SQL statement with SeaQuery then pass it to SeaORM for execution: https://github.com/SeaQL/starfish-ql/blob/459a71609785f74739d4b5ba9a2983af0cedfae6/starfish/starfish-core/src/query/mod.rs#L290-L303

[langzime]

Thank you very much, I'll try it sometime.
Do you have plans to support this feature?

[billy1624]

Do you have plans to support this feature?

It's not on our plan. But if you have an good proposal we're willing to consider :D

[tyt2y3]

Actually, it's possible to have an Entity that has a state, so your Entity is no longer a unit struct, but a struct Entity(String).

[tyt2y3]

Although I am not sure are there other blockers to this approach. But yeah at least it was designed that way, the table_name takes a self

[billy1624]

From table_name perspective, it's designed with runtime in mind.

sea-orm/src/entity/base_entity.rs

Lines 23 to 24 in 6bc7669

	/// Get the name of the table
	fn table_name(&self) -> &str;

However, when constructing query. It doesn't take &self. So, it's solely rely on the Entity defined in compile-time.

Also, EntityName must be Copy but String isn't. So, pub struct Entity(pub String) wouldn't compile.

sea-orm/src/entity/base_entity.rs

Lines 10 to 24 in 6bc7669

	/// Ensure the identifier for an Entity can be converted to a static str
	pub trait IdenStatic: Iden + Copy + Debug + 'static {
	/// Method to call to get the static string identity
	fn as_str(&self) -> &str;
	}
	/// A Trait for mapping an Entity to a database table
	pub trait EntityName: IdenStatic + Default {
	/// Method to get the name for the schema, defaults to [Option::None] if not set
	fn schema_name(&self) -> Option<&str> {
	None
	}
	/// Get the name of the table
	fn table_name(&self) -> &str;

[tyt2y3]

Then probably a lazy_static with Entity(&'static str) would work

[pedrettin]

This issue is blocking us from using sea ORM when the table_name is set at runtime. We create many SQL tables with the same struct dynamically and we don't know the names of the tables before hand.

Setting a variable in Entity won't work as the find() method is called statically on the Entity object, so creating a new entity object with the right "table_name" isn't useful.

[tyt2y3]

Then probably your best bet for now is using SeaQuery, everything is dynamic there.

[penso]

Got a similar issue where I'd like to set the table_name dynamically with a prefix based on a more global app context, I'll have multiple copies of the same tables. But it also means ideally to be able to run migrations multiple times for all the table names.

[theelderbeever]

Has anyone figured out anything on this? We have the situation where we have a set of tables that are individually generated at runtime with a predetermined naming pattern. I looked into SeaQuery however, Query::select().from(<IntoTableRef>) is super restrictive. Is there anyway to just pass a String at runtime for the table that is supposed to be queried?

[theelderbeever]

@penso actually this seems even easier

use sea_query::*;

pub enum Metrics {
    TableName(String),
    Timestamp,
    MetricId,
    Account,
    IntervalMs,
    Counter,
    Gauge,
    MetricTagIds,
}

// Mapping between Enum variant and its corresponding string value
impl Iden for Metrics {
    fn unquoted(&self, s: &mut dyn std::fmt::Write) {
        write!(
            s,
            "{}",
            match self {
                Self::TableName(name) => name,
                Self::Timestamp => "timestamp",
                Self::MetricId => "metric_id",
                Self::Account => "account",
                Self::IntervalMs => "interval_ms",
                Self::Counter => "counter",
                Self::Gauge => "gauge",
                Self::MetricTagIds => "metric_tag_ids",
            }
        )
        .unwrap();
    }
}

fn main() {
    let select = Query::select()
        .columns([Metrics::Account, Metrics::Gauge])
        .from(Metrics::TableName("metric_1".to_string()))
        .build(PostgresQueryBuilder);
    println!("{}", select.0);
}

[tyt2y3]

I think there is a Copy constraint, so only &'static str or usize or ShortString can be used. But yes your idea is right. It's the reason why Iden trait does not simply return &str.

[theelderbeever]

@tyt2y3 the last example I sent compiles and works fwiw. Not sure if I will run into problems later?

[penso]

That's a nice short and concise example, I wonder how it'd work for migrations when I want to write a single migration but have, let's say, 10 of the same tables I need to run migrations on. In your scenario that'd be account and gauge when you create tables through migration files.

[penso]

Also as for my need, I use find() everywhere, rewriting those into sea-query would be such a pain

[tyt2y3]

Oh wow, that’s great news. Not really, once it’s Arced it’ll work as usual

…

________________________________ From: Taylor Beever ***@***.***> Sent: Saturday, April 27, 2024 12:12:12 AM To: SeaQL/sea-orm ***@***.***> Cc: Chris Tsang ***@***.***>; Mention ***@***.***> Subject: Re: [SeaQL/sea-orm] Dynamic Table Name， a lot of tables with the same structure, it is not appropriate to define many structures with macros (Discussion #1202) @tyt2y3<https://github.com/tyt2y3> the last example I sent compiles and works fwiw. Not sure if I will run into problems later? — Reply to this email directly, view it on GitHub<#1202 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AANTHCF4ZXGB73TCXT2IMVLY7LNMZAVCNFSM6AAAAAA2ZD2E5WVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4TENBSGA3DG>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[BB-fat]

In my case the table name was calculated dynamically at runtime.
I wish there was a usage like this:

Entity::from_table("my_table").find()