#!/usr/bin/env python # -*- coding: utf-8 -*- """ Term based tool to view *colored*, *incremental* diff in a *Git/Mercurial/Svn* workspace or from stdin, with *side by side* and *auto pager* support. Requires python (>= 2.5.0) and ``less``. """ import sys import os import re import signal import subprocess import select import difflib META_INFO = { 'version' : '1.1', 'license' : 'BSD-3', 'author' : 'Matthew Wang', 'email' : 'mattwyl(@)gmail(.)com', 'url' : 'https://github.com/ymattw/ydiff', 'keywords' : 'colored incremental side-by-side diff', 'description' : ('View colored, incremental diff in a workspace or from ' 'stdin, with side by side and auto pager support') } if sys.hexversion < 0x02050000: raise SystemExit('*** Requires python >= 2.5.0') # pragma: no cover # Python < 2.6 does not have next() try: next except NameError: def next(obj): return obj.next() try: unicode except NameError: unicode = str COLORS = { 'reset' : '\x1b[0m', 'underline' : '\x1b[4m', 'reverse' : '\x1b[7m', 'red' : '\x1b[31m', 'green' : '\x1b[32m', 'yellow' : '\x1b[33m', 'blue' : '\x1b[34m', 'magenta' : '\x1b[35m', 'cyan' : '\x1b[36m', 'lightred' : '\x1b[1;31m', 'lightgreen' : '\x1b[1;32m', 'lightyellow' : '\x1b[1;33m', 'lightblue' : '\x1b[1;34m', 'lightmagenta' : '\x1b[1;35m', 'lightcyan' : '\x1b[1;36m', } # Keys for revision control probe, diff and log with diff VCS_INFO = { 'Git': { 'probe': ['git', 'rev-parse'], 'diff': ['git', 'diff', '--no-ext-diff'], 'log': ['git', 'log', '--patch'], }, 'Mercurial': { 'probe': ['hg', 'summary'], 'diff': ['hg', 'diff'], 'log': ['hg', 'log', '--patch'], }, 'Svn': { 'probe': ['svn', 'info'], 'diff': ['svn', 'diff'], 'log': ['svn', 'log', '--diff', '--use-merge-history'], }, } def colorize(text, start_color, end_color='reset'): return COLORS[start_color] + text + COLORS[end_color] def strsplit(text, width): r"""strsplit() splits a given string into two substrings, respecting the escape sequences (in a global var COLORS). It returns 3-tuple: (first string, second string, number of visible chars in the first string). If some color was active at the splitting point, then the first string is appended with the resetting sequence, and the second string is prefixed with all active colors. """ first = '' second = '' found_colors = [] chars_cnt = 0 bytes_cnt = 0 while text: # First of all, check if current string begins with any escape # sequence. append_len = 0 for color in COLORS: if text.startswith(COLORS[color]): if color == 'reset': found_colors = [] else: found_colors.append(color) append_len = len(COLORS[color]) break if append_len == 0: # Current string does not start with any escape sequence, so, # either add one more visible char to the "first" string, or # break if that string is already large enough. if chars_cnt >= width: break chars_cnt += 1 append_len = 1 first += text[:append_len] text = text[append_len:] bytes_cnt += append_len second = text # If the first string has some active colors at the splitting point, # reset it and append the same colors to the second string if found_colors: first += COLORS['reset'] for color in found_colors: second = COLORS[color] + second return (first, second, chars_cnt) def strtrim(text, width, wrap_char, pad): r"""strtrim() trims given string respecting the escape sequences (using strsplit), so that if text is larger than width, it's trimmed to have width-1 chars plus wrap_char. Additionally, if pad is True, short strings are padded with space to have exactly needed width. Returns resulting string. """ text, _, tlen = strsplit(text, width + 1) if tlen > width: text, _, _ = strsplit(text, width - 1) text += wrap_char elif pad: # The string is short enough, but it might need to be padded. text = '%s%*s' % (text, width - tlen, '') return text class Hunk(object): def __init__(self, hunk_headers, hunk_meta, old_addr, new_addr): self._hunk_headers = hunk_headers self._hunk_meta = hunk_meta self._old_addr = old_addr # tuple (start, offset) self._new_addr = new_addr # tuple (start, offset) self._hunk_list = [] # list of tuple (attr, line) def append(self, hunk_line): """hunk_line is a 2-element