Make get_page_stream "spawn-friendly"?

[dantengsky]

For code like tokio::spawn(get_page_stream(col_chunk_meta, reader, buffer, page_filter)), which requires the Future being spawned should be Send, rustc complains about the parameter page_filter as follows:

type Arc<dyn for<'r, 's> Fn(&'r ColumnDescriptor, &'s DataPageHeader) -> bool> which is not Send

---

Currently the signature of get_page_stream and the type def of PageFilter are:

parquet2/src/read/page_stream.rs

Lines 16 to 21 in 7661cd4

	pub async fn get_page_stream<'a, RR: AsyncRead + Unpin + Send + AsyncSeek>(
	column_metadata: &'a ColumnChunkMetaData,
	reader: &'a mut RR,
	buffer: Vec<u8>,
	pages_filter: PageFilter,
	) -> Result<impl Stream<Item = Result<CompressedDataPage>> + 'a> {

parquet2/src/read/page_iterator.rs

Line 15 in 5ec7e09

pub type PageFilter = Arc<dyn Fn(&ColumnDescriptor, &DataPageHeader) -> bool>;

---

• Should other APIs be used to get pages asynchronously in this case?

• Is it possible/proper to change the type of PageFilter to

  pub type PageFilter = Arc<dyn Fn(&ColumnDescriptor, &DataPageHeader) -> bool + Send + Sync>; 

  so that PageFilter will be Send

[jorgecarleitao]

Great observation. Thanks for it and for the fix!