How to exclude U+2028 from line separators in Python when reading file?

Question

I have a file in UTF-8, where some lines contain the U+2028 Line Separator character (http://www.fileformat.info/info/unicode/char/2028/index.htm). I don't want it to be treated as a line break when I read lines from the file. Is there a way to exclude it from separators when I iterate over the file or use readlines()? (Besides reading the entire file into a string and then splitting by \n.) Thank you!

Answer 1

I can't duplicate this behaviour in python 2.5, 2.6 or 3.0 on mac os x - U+2028 is always treated as non-endline. Could you go into more detail about where you see this error?

That said, here is a subclass of the "file" class that might do what you want:

#/usr/bin/python
# -*- coding: utf-8 -*-
class MyFile (file):
    def __init__(self, *arg, **kwarg):
        file.__init__(self, *arg, **kwarg)
        self.EOF = False
    def next(self, catchEOF = False):
        if self.EOF:
            raise StopIteration("End of file")
        try:
            nextLine= file.next(self)
        except StopIteration:
            self.EOF = True
            if not catchEOF:
                raise
            return ""
        if nextLine.decode("utf8")[-1] == u'\u2028':
            return nextLine+self.next(catchEOF = True)
        else:
            return nextLine

A = MyFile("someUnicode.txt")
for line in A:
    print line.strip("\n").decode("utf8")

Answer 2

I couldn't reproduce that behavior but here's a naive solution that just merges readline results until they don't end with U+2028.

#!/usr/bin/env python

from __future__ import with_statement

def my_readlines(f):
  buf = u""
  for line in f.readlines():
    uline = line.decode('utf8')
    buf += uline
    if uline[-1] != u'\u2028':
      yield buf
      buf = u""
  if buf:
    yield buf

with open("in.txt", "rb") as fin:
  for l in my_readlines(fin):
    print l

Answer 3

Thanks to everyone for answering. I think I know why you might not have been able to replicate this.I just realized that it happens if I decode the file when opening, as in:

f = codecs.open(filename, encoding='utf-8')
for line in f:
    print line

The lines are not separated on u2028, if I open the file first and then decode individual lines:

f = open(filename)
for line in f:
    print line.decode("utf8")

(I'm using Python 2.6 on Windows. The file was originally UTF16LE and then it was converted into UTF8).

This is very interesting, I guess I won't be using codecs.open much from now on :-).

Answer 4

If you use Python 3.0 (note that I don't, so I can't test), according to the documentation you can pass an optional newline parameter to open to specifify which line seperator to use. However, the documentation doesn't mention U+2028 at all (it only mentions \r, \n, and \r\n as line seperators), so it's actually a suprise to me that this even occurs (although I can confirm this even with Python 2.6).

Answer 5

The codecs module is doing the RIGHT thing. U+2028 is named "LINE SEPARATOR" with the comment "may be used to represent this semantic unambiguously". So treating it as a line separator is sensible.

Presumably the creator would not have put the U+2028 characters there without good reason ... does the file have u"\n" as well? Why do you want lines not to be split on U+2028?