Repository navigation
[Feature Request] Create async reader. #509
Description
Reading data from a stream is potentially a long running I/O bound operation.
Create IExcelDataReader.AsDataSetAsync (or ExcelReaderFactory.CreateReaderAsync, if the file is read at this point).
Reacted by Joakim Malmberg, perason, kronic, ECK, LIN, oof, ENZ, igelineau, Dawid Dyrcz, Emanuel and 1 more
Activity
appel1 commentedon Oct 4, 2026
Async factory methods would be relatively straightforward to add: non-seekable streams could be buffered asynchronously before initializing the reader. Reader initialization and parsing would still be synchronous after buffering completes.
A fully async reader would primarily benefit seekable, I/O-backed streams, such as files on disk or network shares. Non-seekable input would already be buffered in memory, so subsequent cell reading would be CPU-bound rather than waiting for source I/O.
Both single-pass and two-pass modes could support async reads, including the initial scan. However, that would require substantial changes throughout initialization and parsing, and some ZIP operations would remain synchronous on our current target frameworks. The question is whether the benefit for file-backed reading justifies that complexity.
Perhaps we should start with async factory methods and leave async reader methods and AsDataSetAsync for a separate proposal. That would address non-seekable download waits without a large parser refactor. Without async row reading and navigation, AsDataSetAsync would only wrap synchronous work rather than provide meaningful async I/O.