src/etc/unicode.py

   1 #!/usr/bin/env python
   2 #
   3 # Copyright 2011-2013 The Rust Project Developers. See the COPYRIGHT
   4 # file at the top-level directory of this distribution and at
   5 # http://rust-lang.org/COPYRIGHT.
   6 #
   7 # Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
   8 # http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
   9 # <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
  10 # option. This file may not be copied, modified, or distributed
  11 # except according to those terms.
  12
  13 # This script uses the following Unicode tables:
  14 # - DerivedCoreProperties.txt
  15 # - DerivedNormalizationProps.txt
  16 # - EastAsianWidth.txt
  17 # - auxiliary/GraphemeBreakProperty.txt
  18 # - PropList.txt
  19 # - ReadMe.txt
  20 # - Scripts.txt
  21 # - UnicodeData.txt
  22 #
  23 # Since this should not require frequent updates, we just store this
  24 # out-of-line and check the unicode.rs file into git.
  25
  26 import fileinput, re, os, sys, operator
  27
  28 preamble = '''// Copyright 2012-2014 The Rust Project Developers. See the COPYRIGHT
  29 // file at the top-level directory of this distribution and at
  30 // http://rust-lang.org/COPYRIGHT.
  31 //
  32 // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
  33 // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
  34 // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
  35 // option. This file may not be copied, modified, or distributed
  36 // except according to those terms.
  37
  38 // NOTE: The following code was generated by "src/etc/unicode.py", do not edit directly
  39
  40 #![allow(missing_docs, non_upper_case_globals, non_snake_case)]
  41 '''
  42
  43 # Mapping taken from Table 12 from:
  44 # http://www.unicode.org/reports/tr44/#General_Category_Values
  45 expanded_categories = {
  46     'Lu': ['LC', 'L'], 'Ll': ['LC', 'L'], 'Lt': ['LC', 'L'],
  47     'Lm': ['L'], 'Lo': ['L'],
  48     'Mn': ['M'], 'Mc': ['M'], 'Me': ['M'],
  49     'Nd': ['N'], 'Nl': ['N'], 'No': ['No'],
  50     'Pc': ['P'], 'Pd': ['P'], 'Ps': ['P'], 'Pe': ['P'],
  51     'Pi': ['P'], 'Pf': ['P'], 'Po': ['P'],
  52     'Sm': ['S'], 'Sc': ['S'], 'Sk': ['S'], 'So': ['S'],
  53     'Zs': ['Z'], 'Zl': ['Z'], 'Zp': ['Z'],
  54     'Cc': ['C'], 'Cf': ['C'], 'Cs': ['C'], 'Co': ['C'], 'Cn': ['C'],
  55 }
  56
  57 # these are the surrogate codepoints, which are not valid rust characters
  58 surrogate_codepoints = (0xd800, 0xdfff)
  59
  60 def fetch(f):
  61     if not os.path.exists(os.path.basename(f)):
  62         os.system("curl -O http://www.unicode.org/Public/UNIDATA/%s"
  63                   % f)
  64
  65     if not os.path.exists(os.path.basename(f)):
  66         sys.stderr.write("cannot load %s" % f)
  67         exit(1)
  68
  69 def is_surrogate(n):
  70     return surrogate_codepoints[0] <= n <= surrogate_codepoints[1]
  71
  72 def load_unicode_data(f):
  73     fetch(f)
  74     gencats = {}
  75     upperlower = {}
  76     lowerupper = {}
  77     combines = {}
  78     canon_decomp = {}
  79     compat_decomp = {}
  80
  81     udict = {};
  82     range_start = -1;
  83     for line in fileinput.input(f):
  84         data = line.split(';');
  85         if len(data) != 15:
  86             continue
  87         cp = int(data[0], 16);
  88         if is_surrogate(cp):
  89             continue
  90         if range_start >= 0:
  91             for i in xrange(range_start, cp):
  92                 udict[i] = data;
  93             range_start = -1;
  94         if data[1].endswith(", First>"):
