Changeset - 7dbe020e93fe
[Not reviewed]
default
0 1 0
Mads Kiilerich - 8 years ago 2017-10-03 00:14:40
mads@kiilerich.com
diffs: avoid extra copy of diff when splitting into files

Instead of allocating memory for an extra copy of the whole raw diff, just use
buffer() to give read-only string views of the relevant sections of the big
diff string given as input.
1 file changed with 8 insertions and 7 deletions:
0 comments (0 inline, 0 general)
kallithea/lib/diffs.py
Show inline comments
 
@@ -141,70 +141,69 @@ BIN_FILENODE = 7
 

	
 

	
 
class DiffLimitExceeded(Exception):
 
    pass
 

	
 

	
 
class LimitedDiffContainer(object):
 

	
 
    def __init__(self, diff_limit, cur_diff_size, diff):
 
        self.diff = diff
 
        self.diff_limit = diff_limit
 
        self.cur_diff_size = cur_diff_size
 

	
 
    def __iter__(self):
 
        for l in self.diff:
 
            yield l
 

	
 

	
 
class DiffProcessor(object):
 
    """
 
    Give it a unified or git diff and it returns a list of the files that were
 
    mentioned in the diff together with a dict of meta information that
 
    can be used to render it in a HTML template.
 
    """
 
    _diff_git_re = re.compile('^diff --git', re.MULTILINE)
 
    _chunk_re = re.compile(r'^@@ -(\d+)(?:,(\d+))? \+(\d+)(?:,(\d+))? @@(.*)')
 
    _newline_marker = re.compile(r'^\\ No newline at end of file')
 
    _git_header_re = re.compile(r"""
 
        # has already been split on this:
 
        # ^diff[ ]--git
 
        ^diff[ ]--git
 
            [ ]a/(?P<a_path>.+?)[ ]b/(?P<b_path>.+?)\n
 
        (?:^old[ ]mode[ ](?P<old_mode>\d+)\n
 
           ^new[ ]mode[ ](?P<new_mode>\d+)(?:\n|$))?
 
        (?:^similarity[ ]index[ ](?P<similarity_index>\d+)%\n
 
           ^rename[ ]from[ ](?P<rename_from>.+)\n
 
           ^rename[ ]to[ ](?P<rename_to>.+)(?:\n|$))?
 
        (?:^new[ ]file[ ]mode[ ](?P<new_file_mode>.+)(?:\n|$))?
 
        (?:^deleted[ ]file[ ]mode[ ](?P<deleted_file_mode>.+)(?:\n|$))?
 
        (?:^index[ ](?P<a_blob_id>[0-9A-Fa-f]+)
 
            \.\.(?P<b_blob_id>[0-9A-Fa-f]+)[ ]?(?P<b_mode>.+)?(?:\n|$))?
 
        (?:^(?P<bin_patch>GIT[ ]binary[ ]patch)(?:\n|$))?
 
        (?:^---[ ](a/(?P<a_file>.+?)|/dev/null)\t?(?:\n|$))?
 
        (?:^\+\+\+[ ](b/(?P<b_file>.+?)|/dev/null)\t?(?:\n|$))?
 
    """, re.VERBOSE | re.MULTILINE)
 
    _hg_header_re = re.compile(r"""
 
        # has already been split on this:
 
        # ^diff[ ]--git
 
        ^diff[ ]--git
 
            [ ]a/(?P<a_path>.+?)[ ]b/(?P<b_path>.+?)\n
 
        (?:^old[ ]mode[ ](?P<old_mode>\d+)\n
 
           ^new[ ]mode[ ](?P<new_mode>\d+)(?:\n|$))?
 
        (?:^similarity[ ]index[ ](?P<similarity_index>\d+)%(?:\n|$))?
 
        (?:^rename[ ]from[ ](?P<rename_from>.+)\n
 
           ^rename[ ]to[ ](?P<rename_to>.+)(?:\n|$))?
 
        (?:^copy[ ]from[ ](?P<copy_from>.+)\n
 
           ^copy[ ]to[ ](?P<copy_to>.+)(?:\n|$))?
 
