Create a `DataFrame` from a `Vec` of structs with automatically derived schema

[oersted]

Problem description

Say we have a trait like this (there's probably a better name for it)

trait GetSchema {
    fn get_schema() -> Schema;
}

And we have its corresponding derive macro for structs.

It would be nice for a user to create a DataFrame like this:

#[derive(GetSchema)]
struct Record {
    ...
}

let records: Vec<Record> = vec![...];
let df: DataFrame = records.into();

As opposed to having to parse the structs into Rows and manually specifying the schema.

Let me know if I'm missing an existing API that does something similar already.

Perhaps it is possible to leverage the existing interop with serve? I'm not sure since the serde and serde-lazy features are not documented at the moment.

[ritchie46]

You can try this: https://github.com/DataEngineeringLabs/arrow2-convert

[mdekstrand]

arrow2-convert works reasonably well, but is built on arrow2 directly, so its usage path now that Polars has vendored arrow2/parquet2 is unclear.