tuple: (attr, text), where attr is: '-': old, '+': new, ' ': common """ self._hunk_list.append(hunk_line) def mdiff(self): r"""The difflib._mdiff() function returns an interator which returns a tuple: (from line tuple, to line tuple, boolean flag) from/to line tuple -- (line num, line text) line num -- integer or None (to indicate a context separation) line text -- original line text with following markers inserted: '\0+' -- marks start of added text '\0-' -- marks start of deleted text '\0^' -- marks start of changed text '\1' -- marks end of added/deleted/changed text boolean flag -- None indicates context separation, True indicates either "from" or "to" line contains a change, otherwise False. """ return difflib._mdiff(self._get_old_text(), self._get_new_text()) def _get_old_text(self): return [line for (attr, line) in self._hunk_list if attr != '+'] def _get_new_text(self): return [line for (attr, line) in self._hunk_list if attr != '-'] def is_completed(self): old_completed = self._old_addr[1] == len(self._get_old_text()) new_completed = self._new_addr[1] == len(self._get_new_text()) return old_completed and new_completed class UnifiedDiff(object): def __init__(self, headers, old_path, new_path, hunks): self._headers = headers self._old_path = old_path self._new_path = new_path self._hunks = hunks def is_old_path(self, line): return line.startswith('--- ') def is_new_path(self, line): return line.startswith('+++ ') def is_hunk_meta(self, line): """Minimal valid hunk meta is like '@@ -1 +1 @@', note extra chars might occur after the ending @@, e.g. in git log. '## ' usually indicates svn property changes in output from `svn log --diff` """ return (line.startswith('@@ -') and line.find(' @@') >= 8 or line.startswith('## -') and line.find(' ##') >= 8) def parse_hunk_meta(self, hunk_meta): # @@ -3,7 +3,6 @@ a = hunk_meta.split()[1].split(',') # -3 7 if len(a) > 1: old_addr = (int(a[0][1:]), int(a[1])) else: # @@ -1 +1,2 @@ old_addr = (int(a[0][1:]), 1) b = hunk_meta.split()[2].split(',') # +3 6 if len(b) > 1: new_addr = (int(b[0][1:]), int(b[1])) else: # @@ -0,0 +1 @@ new_addr = (int(b[0][1:]), 1) return (old_addr, new_addr) def parse_hunk_line(self, line): return (line[0], line[1:]) def is_old(self, line): """Exclude old path and header line from svn log --diff output, allow '----' likely to see in diff from yaml file """ return (line.startswith('-') and not self.is_old_path(line) and not re.match(r'^-{72}$', line.rstrip())) def is_new(self, line): return line.startswith('+') and not self.is_new_path(line) def is_common(self, line): return line.startswith(' ') def is_eof(self, line): # \ No newline at end of file # \ No newline at end of property return line.startswith(r'\ No newline at end of') def is_only_in_dir(self, line): return line.startswith('Only in ') def is_binary_differ(self, line): return re.match('^Binary files .* differ$', line.rstrip()) class PatchStream(object): def __init__(self, diff_hdl): self._diff_hdl = diff_hdl self._stream_header_size = 0 self._stream_header = [] # Test whether stream is empty by read 1 line line = self._diff_hdl.readline() if not line: self._is_empty = True else: self._stream_header.append(line) self._stream_header_size += 1 self._is_empty = False def is_empty(self): return self._is_empty def read_stream_header(self, stream_header_size): """Returns a small chunk for patch type detect, suppose to call once""" for i in range(1, stream_header_size): line = self._diff_hdl.readline() if not line: break self._stream_header.append(line) self._stream_header_size += 1 return self._stream_header def __iter__(self): for line in self._stream_header: yield line for line in self._diff_hdl: yield line class PatchStreamForwarder(object): """A blocking stream forwarder use `select` and line buffered mode. Feed input stream to a diff format translator and read output stream from it. Note input stream is non-seekable, and upstream has eaten some lines. """ def __init__(self, istream, translator): assert isinstance(istream, PatchStream) assert isinstance(translator, subprocess.Popen) self._istream = iter(istream) self._in = translator.stdin self._out = translator.stdout def _can_read(self, timeout=0): return select.select([self._out.fileno()], [], [], timeout)[0] def _forward_line(self): try: line = next(self._istream) self._in.write(line) except StopIteration: self._in.close() def __iter__(self): while True: if self._can_read(): line = self._out.readline() if line: yield line else: return elif not self._in.closed: self._forward_line() class DiffParser(object): def __init__(self, stream): header = [decode(line) for line in stream.read_stream_header(100)] size = len(header) if size >= 4 and (header[0].startswith('*** ') and header[1].startswith('--- ') and header[2].rstrip() == '***************' and header[3].startswith('*** ') and header[3].rstrip().endswith(' ****')): # For context diff, try use `filterdiff` to translate it to unified # format and provide a new stream # self._type = 'context' try: # Use line buffered mode so that to readline() in block mode self._translator = subprocess.Popen( ['filterdiff', '--format=unified'], stdin=subprocess.PIPE, stdout=subprocess.PIPE, bufsize=1) except OSError: raise SystemExit('*** Context diff support depends on ' 'filterdiff') self._stream = PatchStreamForwarder(stream, self._translator) return for n in range(size): if (header[n].startswith('--- ') and (n < size - 1) and header[n + 1].startswith('+++ ')): self._type = 'unified' self._stream = stream break else: # `filterdiff` translates unknown diff to nothing, fall through to # unified diff give ydiff a chance to show everything as headers # sys.stderr.write("*** unknown format, fall through to 'unified'\n") self._type = 'unified' self._stream = stream def get_diff_generator(self): """parse all diff lines, construct a list of UnifiedDiff objects""" diff = UnifiedDiff([], None, None, []) headers = [] for line in self._stream: line = decode(line) if diff.is_old_path(line): # This is a new diff when current hunk is not yet genreated or # is completed. We yield previous diff if exists and construct # a new one for this case. Otherwise it's acutally an 'old' # line starts with '--- '. # if (not diff._hunks or diff._hunks[-1].is_completed()): if diff._old_path and diff._new_path and diff._hunks: yield diff diff = UnifiedDiff(headers, line, None, []) headers = [] else: diff._hunks[-1].append(diff.parse_hunk_line(line)) elif diff.is_new_path(line) and diff._old_path: if not diff._new_path: diff._new_path = line else: diff._hunks[-1].append(diff.parse_hunk_line(line)) elif diff.is_hunk_meta(line): hunk_meta = line try: old_addr, new_addr = diff.parse_hunk_meta(hunk_meta) except (IndexError, ValueError): raise RuntimeError('invalid hunk meta: %s' % hunk_meta) hunk = Hunk(headers, hunk_meta, old_addr, new_addr) headers = [] diff._hunks.append(hunk) elif diff._hunks and not headers and (diff.is_old(line) or diff.is_new(line) or diff.is_common(line)): diff._hunks[-1].append(diff.parse_hunk_line(line)) elif diff.is_eof(line): # ignore pass elif diff.is_only_in_dir(line) or diff.is_binary_differ(line): # 'Only in foo:' and 'Binary files ... differ' are considered # as separate diffs, so yield current diff, then this line # if diff._old_path and diff._new_path and diff._hunks: # Current diff is comppletely constructed yield diff headers.append(line) yield UnifiedDiff(headers, '', '', []) headers = [] diff = UnifiedDiff([], None, None, []) else: # All other non-recognized lines are considered as headers or # hunk headers respectively # headers.append(line) # Validate and yield the last patch set if it is not yielded yet if diff._old_path: assert diff._new_path is not None if diff._hunks: assert len(diff._hunks[-1]._hunk_meta) > 0 assert len(diff._hunks[-1]._hunk_list) > 0 yield diff if headers: # Tolerate dangling headers, just yield a UnifiedDiff object with # only header lines # yield UnifiedDiff(headers, '', '', []) class DiffMarker(object): def __init__(self, side_by_side=False, width=0, tab_width=8, wrap=False): self._side_by_side = side_by_side self._width = width self._tab_width = tab_width self._wrap = wrap def markup(self, diff): """Returns a generator""" if