From 0d6ec6c5d3ddc42b445fae2c703e91b5870ec624 Mon Sep 17 00:00:00 2001 From: Florent Le Coz Date: Fri, 18 Feb 2011 00:47:44 +0100 Subject: =?UTF-8?q?Properly=20take=20into=20account=20the=20length=20of=20?= =?UTF-8?q?the=20characters,=20fixing=20some=20display=20issues=20(for=20e?= =?UTF-8?q?xample=20the=20cursor=20position=20when=20entering=20stranges?= =?UTF-8?q?=20characters=20likes=20=E7=99=94)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/wcwidth.py | 316 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++ src/windows.py | 47 +++++---- 2 files changed, 342 insertions(+), 21 deletions(-) create mode 100644 src/wcwidth.py diff --git a/src/wcwidth.py b/src/wcwidth.py new file mode 100644 index 00000000..b72c3ca3 --- /dev/null +++ b/src/wcwidth.py @@ -0,0 +1,316 @@ +#/usr/bin/env python +# -*- encoding: utf-8 +""" + * This is a Python implementation of wcwidth() and wcswidth(), based on the + * C implementation of the same functions (defined in IEEE Std 1002.1-2001) + * for Unicode: + * + * http://www.opengroup.org/onlinepubs/007904975/functions/wcwidth.html + * http://www.opengroup.org/onlinepubs/007904975/functions/wcswidth.html + * + * In fixed-width output devices, Latin characters all occupy a single + * "cell" position of equal width, whereas ideographic CJK characters + * occupy two such cells. Interoperability between terminal-line + * applications and (teletype-style) character terminals using the + * UTF-8 encoding requires agreement on which character should advance + * the cursor by how many cell positions. No established formal + * standards exist at present on which Unicode character shall occupy + * how many cell positions on character terminals. These routines are + * a first attempt of defining such behavior based on simple rules + * applied to data provided by the Unicode Consortium. + * + * For some graphical characters, the Unicode standard explicitly + * defines a character-cell width via the definition of the East Asian + * FullWidth (F), Wide (W), Half-width (H), and Narrow (Na) classes. + * In all these cases, there is no ambiguity about which width a + * terminal shall use. For characters in the East Asian Ambiguous (A) + * class, the width choice depends purely on a preference of backward + * compatibility with either historic CJK or Western practice. + * Choosing single-width for these characters is easy to justify as + * the appropriate long-term solution, as the CJK practice of + * displaying these characters as double-width comes from historic + * implementation simplicity (8-bit encoded characters were displayed + * single-width and 16-bit ones double-width, even for Greek, + * Cyrillic, etc.) and not any typographic considerations. + * + * Much less clear is the choice of width for the Not East Asian + * (Neutral) class. Existing practice does not dictate a width for any + * of these characters. It would nevertheless make sense + * typographically to allocate two character cells to characters such + * as for instance EM SPACE or VOLUME INTEGRAL, which cannot be + * represented adequately with a single-width glyph. The following + * routines at present merely assign a single-cell width to all + * neutral characters, in the interest of simplicity. This is not + * entirely satisfactory and should be reconsidered before + * establishing a formal standard in this area. At the moment, the + * decision which Not East Asian (Neutral) characters should be + * represented by double-width glyphs cannot yet be answered by + * applying a