[ARROW-5428] [C++] Add option to set "read extent" in arrow::io::BufferedInputStream

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.14.0
• Component/s: C++
• External issue URL: https://github.com/apache/arrow/issues/21881
• Assignee: Wes McKinney
• Reporter: Wes McKinney
• Created: 27/May/19 20:01
• Updated: 11/Jan/23 07:40
• Resolved: 11/Jun/19 19:57

Description

I'm looking at simplifying libparquet to use common IO interfaces rather than its own custom ones

The parquet::BufferedInputStream interface has an option to not read beyond a particular number of bytes. For example, if we were reading a 32MB block with 1MB buffering, then we will not consume more than 32MB from the raw InputStream.

This seems like a fairly trivial addition to arrow::io::BufferedInputStream to track total read bytes and do not read beyond the configured extent. We'll have to add a method like set_read_extent