How do I round-trip a CSV file while accessing only some of the rows?

[vi]

I want to use the convenience of csv::Reader::deserialize automatically mapping columns to struct fields while also preserving unknown columns found in CSV as is.

Currently I use a hack like this:

#[derive(Serialize,Deserialize)]
struct Entry {
    interesting_column: f64,
    another_interesting_column: f64,

    #[serde(flatten)]
    just_copy_this_from_input_to_output: BTreeMap<String, serde_json::Value>,
}

, but preserving other columns is tricky - I need to write csv manually, without serializing. Also using String instead of serde_json::Value makes it reject numbers in those unknown columns.

Is there some better way to handle this (without manually interpreting csv header)?

Maybe there is other deserialisation function that also gives a csv::ByteRecord / csv::StringRecord with the rest of fields? Or I can somehow #[serde(flatten)] the record type directly instead of resorting to a map?

[BurntSushi]

I'm on mobile or else I would write it out, but have you looked at this? https://docs.rs/csv/latest/csv/struct.ByteRecord.html#method.deserialize

I think that should be able to do what you want.

Otherwise you might need to abdicate serde.