simple rule from the Unicode database content. Setting + * up a proper standard for the behavior of UTF-8 character terminals + * will require a careful analysis not only of each Unicode character, + * but also of each presentation form, something the author of these + * routines has avoided to do so far. + * + * http://www.unicode.org/unicode/reports/tr11/ + * + * Markus Kuhn -- 2007-05-26 (Unicode 5.0) + * Berteun Damman - 2007-06-28 (Python version) + * + * Permission to use, copy, modify, and distribute this software + * for any purpose and without fee is hereby granted. The author + * disclaims all warranties with regard to this software. + * + * Latest C version: http://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c +""" + +# auxiliary function for binary search in interval table, see below +def bisearch(ucs): + mn = 0 + mx = len(combining_table) - 1 + if ucs < combining_table[0][0] or ucs > combining_table[mx][1]: + return False + + while mx >= mn: + mid = (mn + mx) // 2 + if ucs > combining_table[mid][1]: + mn = mid + 1 + elif ucs < combining_table[mid][0]: + mx = mid - 1 + else: + return True + + return False + + +""" + * The following two functions define the column width of an ISO 10646 + * character as follows: + * + * - The null character (U+0000) has a column width of 0. + * + * - Other C0/C1 control characters and DEL will lead to a return + * value of -1. + * + * - Non-spacing and enclosing combining characters (general + * category code Mn or Me in the Unicode database) have a + * column width of 0. + * + * - SOFT HYPHEN (U+00AD) has a column width of 1. + * + * - Other format characters (general category code Cf in the Unicode + * database) and ZERO WIDTH SPACE (U+200B) have a column width of 0. + * + * - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF) + * have a column width of 0. + * + * - Spacing characters in the East Asian Wide (W) or East Asian + * Full-width (F) category as defined in Unicode Technical + * Report #11 have a column width of 2. + * + * - All remaining characters (including all printable + * ISO 8859-1 and WGL4 characters, Unicode control characters, + * etc.) have a column width of 1. + * + * This implementation assumes that wchar_t characters are encoded + * in ISO 10646. +""" +# sorted list of non-overlapping intervals of non-spacing characters +# generated by "uniset +cat=Me +cat=Mn +cat=Cf -00AD +1160-11FF +200B c" +combining_table = [ + ('\u0300', '\u036F'), ('\u0483', '\u0486'), ('\u0488', '\u0489'), + ('\u0591', '\u05BD'), ('\u05BF', '\u05BF'), ('\u05C1', '\u05C2'), + ('\u05C4', '\u05C5'), ('\u05C7', '\u05C7'), ('\u0600', '\u0603'), + ('\u0610', '\u0615'), ('\u064B', '\u065E'), ('\u0670', '\u0670'), + ('\u06D6', '\u06E4'), ('\u06E7', '\u06E8'), ('\u06EA', '\u06ED'), + ('\u070F', '\u070F'), ('\u0711', '\u0711'), ('\u0730', '\u074A'), + ('\u07A6', '\u07B0'), ('\u07EB', '\u07F3'), ('\u0901', '\u0902'), + ('\u093C', '\u093C'), ('\u0941', '\u0948'), ('\u094D', '\u094D'), + ('\u0951', '\u0954'), ('\u0962', '\u0963'), ('\u0981', '\u0981'), + ('\u09BC', '\u09BC'), ('\u09C1', '\u09C4'), ('\u09CD', '\u09CD'), + ('\u09E2', '\u09E3'), ('\u0A01', '\u0A02'), ('\u0A3C', '\u0A3C'), + ('\u0A41', '\u0A42'), ('\u0A47', '\u0A48'), ('\u0A4B', '\u0A4D'), + ('\u0A70', '\u0A71'), ('\u0A81', '\u0A82'), ('\u0ABC', '\u0ABC'), + ('\u0AC1', '\u0AC5'), ('\u0AC7', '\u0AC8'), ('\u0ACD', '\u0ACD'), + ('\u0AE2', '\u0AE3'), ('\u0B01', '\u0B01'), ('\u0B3C', '\u0B3C'), + ('\u0B3F', '\u0B3F'), ('\u0B41', '\u0B43'), ('\u0B4D', '\u0B4D'), + ('\u0B56', '\