BufferedReader.read_line undocumented behavior on new line characters

[einar-hjortdal]

Describe the issue

The documentation says the reader stops at \n but it does not do that: replies from a Redis server always end in \r\n and read_reply does not only stop at \n, it also stops at \r\n and does not return these characters.
This is a problem for me, I will have to rewrite using TcpConn.read instead and providing my own buffer.

Links

Problem was identified here
read_line

[JalonSolov]

Yes, the documentation should say that it reads until EOL, and returns the string without it.

read_line is not undocumented, simply slightly incorrectly documented.

If you want different behaviour, call BufferedReader.read instead of BufferedReader.read_line.

[Casper64]

Maybe I misunderstand the issue, but read_line does also stop at \r\n it's in the code:

// do some checking for whether we hit \r\n or just \n

It will actually check if there was a \r first and then returns the line without \r\n.

Or do you need to check whether the line ended in \r\n?

[einar-hjortdal]

Or do you need to check whether the line ended in \r\n?

Yes I expected read_line to return the line including \n and the characters before it, since it is described as "stopping" there.

I have solved the problem by providing my own buffer and reading implementation on TcpConn.reader (since I am reading a TcpConn stream). Proper documentation would have saved me a lot of time I spent debugging the situation.

[JalonSolov]

All of the read_line/read_lines routines in vlib return the line without the EOL. I don't see this explicitly documented, so the documentation could be improved.

It's done this way because that's what 99.9% of programs expect... the line without the EOL.

And you could still use read instead of read_line.