Use serialize_field without requiring static lifetime?

[agentsim]

I'm trying to write a custom Serialize implementation for the results of a database query, where I do not have the table structure upfront (I cannot just use a struct and #[derive(Serialize)]).

I want to write the output in csv, so I'm using the csv crate. Unfortunately csv doesn't support serializing maps (BurntSushi/rust-csv#98), so I'm trying to use serialize_struct. Unfortunately, serialize_field requires the key to be &'static, but the key is the column name which comes from the database at runtime and therefore is not static. I don't see any particular reason that serialize_field must have a static lifetime on the key (I guess performance), is there a way around this?

This is my Serialize implementation thus far:

impl<T> Serialize for ScrubbableRow<T>
where
    T: Scrubber,
{
    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
    where
        S: serde::Serializer,
    {
        let cols = self.row.columns();
        let mut state = serializer.serialize_struct(Some(cols.len()))?;

        for col in cols {
            state.serialize_field(
                col.name(), // Uhoh
                &T::scrub_column(&self.row, col).map_err(serde::ser::Error::custom)?,
            )?;
        }

        state.end()
    }
}

[charrondev]

I'm running into almost the same thing, although in this case it's to serialize to XML and I'm trying to add extra fields that were originally deserialized into a hash map into the struct.

[charrondev]

#708 (comment) this resolved my issue actually. I was able to just use serializeMap() and serializeEntry() instead which doesn't require static lifetimes.

[dtolnay]

I'll close this in favor of BurntSushi/rust-csv#98. I'm not familiar with the csv crate but from skimming the docs I'm sure you could get this working today by handling the conversion from serialize_map to ByteRecord in your own code regardless of what the csv crate does for serialize_map, and then Writer::write_byte_record to write the already serialized records.