u0B56'), ('\u0B82', '\u0B82'), ('\u0BC0', '\u0BC0'), + ('\u0BCD', '\u0BCD'), ('\u0C3E', '\u0C40'), ('\u0C46', '\u0C48'), + ('\u0C4A', '\u0C4D'), ('\u0C55', '\u0C56'), ('\u0CBC', '\u0CBC'), + ('\u0CBF', '\u0CBF'), ('\u0CC6', '\u0CC6'), ('\u0CCC', '\u0CCD'), + ('\u0CE2', '\u0CE3'), ('\u0D41', '\u0D43'), ('\u0D4D', '\u0D4D'), + ('\u0DCA', '\u0DCA'), ('\u0DD2', '\u0DD4'), ('\u0DD6', '\u0DD6'), + ('\u0E31', '\u0E31'), ('\u0E34', '\u0E3A'), ('\u0E47', '\u0E4E'), + ('\u0EB1', '\u0EB1'), ('\u0EB4', '\u0EB9'), ('\u0EBB', '\u0EBC'), + ('\u0EC8', '\u0ECD'), ('\u0F18', '\u0F19'), ('\u0F35', '\u0F35'), + ('\u0F37', '\u0F37'), ('\u0F39', '\u0F39'), ('\u0F71', '\u0F7E'), + ('\u0F80', '\u0F84'), ('\u0F86', '\u0F87'), ('\u0F90', '\u0F97'), + ('\u0F99', '\u0FBC'), ('\u0FC6', '\u0FC6'), ('\u102D', '\u1030'), + ('\u1032', '\u1032'), ('\u1036', '\u1037'), ('\u1039', '\u1039'), + ('\u1058', '\u1059'), ('\u1160', '\u11FF'), ('\u135F', '\u135F'), + ('\u1712', '\u1714'), ('\u1732', '\u1734'), ('\u1752', '\u1753'), + ('\u1772', '\u1773'), ('\u17B4', '\u17B5'), ('\u17B7', '\u17BD'), + ('\u17C6', '\u17C6'), ('\u17C9', '\u17D3'), ('\u17DD', '\u17DD'), + ('\u180B', '\u180D'), ('\u18A9', '\u18A9'), ('\u1920', '\u1922'), + ('\u1927', '\u1928'), ('\u1932', '\u1932'), ('\u1939', '\u193B'), + ('\u1A17', '\u1A18'), ('\u1B00', '\u1B03'), ('\u1B34', '\u1B34'), + ('\u1B36', '\u1B3A'), ('\u1B3C', '\u1B3C'), ('\u1B42', '\u1B42'), + ('\u1B6B', '\u1B73'), ('\u1DC0', '\u1DCA'), ('\u1DFE', '\u1DFF'), + ('\u200B', '\u200F'), ('\u202A', '\u202E'), ('\u2060', '\u2063'), + ('\u206A', '\u206F'), ('\u20D0', '\u20EF'), ('\u302A', '\u302F'), + ('\u3099', '\u309A'), ('\uA806', '\uA806'), ('\uA80B', '\uA80B'), + ('\uA825', '\uA826'), ('\uFB1E', '\uFB1E'), ('\uFE00', '\uFE0F'), + ('\uFE20', '\uFE23'), ('\uFEFF', '\uFEFF'), ('\uFFF9', '\uFFFB'), +] + + # XXX: There are some issues with Plane 1 Unicode characters on 32-bit + # systems. As these use UTF-16 internally they will use surrogate pairs + # to represent the character. I don't know how this works exactly though, + # therefore, until I've figured it out, if we're on a 32-bit system, + # we won't include these, otherwise we will. +if '\U0000FFFF' < '\U00010000': + combining_table.extend([ + ('\U00010A01', '\U00010A03'), ('\U00010A05', '\U00010A06'), + ('\U00010A0C', '\U00010A0F'), ('\U00010A38', '\U00010A3A'), + ('\U00010A3F', '\U00010A3F'), ('\U0001D167', '\U0001D169'), + ('\U0001D173', '\U0001D182'), ('\U0001D185', '\U0001D18B'), + ('\U0001D1AA', '\U0001D1AD'), ('\U0001D242', '\U0001D244'), + ('\U000E0001', '\U000E0001'), ('\U000E0020', '\U000E007F'), + ('\U000E0100', '\U000E01EF'), + ]) + +def wcwidth(ucs): + if len(ucs) > 1: + raise TypeError('wcwidth() expected a character, ' + 'but string of length %d found' % (len(ucs),)) + # test for 8-bit control characters + if ucs == '\u0000': + return 0 + + # non-printable chars. + if ucs < '\u0020' or (ucs >= '\u007f' and ucs < '\u00a0'): + return -1 + + # binary search in table of non-spacing characters + if bisearch(ucs): + return 0 + + # if we arrive here, ucs is not a combining or C0/C1 control character + + return (1 + + (ucs >= '\u1100' and + (ucs <= '\u115f' or # Hangul Jamo init. consonants + ucs == '\u2329' or ucs == '\u232a' or + (ucs >= '\u2e80' and ucs <= '\ua4cf' and + ucs != '\u303f') or # CJK ... Yi + (ucs >= '\uac00' and ucs <= '\ud7a3') or # Hangul Syllables + (ucs >= '\uf900' and ucs <= '\ufaff') or # CJK Comp. Ideographs + (ucs >= '\ufe10' and ucs <= '\ufe19') or # Vertical forms + (ucs >= '\ufe30' and ucs <= '\ufe6f') or # CJK Comp. Forms + (ucs >= '\uff00' and ucs <= '\uff60') or # Fullwidth Forms + (ucs >= '\uffe0' and ucs <= '\uffe6') or + # XXX: '\U0000FFFF' < '\U00010000' is only True on 64-bit systems. + # On 32 bit systems it fails, but hopefully it won't cause chars to be + # misrepresented. It has to do with surrogate pairs, but I don't know + # how to fix this. + (('\U0000FFFF' < '\U00010000') and + (ucs >= '\U00020000' and ucs <= '\U0002fffd') or + (ucs >= '\U00030000' and ucs <= '\U0003fffd'))))) + + +def wcswidth(s): + """ + Return the length of the passed string, using wcwidth on each char + instead of couting 1 for each one. + """ + width = 0 + for c in s: + w = wcwidth(c) + if w < 0: + # If s contains a non-printable char, we should return -1. + # This includes newlines and tabs! + return -1 + else: + width += w + return width + +def widthcut(s, m): + """ + Return the first characters of s that can be contained in + a m length + """ + i = 0 + width = 0 + for c in s: + w = wcwidth(c) + if w < 0: + return None + else: + width += w + + i += 1 + if width > m: + return s[:i-1] + return s + +def ljust(s, max, fillchar=" "): + """ + Like widthcut but adding chars at the end of the string until + max is reached + """ + if wcwidth(fillchar)!=1: + raise TypeError('widthpad() expected fillchar as a character, ' + 'but string of length %d found' % (len(fillchar),)) + i = 0 + width = 0 + for c in s: + w = wcwidth(c) + if w < 0: + return None + else: + width += w + + i += 1 + if width==max: + return s[:i] + if width > max: + return s[:i-1]+fillchar + + return s + fillchar*(max-width) + +def rjust(s, max, fillchar=" "): + if wcwidth(fillchar)!=1: + raise TypeError('widthpad() expected fillchar as a character, ' + 'but string of length %d found' % (len(fillchar),)) + i = 0 + width = 0 + for c in s: + w = wcwidth(c) + if w < 0: + return None + else: + width += w + + i += 1 + if width==max: + return s[:i] + if width > max: + return fillchar+s[:i-1] + + return fillchar*(max-width) + s + +if __name__ == '__main__': + import unicodedata + test_strings = [ + 'Pál Erdős', 'Kurt Gödel', 'Évariste Galois', + "Guillaume de l'Hôpital", + 'ἄνδρα μοι ἔννεπε, μοῦσα, πολύτροπον, ὃς μάλα πολλὰ πλάγχθη', + ] + for s in test_strings: + # d will be the decomposed version, this one should have the + # same display width, but it should have more characters. + d = unicodedata.normalize('NFD', s) + assert wcswidth(s) == wcswidth(d) + assert len(s) != len(d) + assert wcswidth('string with \n char') == -1 + assert wcswidth('string with \t char') == -1 + print('Minor testcase succeeded') diff --git a/src/windows.py b/src/windows.py index fa17d8ff..aab254c0 100644 --- a/src/windows.py +++ b/src/windows.py @@ -48,6 +48,7 @@ from sleekxmpp.xmlstream.stanzabase import JID import theme import common +import wcwidth g_lock = Lock() @@ -391,10 +392,8 @@ class MucInfoWin(InfoWin): self._refresh() def write_room_name(self, room): - """ - """ self.addstr('[', common.curses_color_pair(theme.COLOR_INFORMATION_BAR)) - self.addnstr(room.name, len(room.name), common.curses_color_pair(theme.COLOR_GROUPCHAT_NAME)) + self.addstr(room.name, common.curses_color_pair(theme.COLOR_GROUPCHAT_NAME)) self.addstr('] ', common.curses_color_pair(theme.COLOR_INFORMATION_BAR)) def write_disconnected(self, room): @@ -491,23 +490,27 @@ class TextWin(Win): txt = message.txt if not txt: return 0 - # length of the time + else: + txt = txt.replace('\t', ' ') + # length of the time offset = 9+len(theme.CHAR_TIME_LEFT[:1])+len(theme.CHAR_TIME_RIGHT[:1]) - if message.nickname and len(message.nickname) >= 30: - nick = message.nickname[:30]+'…' + if message.nickname and wcwidth.wcswidth(message.nickname) >= 25: + nick = message.nickname[:25]+'…' else: nick = message.nickname if nick: - offset += len(nick) + 2 # + nick + spaces length + offset += wcwidth.wcswidth(nick) + 2 # + nick + spaces length first = True this_line_was_broken_by_space = False nb = 0 while txt != '': - limit = txt[:self.width-offset].find('\n') + cutted_txt = wcwidth.widthcut(txt, self.width-offset) or txt[:self.width-offset] + limit = cutted_txt.find('\n') if limit < 0: # break between words if possible - if len(txt) >= self.width-offset: - limit = txt[:self.width-offset].rfind(' ') + if wcwidth.wcswidth(txt) >= self.width-offset: + cutted_txt = wcwidth.widthcut(txt, self.width-offset) or txt[:self.width-offset] + limit = cutted_txt.rfind(' ') this_line_was_broken_by_space = True if limit <= 0: limit = self.width-offset @@ -528,15 +531,18 @@ class TextWin(Win): } l = Line(nick, color, time, - txt[:limit], message.color, + wcwidth.widthcut(txt, limit) or txt[:limit], message.color, offset, message.colorized) self.built_lines.append(l) nb += 1 if this_line_was_broken_by_space: - txt = txt[limit+1:] # jump the space at the start of the line - else: + limit += 1 # jump the space at the start of the line + cutted_txt = wcwidth.widthcut(txt, limit) + if not cutted_txt: txt = txt[limit:] + else: + txt = txt[len(cutted_txt):] if txt.startswith('\n'): txt = txt[1:] first = False @@ -567,13 +573,11 @@ class TextWin(Win): if line.nickname: self.write_nickname(line.nickname, line.nickname_color) self.write_text(y, line.text_offset, line.text, line.text_color, line.colorized) - if y != self.height-1 or (not line or line.text_offset+len(line.text) < self.width): + if y != self.height-1 or (not line or line.text_offset+wcwidth.wcswidth(line.text) < self.width): self.addstr('\n') self._refresh() def write_line_separator(self): - """ - """ self.addnstr('- '*(self.width//2-1)+'-', self.width, common.curses_color_pair(theme.COLOR_NEW_TEXT_SEPARATOR)) def write_text(self, y, x, txt, color, colorized): @@ -603,7 +607,7 @@ class TextWin(Win): .replace('"(', '').replace(')"', '') splitted = txt.split() for word in splitted: - if word in list(special_words.keys()): + if word in special_words.keys(): self.addstr(word, common.curses_color_pair(special_words[word])) elif word.startswith('(') and word.endswith(')'): self.addstr('(', common.curses_color_pair(color)) @@ -1000,9 +1004,9 @@ class Input(Win): self.text = self.text[:self.pos+self.line_pos]+key+self.text[self.pos+self.line_pos:] (y, x) = self._win.getyx() if x == self.width-1: - self.line_pos += len(key) + self.line_pos += 1 # wcwidth.wcswidth(key) else: - self.pos += len(key) + self.pos += 1 # wcwidth.wcswidth(key) if reset: self.rewrite_text() if self.on_input: @@ -1030,12 +1034,13 @@ class Input(Win): self._win.erase() if self.color: self._win.attron(common.curses_color_pair(self.color)) - self.addstr(text[self.line_pos:self.line_pos+self.width-1]) + displayed_text = text[self.line_pos:self.line_pos+self.width-1] + self.addstr(displayed_text) if self.color: (y, x) = self._win.getyx() size = self.width-x self.addnstr(' '*size, size, common.curses_color_pair(self.color)) - self.addstr(0, self.pos, '') # WTF, this works but .move() doesn't… + self.addstr(0, wcwidth.wcswidth(displayed_text[:self.pos]), '') if self.color: self._win.attroff(common.curses_color_pair(self.color)) self._refresh() -- cgit v1.2.3