FileSource and DataSource traits require deep copies

[alamb]

Is your feature request related to a problem or challenge?

While working on various upgrade PRs and preparing for the DataFusion 46 release, I have noticed something I would like to change before we release

The FileSource and DataSource traits were introduced in the datasource refactor

• Introduce unified DataSourceExec for provided datasources, remove ParquetExec, CsvExec, etc #14224

They have APIs to update the underlying source in a few ways, but the APIs require cloning. For example, FileSource looks like this:

/// Common behaviors that every file format needs to implement.
///
/// See initialization examples on `ParquetSource`, `CsvSource`
pub trait FileSource: Send + Sync {
...
    /// Initialize new type with batch size configuration
    fn with_batch_size(&self, batch_size: usize) -> Arc<dyn FileSource>;
...
}

The only way to implement with_batch_size is to (deep) clone the object

    fn with_batch_size(&self, batch_size: usize) -> Arc<dyn FileSource> {
        let mut conf = self.clone();
        conf.batch_size = Some(batch_size);
        Arc::new(conf)
    }

datafusion/datafusion/core/src/datasource/physical_plan/csv.rs

Lines 584 to 588 in 1ae06a4

	fn with_batch_size(&self, batch_size: usize) -> Arc<dyn FileSource> {
	let mut conf = self.clone();
	conf.batch_size = Some(batch_size);
	Arc::new(conf)
	}

Describe the solution you'd like

I would like to avoid having to deep clone the object

Describe alternatives you've considered

No response

Additional context

No response

[alamb]

Maybe we can try a mutable API instead of a builder style API 🤔

[Standing-Man]

take