  95             range_start = cp;
  96             continue;
  97         udict[cp] = data;
  98
  99     for code in udict:
 100         [code_org, name, gencat, combine, bidi,
 101          decomp, deci, digit, num, mirror,
 102          old, iso, upcase, lowcase, titlecase ] = udict[code];
 103
 104         # generate char to char direct common and simple conversions
 105         # uppercase to lowercase
 106         if gencat == "Lu" and lowcase != "" and code_org != lowcase:
 107             upperlower[code] = int(lowcase, 16)
 108
 109         # lowercase to uppercase
 110         if gencat == "Ll" and upcase != "" and code_org != upcase:
 111             lowerupper[code] = int(upcase, 16)
 112
 113         # store decomposition, if given
 114         if decomp != "":
 115             if decomp.startswith('<'):
 116                 seq = []
 117                 for i in decomp.split()[1:]:
 118                     seq.append(int(i, 16))
 119                 compat_decomp[code] = seq
 120             else:
 121                 seq = []
 122                 for i in decomp.split():
 123                     seq.append(int(i, 16))
 124                 canon_decomp[code] = seq
 125
 126         # place letter in categories as appropriate
 127         for cat in [gencat, "Assigned"] + expanded_categories.get(gencat, []):
 128             if cat not in gencats:
 129                 gencats[cat] = []
 130             gencats[cat].append(code)
 131
 132         # record combining class, if any
 133         if combine != "0":
 134             if combine not in combines:
 135                 combines[combine] = []
 136             combines[combine].append(code)
 137
 138     # generate Not_Assigned from Assigned
 139     gencats["Cn"] = gen_unassigned(gencats["Assigned"])
 140     # Assigned is not a real category
 141     del(gencats["Assigned"])
 142     # Other contains Not_Assigned
 143     gencats["C"].extend(gencats["Cn"])
 144     gencats = group_cats(gencats)
 145     combines = to_combines(group_cats(combines))
 146
 147     return (canon_decomp, compat_decomp, gencats, combines, lowerupper, upperlower)
 148
 149 def group_cats(cats):
 150     cats_out = {}
 151     for cat in cats:
 152         cats_out[cat] = group_cat(cats[cat])
 153     return cats_out
 154
 155 def group_cat(cat):
 156     cat_out = []
 157     letters = sorted(set(cat))
 158     cur_start = letters.pop(0)
 159     cur_end = cur_start
 160     for letter in letters:
 161         assert letter > cur_end, \
 162             "cur_end: %s, letter: %s" % (hex(cur_end), hex(letter))
 163         if letter == cur_end + 1:
 164             cur_end = letter
 165         else:
 166             cat_out.append((cur_start, cur_end))
 167             cur_start = cur_end = letter
 168     cat_out.append((cur_start, cur_end))
 169     return cat_out
 170
 171 def ungroup_cat(cat):
 172     cat_out = []
 173     for (lo, hi) in cat:
 174         while lo <= hi:
 175             cat_out.append(lo)
 176             lo += 1
 177     return cat_out
 178
 179 def gen_unassigned(assigned):
 180     assigned = set(assigned)
 181     return ([i for i in range(0, 0xd800) if i not in assigned] +
 182             [i for i in range(0xe000, 0x110000) if i not in assigned])
 183
 184 def to_combines(combs):
 185     combs_out = []
 186     for comb in combs:
 187         for (lo, hi) in combs[comb]:
 188             combs_out.append((lo, hi, comb))
 189     combs_out.sort(key=lambda comb: comb[0])
 190     return combs_out
 191
 192 def format_table_content(f, content, indent):
