Unbuffered Stream Reader

[mbaker3]

Adds UnbufferedStreamReader. A stream reader that does not buffer data from the backing BaseStream.
This implementation is less performant than StreamReader but doesn't read ahead on the underlying stream ensuring the BaseStream's position to remains lockstep with the characters read out of UnbufferedStreamReader.

This stream is useful when reading files with mixed encoding or where the BaseStream needs to be advanced the minimum amount required.

What is the current behaviour?

StreamReader fills a buffer of characters when responding to read requests. The characters buffer is filled by repeatedly reading blocks of its BaseStream and converting them into characters until the character buffer is full.

This means that the BaseStream is advanced beyond the point that the StreamReader has been read.

What is the new behaviour?

UnbufferedStreamReader effectively functions the same as StreamReader with a byte and character buffer size set to 1. The UnbufferedStreamReader reads one byte at a time out of BaseStream until it has enough to resolve a full character. That one character is then evaluated by the read method to decide whether to continue reading or return.

Note: The implementation does not currently cover the complete features set of StreamReader. New features will be developed as needed. This effort is captured in #87

Why don't you just calculate actual the BaseStream position based on StreamReader's unused buffer contents?

TL;DR

• Invalid bytes are replaced with a pre-defined fallback character which may not equal the byte size of the invalid bytes.
• The reflection relies heavily on Microsoft not changing the way StreamReader is implemented
• It does not work for variable-byte encodings (Ex: UTF-7)

I went down that path. It's not possible. This StackOverflow post covers the issue in detail. Read the linked post + all comments to understand.

If you really want to dig into this read through StreamReader's source: https://referencesource.microsoft.com/#mscorlib/system/io/streamreader.cs,b5fe1efcec14de32

What issues does this resolve?

None, This class allows developers to continue reading a BaseStream beyond the lifecycle of the UnbufferedStreamReader.

What PRs does this depend on?

• None

Does this introduce a breaking change?

• Yes
• No

[jkeon]

Minor comments, otherwise looks good

[jkeon]

Should we link that SO post in here too? Useful background info.

[mbaker3]

What SO post, the one about StreamReader position issues?
I don't think it's necessary to teach the user why position can't be calculated on a different class.

[jkeon]

I thought it was the reason this class existed. Normally you'd use StreamReader but sometimes you can't and good background reading on why that might be the case is in that SO post.

But if you don't agree, I'm a-ok with that too.

[mbaker3]

Feedback addressed