aeidon/liner.py¶
Source code for module aeidon.liner from file aeidon/liner.py.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 | # -*- coding: utf-8 -*-
# Copyright (C) 2011 Osmo Salomaa
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
"""Breaking lines to a specified width."""
import aeidon
import re
import sys
__all__ = ("Liner",)
class Liner(aeidon.Parser):
"""
Breaking lines to a specified width.
:ivar length_func: A function that returns the length of its argument
:ivar max_length: Maximum length of a line in units of :attr:`length_func`
:ivar max_lines: Maximum preferred amount of lines (may be exceeded)
:ivar _penalties: List of penalty pattern dictionaries
"""
# Reading Donald E. Knuth and Michael F. Plass's "Breaking Paragraphs into
# Lines" from "Software--Practice and Experience" vol. 11 from 1981 is
# recommended to understand the general problem of breaking a paragraph
# of text into lines, the terminology used here as well of boxes, penalties
# and demerits and the computational complexity of finding the optimal
# solution in each case. Subtitling is in many ways a different
# application, which is a lot simpler, but also more ambiguous and
# subjective in terms of the minimized demerit measure.
_re_multi_space = re.compile(r" {2,}")
def __init__(self, re_tag=None, clean_func=None):
"""Initialize a :class:`Liner` instance."""
aeidon.Parser.__init__(self, re_tag, clean_func)
self._penalties = []
self.length_func = len
self.max_length = 40
self.max_lines = 3
def _boxes_to_lines(self, boxes, breaks):
"""Return `boxes` joined to form lines."""
edges = [0] + [x+1 for x in breaks] + [len(boxes)]
return [" ".join(boxes[edges[i]:edges[i+1]])
for i in range(len(edges) - 1)]
def _break_lines(self, boxes, penalties, nlines):
"""
Break `boxes` into lines and return break points and demerit.
If keeping all boxes on a single line results in a valid and better
result than splitting to `nlines` return an empty list. If no
valid break points can be found, return ``None``.
"""
breaks = self._list_possible_breaks(boxes, penalties, nlines)
best_breaks = None
best_demerit = sys.maxsize
text = " ".join(boxes)
if self.length_func(text) <= self.max_length:
# Use a valid one-line solution as the benchmark,
# that any more-line solution must beat.
best_breaks = []
best_demerit = self._calculate_demerit(boxes, penalties, [])
if nlines == 1:
return best_breaks, best_demerit
if nlines == 2:
for i in breaks:
if penalties[i] > best_demerit: break
demerit = self._calculate_demerit(boxes, penalties, [i])
if demerit < best_demerit:
best_breaks = [i]
best_demerit = demerit
return best_breaks, best_demerit
# For more than two lines, loop over first break points and
# recursively figure out rest of the breaks in each case.
for i in breaks:
# Use the maximum total negative penalty
# that can accumulate from all later breaks.
negpen = sorted(x for x in penalties[i+1:] if x < 0)
negpen = sum(negpen[:min(len(negpen), nlines - 2)])
if (penalties[i] + negpen) > best_demerit: break
value = self._break_lines(boxes[i+1:], penalties[i+1:], nlines-1)
if value[0] is None: continue
later = [i+1+x for x in value[0]]
demerit = self._calculate_demerit(boxes, penalties, [i] + later)
if demerit < best_demerit:
best_breaks = [i] + later
best_demerit = demerit
return best_breaks, best_demerit
def break_lines(self):
"""Break lines and return text."""
self.text = self.text.replace("\n", " ")
self.pattern = self._re_multi_space
self.replacement = " "
self.replace_all()
boxes = self.text.split(" ")
if len(boxes) == 1:
return self.get_text()
penalties = self._detect_penalties(boxes)
best_breaks = None
best_demerit = sys.maxsize
# We can probably handle up to ten lines of text
# before finding break points gets intolerably slow.
min_nlines = min(2, self.max_lines)
max_nlines = min(10, len(boxes))
for nlines in range(min_nlines, max_nlines+1):
breaks, demerit = self._break_lines(boxes, penalties, nlines)
if breaks is None: continue
if demerit < best_demerit:
best_breaks = breaks
best_demerit = demerit
if nlines < self.max_lines:
continue
pos = -1
for i in range(len(boxes)):
pos = pos + 1 + len(boxes[i])
text = self.text
if i in best_breaks:
text = text[:pos] + "\n" + text[pos+1:]
self.text = text
return self.get_text()
# If text cannot be broken, return original text.
return self.get_text()
def _calculate_demerit(self, boxes, penalties, breaks):
"""Return demerit measure for `boxes` broken by `breaks`."""