 193     line = " "*indent
 194     first = True
 195     for chunk in content.split(","):
 196         if len(line) + len(chunk) < 98:
 197             if first:
 198                 line += chunk
 199             else:
 200                 line += ", " + chunk
 201             first = False
 202         else:
 203             f.write(line + ",\n")
 204             line = " "*indent + chunk
 205     f.write(line)
 206
 207 def load_properties(f, interestingprops):
 208     fetch(f)
 209     props = {}
 210     re1 = re.compile("^([0-9A-F]+) +; (\w+)")
 211     re2 = re.compile("^([0-9A-F]+)\.\.([0-9A-F]+) +; (\w+)")
 212
 213     for line in fileinput.input(os.path.basename(f)):
 214         prop = None
 215         d_lo = 0
 216         d_hi = 0
 217         m = re1.match(line)
 218         if m:
 219             d_lo = m.group(1)
 220             d_hi = m.group(1)
 221             prop = m.group(2)
 222         else:
 223             m = re2.match(line)
 224             if m:
 225                 d_lo = m.group(1)
 226                 d_hi = m.group(2)
 227                 prop = m.group(3)
 228             else:
 229                 continue
 230         if interestingprops and prop not in interestingprops:
 231             continue
 232         d_lo = int(d_lo, 16)
 233         d_hi = int(d_hi, 16)
 234         if prop not in props:
 235             props[prop] = []
 236         props[prop].append((d_lo, d_hi))
 237     return props
 238
 239 # load all widths of want_widths, except those in except_cats
 240 def load_east_asian_width(want_widths, except_cats):
 241     f = "EastAsianWidth.txt"
 242     fetch(f)
 243     widths = {}
 244     re1 = re.compile("^([0-9A-F]+);(\w+) +# (\w+)")
 245     re2 = re.compile("^([0-9A-F]+)\.\.([0-9A-F]+);(\w+) +# (\w+)")
 246
 247     for line in fileinput.input(f):
 248         width = None
 249         d_lo = 0
 250         d_hi = 0
 251         cat = None
 252         m = re1.match(line)
 253         if m:
 254             d_lo = m.group(1)
 255             d_hi = m.group(1)
 256             width = m.group(2)
 257             cat = m.group(3)
 258         else:
 259             m = re2.match(line)
 260             if m:
 261                 d_lo = m.group(1)
 262                 d_hi = m.group(2)
 263                 width = m.group(3)
 264                 cat = m.group(4)
 265             else:
 266                 continue
 267         if cat in except_cats or width not in want_widths:
 268             continue
 269         d_lo = int(d_lo, 16)
 270         d_hi = int(d_hi, 16)
 271         if width not in widths:
 272             widths[width] = []
 273         widths[width].append((d_lo, d_hi))
 274     return widths
 275
 276 def escape_char(c):
 277     return "'\\u{%x}'" % c
 278
 279 def emit_bsearch_range_table(f):
 280     f.write("""
 281 fn bsearch_range_table(c: char, r: &'static [(char,char)]) -> bool {
 282     use core::cmp::Ordering::{Equal, Less, Greater};
 283     use core::slice::SliceExt;
 284     r.binary_search_by(|&(lo,hi)| {
 285         if lo <= c && c <= hi { Equal }
 286         else if hi < c { Less }
 287         else { Greater }
 288     }).is_ok()
 289 }\n
 290 """)
 291
 292 def emit_table(f, name, t_data, t_type = "&'static [(char, char)]", is_pub=True,
 293         pfun=lambda x: "(%s,%s)" % (escape_char(x[0]), escape_char(x[1]))):
 294     pub_string = ""
 295     if is_pub:
 296         pub_string = "pub "
 297     f.write("    %sconst %s: %s = &[\n" % (pub_string, name, t_type))
 298     data = ""
 299     first = True
 300     for dat in t_data:
