Add Upgrade Guide for DataFusion 46.0.0

[alamb]

Which issue does this PR close?

• Related to Release DataFusion 46.0.0 #14123
• Related to Introduce unified DataSourceExec for provided datasources, remove ParquetExec, CsvExec, etc #14224

Rationale for this change

As we have discussed in tickets, it would be good to have help users upgrade to new DataFusion versions, especially when major / disruptive API changes like #14224 are included

• Related to [DISCUSSION] Making it easier to use DataFusion (lessons from GlareDB) #13525
• Related to [DISCUSS] More extensive pre-release testing #13661

What changes are included in this PR?

1. Add a upgrade guide section:
   
2. Add details on the upgrade to 46.0.0

Are these changes tested?

I build the docs manually

Are there any user-facing changes?

[alamb]

I added this section based on the report from @shehabgamin in #14123 (comment)

@shehabgamin any chance you have an example of the kind of error you saw?

[shehabgamin]

@alamb Yes, users may see an error similar to something like this (depending on which UDF they are calling .invoke() or .invoke_batch() with):

This feature is not implemented: Function concat does not implement invoke but called

Here is an example of resolving the error:

Old Code:

impl ScalarUDFImpl for SparkConcat {
...
    fn invoke_batch(&self, args: &[ColumnarValue], number_rows: usize) -> Result<ColumnarValue> {
        if args
            .iter()
            .any(|arg| matches!(arg.data_type(), DataType::List(_)))
        {
            ArrayConcat::new().invoke_batch(args, number_rows)
        } else {
            ConcatFunc::new().invoke_batch(args, number_rows)
        }
    }
}

New Code:

impl ScalarUDFImpl for SparkConcat {
...
    fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> {
        if args
            .args
            .iter()
            .any(|arg| matches!(arg.data_type(), DataType::List(_)))
        {
            ArrayConcat::new().invoke_with_args(args)
        } else {
            ConcatFunc::new().invoke_with_args(args)
        }
    }
}

Changes:
1.
From:
fn invoke_batch(&self, args: &[ColumnarValue], number_rows: usize) OR
fn invoke(&self, args: &[ColumnarValue])
To:
fn invoke_with_args(&self, args: ScalarFunctionArgs)

2.
From:
.invoke_batch(args, number_rows) OR
.invoke(args)
To:
.invoke_with_args(args)

[xudong963]

This will be definitely helpful for the user's upgrade, thank you @alamb

[comphead]

Thanks @alamb for proceeding with this. Imho we still need to modify contribution guide to let contributors to maintain the upgration guide. Leaving this to one person is might be a huge burden

[alamb]

Thanks @alamb for proceeding with this. Imho we still need to modify contribution guide to let contributors to maintain the upgration guide. Leaving this to one person is might be a huge burden

Sounds good . Maybe once we have an example of what an upgrade guide looks like it will be easier to add / point people to an example

In general trying to change people's behavior with documentation is a bit diecy in my opinion

[comphead]

Thanks @alamb for proceeding with this. Imho we still need to modify contribution guide to let contributors to maintain the upgration guide. Leaving this to one person is might be a huge burden

Sounds good . Maybe once we have an example of what an upgrade guide looks like it will be easier to add / point people to an example

In general trying to change people's behavior with documentation is a bit diecy in my opinion

Yeah, example is awesome. The upgrade guide would be totally voluntarily but if anyone feels their changes are breaking and wanna give a downstream user a hint how to migrate they can refer to the example as you propose.