self._side_by_side: for line in self._markup_side_by_side(diff): yield line else: for line in self._markup_traditional(diff): yield line def _markup_traditional(self, diff): """Returns a generator""" for line in diff._headers: yield self._markup_header(line) yield self._markup_old_path(diff._old_path) yield self._markup_new_path(diff._new_path) for hunk in diff._hunks: for hunk_header in hunk._hunk_headers: yield self._markup_hunk_header(hunk_header) yield self._markup_hunk_meta(hunk._hunk_meta) for old, new, changed in hunk.mdiff(): if changed: if not old[0]: # The '+' char after \x00 is kept # DEBUG: yield 'NEW: %s %s\n' % (old, new) line = new[1].strip('\x00\x01') yield self._markup_new(line) elif not new[0]: # The '-' char after \x00 is kept # DEBUG: yield 'OLD: %s %s\n' % (old, new) line = old[1].strip('\x00\x01') yield self._markup_old(line) else: # DEBUG: yield 'CHG: %s %s\n' % (old, new) yield (self._markup_old('-') + self._markup_mix(old[1], 'red')) yield (self._markup_new('+') + self._markup_mix(new[1], 'green')) else: yield self._markup_common(' ' + old[1]) def _markup_side_by_side(self, diff): """Returns a generator""" def _normalize(line): return (line .replace('\t', ' ' * self._tab_width) .replace('\n', '') .replace('\r', '')) def _fit_with_marker_mix(text, base_color): """Wrap input text which contains mdiff tags, markup at the meantime """ out = [COLORS[base_color]] tag_re = re.compile(r'\x00[+^-]|\x01') while text: if text.startswith('\x00-'): # del out.append(COLORS['reverse'] + COLORS[base_color]) text = text[2:] elif text.startswith('\x00+'): # add out.append(COLORS['reverse'] + COLORS[base_color]) text = text[2:] elif text.startswith('\x00^'): # change out.append(COLORS['underline'] + COLORS[base_color]) text = text[2:] elif text.startswith('\x01'): # reset if len(text) > 1: out.append(COLORS['reset'] + COLORS[base_color]) text = text[1:] else: # FIXME: utf-8 wchar might break the rule here, e.g. # u'\u554a' takes double width of a single letter, also # this depends on your terminal font. I guess audience of # this tool never put that kind of symbol in their code :-) # out.append(text[0]) text = text[1:] out.append(COLORS['reset']) return ''.join(out) # Set up number width, note last hunk might be empty try: (start, offset) = diff._hunks[-1]._old_addr max1 = start + offset - 1 (start, offset) = diff._hunks[-1]._new_addr max2 = start + offset - 1 except IndexError: max1 = max2 = 0 num_width = max(len(str(max1)), len(str(max2))) # Set up line width width = self._width if width <= 0: # Autodetection of text width according to terminal size try: # Each line is like 'nnn TEXT nnn TEXT\n', so width is half of # [terminal size minus the line number columns and 3 separating # spaces # width = (terminal_size()[0] - num_width * 2 - 3) // 2 except Exception: # If terminal detection failed, set back to default width = 80 # Setup lineno and line format left_num_fmt = colorize('%%(left_num)%ds' % num_width, 'yellow') right_num_fmt = colorize('%%(right_num)%ds' % num_width, 'yellow') line_fmt = (left_num_fmt + ' %(left)s ' + COLORS['reset'] + right_num_fmt + ' %(right)s\n') # yield header, old path and new path for line in diff._headers: yield self._markup_header(line) yield self._markup_old_path(diff._old_path) yield self._markup_new_path(diff._new_path) # yield hunks for hunk in diff._hunks: for hunk_header in hunk._hunk_headers: yield self._markup_hunk_header(hunk_header) yield self._markup_hunk_meta(hunk._hunk_meta) for old, new, changed in hunk.mdiff(): if old[0]: left_num = str(hunk._old_addr[0] + int(old[0]) - 1) else: left_num = ' ' if new[0]: right_num = str(hunk._new_addr[0] + int(new[0]) - 1) else: right_num = ' ' left = _normalize(old[1]) right = _normalize(new[1]) if changed: if not old[0]: left = '' right = right.rstrip('\x01') if right.startswith('\x00+'): right = right[2:] right = self._markup_new(right) elif not new[0]: left = left.rstrip('\x01') if left.startswith('\x00-'): left = left[2:] left = self._markup_old(left) right = '' else: left = _fit_with_marker_mix(left, 'red') right = _fit_with_marker_mix(right, 