 301         if not first:
 302             data += ","
 303         first = False
 304         data += pfun(dat)
 305     format_table_content(f, data, 8)
 306     f.write("\n    ];\n\n")
 307
 308 def emit_property_module(f, mod, tbl, emit):
 309     f.write("pub mod %s {\n" % mod)
 310     for cat in sorted(emit):
 311         emit_table(f, "%s_table" % cat, tbl[cat])
 312         f.write("    pub fn %s(c: char) -> bool {\n" % cat)
 313         f.write("        super::bsearch_range_table(c, %s_table)\n" % cat)
 314         f.write("    }\n\n")
 315     f.write("}\n\n")
 316
 317 def emit_conversions_module(f, lowerupper, upperlower):
 318     f.write("pub mod conversions {")
 319     f.write("""
 320     use core::cmp::Ordering::{Equal, Less, Greater};
 321     use core::slice::SliceExt;
 322     use core::option::Option;
 323     use core::option::Option::{Some, None};
 324     use core::result::Result::{Ok, Err};
 325
 326     pub fn to_lower(c: char) -> char {
 327         match bsearch_case_table(c, LuLl_table) {
 328           None        => c,
 329           Some(index) => LuLl_table[index].1
 330         }
 331     }
 332
 333     pub fn to_upper(c: char) -> char {
 334         match bsearch_case_table(c, LlLu_table) {
 335             None        => c,
 336             Some(index) => LlLu_table[index].1
 337         }
 338     }
 339
 340     fn bsearch_case_table(c: char, table: &'static [(char, char)]) -> Option<usize> {
 341         match table.binary_search_by(|&(key, _)| {
 342             if c == key { Equal }
 343             else if key < c { Less }
 344             else { Greater }
 345         }) {
 346             Ok(i) => Some(i),
 347             Err(_) => None,
 348         }
 349     }
 350
 351 """)
 352     emit_table(f, "LuLl_table",
 353         sorted(upperlower.iteritems(), key=operator.itemgetter(0)), is_pub=False)
 354     emit_table(f, "LlLu_table",
 355         sorted(lowerupper.iteritems(), key=operator.itemgetter(0)), is_pub=False)
 356     f.write("}\n\n")
 357
 358 def emit_grapheme_module(f, grapheme_table, grapheme_cats):
 359     f.write("""pub mod grapheme {
 360     use core::slice::SliceExt;
 361     pub use self::GraphemeCat::*;
 362     use core::result::Result::{Ok, Err};
 363
 364     #[allow(non_camel_case_types)]
 365     #[derive(Clone, Copy)]
 366     pub enum GraphemeCat {
 367 """)
 368     for cat in grapheme_cats + ["Any"]:
 369         f.write("        GC_" + cat + ",\n")
 370     f.write("""    }
 371
 372     fn bsearch_range_value_table(c: char, r: &'static [(char, char, GraphemeCat)]) -> GraphemeCat {
 373         use core::cmp::Ordering::{Equal, Less, Greater};
 374         match r.binary_search_by(|&(lo, hi, _)| {
 375             if lo <= c && c <= hi { Equal }
 376             else if hi < c { Less }
 377             else { Greater }
 378         }) {
 379             Ok(idx) => {
 380                 let (_, _, cat) = r[idx];
 381                 cat
 382             }
 383             Err(_) => GC_Any
 384         }
 385     }
 386
 387     pub fn grapheme_category(c: char) -> GraphemeCat {
 388         bsearch_range_value_table(c, grapheme_cat_table)
 389     }
 390
 391 """)
 392
 393     emit_table(f, "grapheme_cat_table", grapheme_table, "&'static [(char, char, GraphemeCat)]",
 394         pfun=lambda x: "(%s,%s,GC_%s)" % (escape_char(x[0]), escape_char(x[1]), x[2]),
 395         is_pub=False)
 396     f.write("}\n")
 397
 398 def emit_charwidth_module(f, width_table):
 399     f.write("pub mod charwidth {\n")
 400     f.write("    use core::option::Option;\n")
 401     f.write("    use core::option::Option::{Some, None};\n")
 402     f.write("    use core::slice::SliceExt;\n")
 403     f.write("    use core::result::Result::{Ok, Err};\n")
 404     f.write("""
 405     fn bsearch_range_value_table(c: char, is_cjk: bool, r: &'static [(char, char, u8, u8)]) -> u8 {
 406         use core::cmp::Ordering::{Equal, Less, Greater};
 407         match r.binary_search_by(|&(lo, hi, _, _)| {
 408             if lo <= c && c <= hi { Equal }
 409             else if hi < c { Less }
 410             else { Greater }
 411         }) {
 412             Ok(idx) => {
 413                 let (_, _, r_ncjk, r_cjk) = r[idx];
 414                 if is_cjk { r_cjk } else { r_ncjk }
 415             }
 416             Err(_) => 1
 417         }
 418     }
 419 """)
 420
 421     f.write("""
 422     pub fn width(c: char, is_cjk: bool) -> Option<usize> {
 423         match c as usize {
 424             _c @ 0 => Some(0),          // null is zero width
 425             cu if cu < 0x20 => None,    // control sequences have no width
 426             cu if cu < 0x7F => Some(1), // ASCII
 427             cu if cu < 0xA0 => None,    // more control sequences
 428             _ => Some(bsearch_range_value_table(c, is_cjk, charwidth_table) as usize)
 429         }
 430     }
 431
 432 """)
 433
 434     f.write("    // character width table. Based on Markus Kuhn's free wcwidth() implementation,\n")
 435     f.write("    //     http://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c\n")
 436     emit_table(f, "charwidth_table", width_table, "&'static [(char, char, u8, u8)]", is_pub=False,
 437             pfun=lambda x: "(%s,%s,%s,%s)" % (escape_char(x[0]), escape_char(x[1]), x[2], x[3]))
 438     f.write("}\n\n")
 439
 440 def emit_norm_module(f, canon, compat, combine, norm_props):
 441     canon_keys = canon.keys()
 442     canon_keys.sort()
 443
 444     compat_keys = compat.keys()
 445     compat_keys.sort()
 446
 447     canon_comp = {}
 448     comp_exclusions = norm_props["Full_Composition_Exclusion"]
 449     for char in canon_keys:
 450         if True in map(lambda (lo, hi): lo <= char <= hi, comp_exclusions):
 451             continue
 452         decomp = canon[char]
 453         if len(decomp) == 2:
 454             if not canon_comp.has_key(decomp[0]):
 455                 canon_comp[decomp[0]] = []
 456             canon_comp[decomp[0]].append( (decomp[1], char) )
 457     canon_comp_keys = canon_comp.keys()
 458     canon_comp_keys.sort()
 459
 460     f.write("pub mod normalization {\n")
 461
 462     def mkdata_fun(table):
 463         def f(char):
 464             data = "(%s,&[" % escape_char(char)
 465             first = True
 466             for d in table[char]:
 467                 if not first:
 468                     data += ","
 469                 first = False
 470                 data += escape_char(d)
 471             data += "])"
 472             return data
 473         return f
 474
 475     f.write("    // Canonical decompositions\n")
 476     emit_table(f, "canonical_table", canon_keys, "&'static [(char, &'static [char])]",
 477         pfun=mkdata_fun(canon))
 478
 479     f.write("    // Compatibility decompositions\n")
 480     emit_table(f, "compatibility_table", compat_keys, "&'static [(char, &'static [char])]",
 481         pfun=mkdata_fun(compat))
 482
 483     def comp_pfun(char):
 484         data = "(%s,&[" % escape_char(char)
 485         canon_comp[char].sort(lambda x, y: x[0] - y[0])
 486         first = True
 487         for pair in canon_comp[char]:
 488             if not first:
 489                 data += ","
 490             first = False
 491             data += "(%s,%s)" % (escape_char(pair[0]), escape_char(pair[1]))
 492         data += "])"
 493         return data
 494
 495     f.write("    // Canonical compositions\n")
 496     emit_table(f, "composition_table", canon_comp_keys,
 497         "&'static [(char, &'static [(char, char)])]", pfun=comp_pfun)
 498
 499     f.write("""
 500     fn bsearch_range_value_table(c: char, r: &'static [(char, char, u8)]) -> u8 {
 501         use core::cmp::Ordering::{Equal, Less, Greater};
 502         use core::slice::SliceExt;
 503         use core::result::Result::{Ok, Err};
 504         match r.binary_search_by(|&(lo, hi, _)| {
 505             if lo <= c && c <= hi { Equal }
 506             else if hi < c { Less }
 507             else { Greater }
 508         }) {
 509             Ok(idx) => {
 510                 let (_, _, result) = r[idx];
 511                 result
 512             }
 513             Err(_) => 0
 514         }
 515     }\n
 516 """)
 517
 518     emit_table(f, "combining_class_table", combine, "&'static [(char, char, u8)]", is_pub=False,
 519             pfun=lambda x: "(%s,%s,%s)" % (escape_char(x[0]), escape_char(x[1]), x[2]))
 520
 521     f.write("    pub fn canonical_combining_class(c: char) -> u8 {\n"
 522         + "        bsearch_range_value_table(c, combining_class_table)\n"
 523         + "    }\n")
 524
 525     f.write("""
 526 }
 527
 528 """)
 529
 530 def remove_from_wtable(wtable, val):
 531     wtable_out = []
 532     while wtable:
 533         if wtable[0][1] < val:
 534             wtable_out.append(wtable.pop(0))
 535         elif wtable[0][0] > val:
 536             break
 537         else:
 538             (wt_lo, wt_hi, width, width_cjk) = wtable.pop(0)
 539             if wt_lo == wt_hi == val:
 540                 continue
 541             elif wt_lo == val:
 542                 wtable_out.append((wt_lo+1, wt_hi, width, width_cjk))
 543             elif wt_hi == val:
 544                 wtable_out.append((wt_lo, wt_hi-1, width, width_cjk))
 545             else:
 546                 wtable_out.append((wt_lo, val-1, width, width_cjk))
 547                 wtable_out.append((val+1, wt_hi, width, width_cjk))
 548     if wtable:
 549         wtable_out.extend(wtable)
 550     return wtable_out
 551
 552
 553
 554 def optimize_width_table(wtable):
 555     wtable_out = []
 556     w_this = wtable.pop(0)
 557     while wtable:
 558         if w_this[1] == wtable[0][0] - 1 and w_this[2:3] == wtable[0][2:3]:
 559             w_tmp = wtable.pop(0)
 560             w_this = (w_this[0], w_tmp[1], w_tmp[2], w_tmp[3])
 561         else:
 562             wtable_out.append(w_this)
 563             w_this = wtable.pop(0)
 564     wtable_out.append(w_this)
 565     return wtable_out
 566
 567 if __name__ == "__main__":
 568     r = "tables.rs"
 569     if os.path.exists(r):
 570         os.remove(r)
 571     with open(r, "w") as rf:
 572         # write the file's preamble
 573         rf.write(preamble)
 574
 575         # download and parse all the data
 576         fetch("ReadMe.txt")
 577         with open("ReadMe.txt") as readme:
 578             pattern = "for Version (\d+)\.(\d+)\.(\d+) of the Unicode"
 579             unicode_version = re.search(pattern, readme.read()).groups()
 580         rf.write("""
 581 /// The version of [Unicode](http://www.unicode.org/)
 582 /// that the unicode parts of `CharExt` and `UnicodeStrPrelude` traits are based on.