        (?:^new[ ]file[ ]mode[ ](?P<new_file_mode>.+)(?:\n|$))?
 
        (?:^deleted[ ]file[ ]mode[ ](?P<deleted_file_mode>.+)(?:\n|$))?
 
        (?:^index[ ](?P<a_blob_id>[0-9A-Fa-f]+)
 
            \.\.(?P<b_blob_id>[0-9A-Fa-f]+)[ ]?(?P<b_mode>.+)?(?:\n|$))?
 
        (?:^(?P<bin_patch>GIT[ ]binary[ ]patch)(?:\n|$))?
 
        (?:^---[ ](a/(?P<a_file>.+?)|/dev/null)\t?(?:\n|$))?
 
        (?:^\+\+\+[ ](b/(?P<b_file>.+?)|/dev/null)\t?(?:\n|$))?
 
    """, re.VERBOSE | re.MULTILINE)
 

	
 
    # Used for inline highlighter word split, must match the substitutions in _escaper
 
    _token_re = re.compile(r'()(&amp;|&lt;|&gt;|<u>\t</u>|<u class="cr"></u>| <i></i>|\W+?)')
 

	
 
    _escape_re = re.compile(r'(&)|(<)|(>)|(\t)|(\r)|(?<=.)( \n| $)')
 

	
 
    def __init__(self, diff, vcs='hg', diff_limit=None):
 
        """
 
@@ -297,51 +296,53 @@ class DiffProcessor(object):
 
        2. An iterator yielding lines with simple HTML markup.
 
        """
 
        match = None
 
        if self.vcs == 'git':
 
            match = self._git_header_re.match(diff_chunk)
 
        elif self.vcs == 'hg':
 
            match = self._hg_header_re.match(diff_chunk)
 
        if match is None:
 
            raise Exception('diff not recognized as valid %s diff' % self.vcs)
 
        meta_info = match.groupdict()
 
        rest = diff_chunk[match.end():]
 
        if rest and not rest.startswith('@') and not rest.startswith('literal ') and not rest.startswith('delta '):
 
            raise Exception('cannot parse %s diff header: %r followed by %r' % (self.vcs, diff_chunk[:match.end()], rest[:1000]))
 
        difflines = imap(self._escaper, re.findall(r'.*\n|.+$', rest)) # don't split on \r as str.splitlines do
 
        return meta_info, difflines
 

	
 
    def _parse_gitdiff(self, inline_diff=True):
 
        """Parse self._diff and return a list of dicts with meta info and chunks for each file.
 
        If diff is truncated, wrap it in LimitedDiffContainer.
 
        Optionally, do an extra pass and to extra markup of one-liner changes.
 
        """
 
        _files = [] # list of dicts with meta info and chunks
 
        diff_container = lambda arg: arg
 

	
 
        # split the diff in chunks of separate --git a/file b/file chunks
 
        for raw_diff in ('\n' + self._diff).split('\ndiff --git')[1:]:
 
            head, diff = self._get_header(raw_diff)
 
        starts = [m.start() for m in self._diff_git_re.finditer(self._diff)]
 
        starts.append(len(self._diff))
 

	
 
        for start, end in zip(starts, starts[1:]):
 
            head, diff = self._get_header(buffer(self._diff, start, end - start))
 

	
 
            op = None
 
            stats = {
 
                'added': 0,
 
                'deleted': 0,
 
                'binary': False,
 
                'ops': {},
 
            }
 

	
 
            if head['deleted_file_mode']:
 
                op = 'D'
 
                stats['binary'] = True
 
                stats['ops'][DEL_FILENODE] = 'deleted file'
 

	
 
            elif head['new_file_mode']:
 
                op = 'A'
 
                stats['binary'] = True
 
                stats['ops'][NEW_FILENODE] = 'new file %s' % head['new_file_mode']
 
            else:  # modify operation, can be cp, rename, chmod
 
                # CHMOD
 
                if head['new_mode'] and head['old_mode']:
 
                    op = 'M'
 
                    stats['binary'] = True
 
                    stats['ops'][CHMOD_FILENODE] = ('modified file chmod %s => %s'
0 comments (0 inline, 0 general)