nlines = len(breaks) + 1
penalties = [penalties[i] for i in breaks]
lines = self._boxes_to_lines(boxes, breaks)
lengths = list(map(self.length_func, lines))
mlength = sum(lengths) / len(lengths)
xlength = self.max_length
# Use two subjective measures of badness: (1) 'deviation',
# which is the variance of line lengths relative to the
# maximum line length and (2) upside-down 'pyramid', which
# is the sum of how much longer each line is than the next.
return (sum(penalties)
+ 50 * sum(((x - mlength) / xlength)**2 for x in lengths)
+ 50 * sum(((lengths[i] - lengths[i+1]) / xlength)**2
for i in range(len(lengths) - 1)
if lengths[i] > lengths[i+1])
+ 100 * (nlines-1)**3
+ 1000 * max(0, nlines - self.max_lines)**3)
def _detect_penalties(self, boxes):
"""Detect penalties for break points following `boxes`."""
text = " ".join(self._boxes_to_lines(boxes, breaks=[]))
textpen = [0] * len(text)
for penalty in self._penalties:
self.pattern = penalty["regex"]
self.pos = 0
while True:
try:
self.next()
except StopIteration:
break
start, end = self.match.span(penalty["group"])
# Use sum, since in some rare cases multiple
# patterns can match the same space.
textpen[start] += penalty["value"]
penalties = [0] * len(boxes)
pos = -1
for i in range(len(boxes) - 1):
pos = pos + 1 + len(boxes[i])
penalties[i] = textpen[pos]
return penalties
@aeidon.deco.memoize(100)
def _list_possible_breaks(self, boxes, penalties, nlines):
"""
Return a list of all possible break points for `boxes`.
All break points that would necessarily cause `max_length` to be
violated are discarded. Breaks are returned sorted in ascending order
of associated `penalties`, so that all remaining breaks can be
discarded once a demerit threshold is crossed.
"""
breaks = list(range(len(boxes) - (nlines - 1)))
breakpen = penalties[:len(breaks)]
if nlines == 1:
return []
if nlines == 2:
keep = [False] * len(breaks)
for i in range(len(breaks)):
lines = self._boxes_to_lines(boxes, breaks=[i])
lengths = map(self.length_func, lines)
keep[i] = max(lengths) <= self.max_length
breaks = [breaks[i] for i in range(len(breaks)) if keep[i]]
breakpen = [breakpen[i] for i in range(len(breakpen)) if keep[i]]
# Sort breaks in ascending order by penalties,
# so that all remaining breaks can be discarded once
# a demerit threshold is crossed.
if not breaks: return []
points = sorted(zip(breakpen, breaks))
breakpen, breaks = zip(*points)
return(breaks)
# For more than two lines, loop over first break points and
# recursively figure out rest of the breaks in each case.
keep = [False] * len(breaks)
for i in range(len(breaks)):
lines = self._boxes_to_lines(boxes, breaks=[i])
alength = self.length_func(lines[0])
if alength > self.max_length: break
later = self._list_possible_breaks(boxes[i+1:],
penalties[i+1:],
nlines-1)
keep[i] = bool(later)
breaks = [breaks[i] for i in range(len(breaks)) if keep[i]]
breakpen = [breakpen[i] for i in range(len(breakpen)) if keep[i]]
# Sort breaks in ascending order by penalties,
# so that all remaining breaks can be discarded once
# a demerit threshold is crossed.
if not breaks: return []
points = sorted(zip(breakpen, breaks))
breakpen, breaks = zip(*points)
return(breaks)
def set_penalties(self, penalties):
"""
Set penalty patterns.
`penalties` should be a list of dictionaries with items "pattern",
"flags", "group" and "value", where pattern is a regular expression
with group parentheses around a space, group is the number of the group
in pattern to hold the penalty of value. A negative penalty encourages
a break and a positive penalty discourages.
"""
self._penalties = []
for penalty in penalties:
regex = re.compile(penalty["pattern"], penalty["flags"])
self._penalties.append(dict(regex=regex,
group=penalty["group"],
value=penalty["value"]))
def set_text(self, text):
"""Set the target text to search in and parse it."""
aeidon.Parser.set_text(self, text.strip())
self.text = self.text.strip()
|