'green') else: left = self._markup_common(left) right = self._markup_common(right) if self._wrap: # Need to wrap long lines, so here we'll iterate, # shaving off `width` chars from both left and right # strings, until both are empty. Also, line number needs to # be printed only for the first part. lncur = left_num rncur = right_num while left or right: # Split both left and right lines, preserving escaping # sequences correctly. lcur, left, llen = strsplit(left, width) rcur, right, rlen = strsplit(right, width) # Pad left line with spaces if needed if llen < width: lcur = '%s%*s' % (lcur, width - llen, '') yield line_fmt % { 'left_num': lncur, 'left': lcur, 'right_num': rncur, 'right': rcur } # Clean line numbers for further iterations lncur = '' rncur = '' else: # Don't need to wrap long lines; instead, a trailing '>' # char needs to be appended. wrap_char = colorize('>', 'lightmagenta') left = strtrim(left, width, wrap_char, len(right) > 0) right = strtrim(right, width, wrap_char, False) yield line_fmt % { 'left_num': left_num, 'left': left, 'right_num': right_num, 'right': right } def _markup_header(self, line): return colorize(line, 'cyan') def _markup_old_path(self, line): return colorize(line, 'yellow') def _markup_new_path(self, line): return colorize(line, 'yellow') def _markup_hunk_header(self, line): return colorize(line, 'lightcyan') def _markup_hunk_meta(self, line): return colorize(line, 'lightblue') def _markup_common(self, line): return colorize(line, 'reset') def _markup_old(self, line): return colorize(line, 'lightred') def _markup_new(self, line): return colorize(line, 'green') def _markup_mix(self, line, base_color): del_code = COLORS['reverse'] + COLORS[base_color] add_code = COLORS['reverse'] + COLORS[base_color] chg_code = COLORS['underline'] + COLORS[base_color] rst_code = COLORS['reset'] + COLORS[base_color] line = line.replace('\x00-', del_code) line = line.replace('\x00+', add_code) line = line.replace('\x00^', chg_code) line = line.replace('\x01', rst_code) return colorize(line, base_color) def markup_to_pager(stream, opts): """Pipe unified diff stream to pager (less). Note: have to create pager Popen object before the translator Popen object in PatchStreamForwarder, otherwise the `stdin=subprocess.PIPE` would cause trouble to the translator pipe (select() never see EOF after input stream ended), most likely python bug 12607 (http://bugs.python.org/issue12607) which was fixed in python 2.7.3. See issue #30 (https://github.com/ymattw/ydiff/issues/30) for more information. """ pager_cmd = ['less'] if not os.getenv('LESS'): # Args stolen from git source: github.com/git/git/blob/master/pager.c pager_cmd.extend(['-FRSX', '--shift 1']) pager = subprocess.Popen( pager_cmd, stdin=subprocess.PIPE, stdout=sys.stdout) diffs = DiffParser(stream).get_diff_generator() for diff in diffs: marker = DiffMarker(side_by_side=opts.side_by_side, width=opts.width, tab_width=opts.tab_width, wrap=opts.wrap) color_diff = marker.markup(diff) for line in color_diff: pager.stdin.write(line.encode('utf-8')) pager.stdin.close() pager.wait() def check_command_status(arguments): """Return True if command returns 0.""" try: return subprocess.call( arguments, stdout=subprocess.PIPE, stderr=subprocess.PIPE) == 0 except OSError: return False def revision_control_diff(args): """Return diff from revision control system.""" for _, ops in VCS_INFO.items(): if check_command_status(ops['probe']): return subprocess.Popen( ops['diff'] + args, stdout=subprocess.PIPE).stdout def revision_control_log(args): """Return log from revision control system.""" for _, ops in VCS_INFO.items(): if check_command_status(ops['probe']): return subprocess.Popen( ops['log'] + args, stdout=subprocess.PIPE).stdout def decode(line): """Decode UTF-8 if necessary.""" if isinstance(line, unicode): return line for encoding in ['utf-8', 'latin1']: try: return line.decode(encoding) except UnicodeDecodeError: pass return '*** ydiff: undecodable bytes ***\n' def terminal_size(): """Returns terminal size. Taken from https://gist.github.com/marsam/7268750 but removed win32 support which depends