lines & unlines

[chemist]

In Bytestring, String and Text EOL always '\n', but in windows its \r\n
how about separate version for lines, unlines?
lines can be universal, and unlines like
unlinesUniversal :: EOL -> [ByteString] -> ByteString

Its not critical, but can be useful.

[dcoutts]

I don't like this very much, even on Windows if you open a file in text mode then you'll get \r\n conversion. The more general thing is then splitting, which we should do by integrating or extending the split package for ByteString (and Text).

Leaving this open for discussion though.

[rahulmutt]

It would be really useful to have unlinesUniversal or even patch unlines to handle \r too because the current implementation violates the principle of least astonishment.

The intuition any beginner Haskeller gets from using the lines function for String is that it returns proper lines without the \r character. You'll write a code assuming that ByteString's version of lines works in a cross-platform way.

I'd be happy to send a PR if the maintainers are fine with it. I wasted half a day finding a bug in a parser only to find that it was working under the assumption that the lines function strips away the \r as well for Windows. I hope to prevent this for anyone else in the future.

There should at least be a line in the documentation in bold letters saying that lines doesn't strip off \r because that's very important to know.

[Bodigrim]

The intuition any beginner Haskeller gets from using the lines function for String is that it returns proper lines without the \r character.

I cannot find any special treatment for \r in Data.List.lines:

> lines "foo\r\nbar"
["foo\r","bar"]

So Data.ByteString.lines is in line with its Prelude counterpart. This behaviour has been recently reflected in the documentation. A general splitOn function is discussed in #100.

I'm in favor of closing this.

[sjakobi]

I agree with closing this issue.