 583 pub const UNICODE_VERSION: (u64, u64, u64) = (%s, %s, %s);
 584 """ % unicode_version)
 585         (canon_decomp, compat_decomp, gencats, combines,
 586                 lowerupper, upperlower) = load_unicode_data("UnicodeData.txt")
 587         want_derived = ["XID_Start", "XID_Continue", "Alphabetic", "Lowercase", "Uppercase"]
 588         derived = load_properties("DerivedCoreProperties.txt", want_derived)
 589         scripts = load_properties("Scripts.txt", [])
 590         props = load_properties("PropList.txt",
 591                 ["White_Space", "Join_Control", "Noncharacter_Code_Point"])
 592         norm_props = load_properties("DerivedNormalizationProps.txt",
 593                      ["Full_Composition_Exclusion"])
 594
 595         # bsearch_range_table is used in all the property modules below
 596         emit_bsearch_range_table(rf)
 597
 598         # category tables
 599         for (name, cat, pfuns) in ("general_category", gencats, ["N", "Cc"]), \
 600                                   ("derived_property", derived, want_derived), \
 601                                   ("property", props, ["White_Space"]):
 602             emit_property_module(rf, name, cat, pfuns)
 603
 604         # normalizations and conversions module
 605         emit_norm_module(rf, canon_decomp, compat_decomp, combines, norm_props)
 606         emit_conversions_module(rf, lowerupper, upperlower)
 607
 608         ### character width module
 609         width_table = []
 610         for zwcat in ["Me", "Mn", "Cf"]:
 611             width_table.extend(map(lambda (lo, hi): (lo, hi, 0, 0), gencats[zwcat]))
 612         width_table.append((4448, 4607, 0, 0))
 613
 614         # get widths, except those that are explicitly marked zero-width above
 615         ea_widths = load_east_asian_width(["W", "F", "A"], ["Me", "Mn", "Cf"])
 616         # these are doublewidth
 617         for dwcat in ["W", "F"]:
 618             width_table.extend(map(lambda (lo, hi): (lo, hi, 2, 2), ea_widths[dwcat]))
 619         width_table.extend(map(lambda (lo, hi): (lo, hi, 1, 2), ea_widths["A"]))
 620
 621         width_table.sort(key=lambda w: w[0])
 622
 623         # soft hyphen is not zero width in preformatted text; it's used to indicate
 624         # a hyphen inserted to facilitate a linebreak.
 625         width_table = remove_from_wtable(width_table, 173)
 626
 627         # optimize the width table by collapsing adjacent entities when possible
 628         width_table = optimize_width_table(width_table)
 629         emit_charwidth_module(rf, width_table)
 630
 631         ### grapheme cluster module
 632         # from http://www.unicode.org/reports/tr29/#Grapheme_Cluster_Break_Property_Values
 633         grapheme_cats = load_properties("auxiliary/GraphemeBreakProperty.txt", [])
 634
 635         # Control
 636         #  Note 1:
 637         # This category also includes Cs (surrogate codepoints), but Rust's `char`s are
 638         # Unicode Scalar Values only, and surrogates are thus invalid `char`s.
 639         # Thus, we have to remove Cs from the Control category
 640         #  Note 2:
 641         # 0x0a and 0x0d (CR and LF) are not in the Control category for Graphemes.
 642         # However, the Graphemes iterator treats these as a special case, so they
 643         # should be included in grapheme_cats["Control"] for our implementation.
 644         grapheme_cats["Control"] = group_cat(list(
 645             (set(ungroup_cat(grapheme_cats["Control"]))
 646              | set(ungroup_cat(grapheme_cats["CR"]))
 647              | set(ungroup_cat(grapheme_cats["LF"])))
 648             - set(ungroup_cat([surrogate_codepoints]))))
 649         del(grapheme_cats["CR"])
 650         del(grapheme_cats["LF"])
 651
 652         grapheme_table = []
 653         for cat in grapheme_cats:
 654             grapheme_table.extend([(x, y, cat) for (x, y) in grapheme_cats[cat]])
 655         grapheme_table.sort(key=lambda w: w[0])
 656         emit_grapheme_module(rf, grapheme_table, grapheme_cats.keys())