on 3rd party extension. """ width, height = None, None try: import struct import fcntl import termios s = struct.pack('HHHH', 0, 0, 0, 0) x = fcntl.ioctl(1, termios.TIOCGWINSZ, s) height, width = struct.unpack('HHHH', x)[0:2] except (IOError, AttributeError): pass return width, height def main(): signal.signal(signal.SIGPIPE, signal.SIG_DFL) signal.signal(signal.SIGINT, signal.SIG_DFL) from optparse import (OptionParser, BadOptionError, AmbiguousOptionError, OptionGroup) class PassThroughOptionParser(OptionParser): """Stop parsing on first unknown option (e.g. --cached, -U10) and pass them down. Note the `opt_str` in exception object does not give us chance to take the full option back, e.g. for '-U10' it will only contain '-U' and the '10' part will be lost. Ref: http://goo.gl/IqY4A (on stackoverflow). My hack is to try parse and insert a '--' in place and parse again. Let me know if someone has better solution. """ def _process_args(self, largs, rargs, values): left = largs[:] right = rargs[:] try: OptionParser._process_args(self, left, right, values) except (BadOptionError, AmbiguousOptionError): parsed_num = len(rargs) - len(right) - 1 rargs.insert(parsed_num, '--') OptionParser._process_args(self, largs, rargs, values) supported_vcs = sorted(VCS_INFO.keys()) usage = """%prog [options] [file|dir ...]""" parser = PassThroughOptionParser( usage=usage, description=META_INFO['description'], version='%%prog %s' % META_INFO['version']) parser.add_option( '-s', '--side-by-side', action='store_true', help='enable side-by-side mode') parser.add_option( '-w', '--width', type='int', default=80, metavar='N', help='set text width for side-by-side mode, 0 for auto detection, ' 'default is 80') parser.add_option( '-l', '--log', action='store_true', help='show log with changes from revision control') parser.add_option( '-c', '--color', default='auto', metavar='M', help="""colorize mode 'auto' (default), 'always', or 'never'""") parser.add_option( '-t', '--tab-width', type='int', default=8, metavar='N', help="""convert tab characters to this many spcaes (default: 8)""") parser.add_option( '', '--wrap', action='store_true', help='wrap long lines in side-by-side view') # Hack: use OptionGroup text for extra help message after option list option_group = OptionGroup( parser, 'Note', ('Option parser will stop on first unknown option ' 'and pass them down to underneath revision control. ' 'Environment variable YDIFF_OPTIONS may be used to ' 'specify default options that will be placed at the ' 'beginning of the argument list.')) parser.add_option_group(option_group) # Place possible options defined in YDIFF_OPTIONS at the beginning of argv ydiff_opts = [x for x in os.getenv('YDIFF_OPTIONS', '').split(' ') if x] # TODO: Deprecate CDIFF_OPTIONS. Fall back to it and warn (for now). if not ydiff_opts: cdiff_opts = [x for x in os.getenv('CDIFF_OPTIONS', '').split(' ') if x] if cdiff_opts: sys.stderr.write('*** CDIFF_OPTIONS will be depreated soon, ' 'please use YDIFF_OPTIONS instead\n') ydiff_opts = cdiff_opts opts, args = parser.parse_args(ydiff_opts + sys.argv[1:]) if opts.log: diff_hdl = revision_control_log(args) if not diff_hdl: sys.stderr.write(('*** Not in a supported workspace, supported ' 'are: %s\n') % ', '.join(supported_vcs)) return 1 elif sys.stdin.isatty(): diff_hdl = revision_control_diff(args) if not diff_hdl: sys.stderr.write(('*** Not in a supported workspace, supported ' 'are: %s\n\n') % ', '.join(supported_vcs)) parser.print_help() return 1 else: diff_hdl = (sys.stdin.buffer if hasattr(sys.stdin, 'buffer') else sys.stdin) stream = PatchStream(diff_hdl) # Don't let empty diff pass thru if stream.is_empty(): return 0 if (opts.color == 'always' or (opts.color == 'auto' and sys.stdout.isatty())): markup_to_pager(stream, opts) else: # pipe out stream untouched to make sure it is still a patch byte_output = (sys.stdout.buffer if hasattr(sys.stdout, 'buffer') else sys.stdout) for line in stream: byte_output.write(line) if diff_hdl is not sys.stdin: diff_hdl.close() return 0 if __name__ == '__main__': sys.exit(main()) # vim:set et sts=4 sw=4 tw=79: