~bzr-pqm/bzr/bzr.dev

Committer: John Arbash Meinel
Date: 2008-12-11 02:02:07 UTC
mto: This revision was merged to the branch mainline in revision 3895.
Revision ID: john@arbash-meinel.com-20081211020207-rrgdcyqc344zo5q1

Change name to 'chunks_to_lines', and find an optimized form.

It is a little bit ugly, but it is faster than join & split, and means
we get to leave the strings untouched.

files modified:
bzrlib/osutils.py

bzrlib/tests/test_osutils.py

Show diffs side-by-side

added added

removed removed

bzrlib/osutils.py

820

return pathjoin(*p)

821

822

823

def chunked_to_lines(chunks):

823

def chunks_to_lines(chunks):

824

"""Ensure that chunks is split cleanly into lines.

825

826

Each entry in the result should contain a single newline at the end. Except

829

:param chunks: An iterable of strings

830

:return: A list of strings.

831

"""

832

# Optimize for a very common case when chunks are already lines

833

def fail():

834

raise IndexError

835

try:

836

# This is a bit ugly, but is the fastest way to check if all of the

837

# chunks are individual lines.

838

# You can't use function calls like .count(), .index(), or endswith()

839

# because they incur too much python overhead.

840

# It works because

841

# if chunk is an empty string, it will raise IndexError, which will

842

# be caught.

843

# if chunk doesn't end with '\n' then we hit fail()

844

# if there is more than one '\n' then we hit fail()

845

# timing shows this loop to take 2.58ms rather than 3.18ms for

846

# split_lines(''.join(chunks))

847

# Further, it means we get to preserve the original lines, rather than

848

# expanding memory

849

[(chunk[-1] == '\n' and '\n' not in chunk[:-1]) or fail()

850

for chunk in chunks]

851

return chunks

852

except IndexError:

853

pass

832

854

return split_lines(''.join(chunks))

833

855

834

856

Older »