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.
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.
13 # This script uses the following Unicode tables:
14 # - DerivedCoreProperties.txt
15 # - DerivedNormalizationProps.txt
16 # - EastAsianWidth.txt
17 # - auxiliary/GraphemeBreakProperty.txt
23 # Since this should not require frequent updates, we just store this
24 # out-of-line and check the unicode.rs file into git.
26 import fileinput, re, os, sys, operator, math
28 preamble = '''// Copyright 2012-2016 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.
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.
38 // NOTE: The following code was generated by "./unicode.py", do not edit directly
40 #![allow(missing_docs, non_upper_case_globals, non_snake_case)]
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'],
57 # these are the surrogate codepoints, which are not valid rust characters
58 surrogate_codepoints = (0xd800, 0xdfff)
61 if not os.path.exists(os.path.basename(f)):
62 os.system("curl -O http://www.unicode.org/Public/UNIDATA/%s"
65 if not os.path.exists(os.path.basename(f)):
66 sys.stderr.write("cannot load %s" % f)
70 return surrogate_codepoints[0] <= n <= surrogate_codepoints[1]
72 def load_unicode_data(f):
84 for line in fileinput.input(f):
85 data = line.split(';')
92 for i in xrange(range_start, cp):
95 if data[1].endswith(", First>"):
101 (code_org, name, gencat, combine, bidi,
102 decomp, deci, digit, num, mirror,
103 old, iso, upcase, lowcase, titlecase) = udict[code]
105 # generate char to char direct common and simple conversions
106 # uppercase to lowercase
107 if lowcase != "" and code_org != lowcase:
108 to_lower[code] = (int(lowcase, 16), 0, 0)
110 # lowercase to uppercase
111 if upcase != "" and code_org != upcase:
112 to_upper[code] = (int(upcase, 16), 0, 0)
115 if titlecase.strip() != "" and code_org != titlecase:
116 to_title[code] = (int(titlecase, 16), 0, 0)
118 # store decomposition, if given
120 if decomp.startswith('<'):
122 for i in decomp.split()[1:]:
123 seq.append(int(i, 16))
124 compat_decomp[code] = seq
127 for i in decomp.split():
128 seq.append(int(i, 16))
129 canon_decomp[code] = seq
131 # place letter in categories as appropriate
132 for cat in [gencat, "Assigned"] + expanded_categories.get(gencat, []):
133 if cat not in gencats:
135 gencats[cat].append(code)
137 # record combining class, if any
139 if combine not in combines:
140 combines[combine] = []
141 combines[combine].append(code)
143 # generate Not_Assigned from Assigned
144 gencats["Cn"] = gen_unassigned(gencats["Assigned"])
145 # Assigned is not a real category
146 del(gencats["Assigned"])
147 # Other contains Not_Assigned
148 gencats["C"].extend(gencats["Cn"])
149 gencats = group_cats(gencats)
150 combines = to_combines(group_cats(combines))
152 return (canon_decomp, compat_decomp, gencats, combines, to_upper, to_lower, to_title)
154 def load_special_casing(f, to_upper, to_lower, to_title):
156 for line in fileinput.input(f):
157 data = line.split('#')[0].split(';')
159 code, lower, title, upper, _comment = data
161 code, lower, title, upper, condition, _comment = data
162 if condition.strip(): # Only keep unconditional mappins
167 lower = lower.strip()
168 title = title.strip()
169 upper = upper.strip()
171 for (map_, values) in [(to_lower, lower), (to_upper, upper), (to_title, title)]:
173 values = [int(i, 16) for i in values.split()]
174 for _ in range(len(values), 3):
176 assert len(values) == 3
179 def group_cats(cats):
182 cats_out[cat] = group_cat(cats[cat])
187 letters = sorted(set(cat))
188 cur_start = letters.pop(0)
190 for letter in letters:
191 assert letter > cur_end, \
192 "cur_end: %s, letter: %s" % (hex(cur_end), hex(letter))
193 if letter == cur_end + 1:
196 cat_out.append((cur_start, cur_end))
197 cur_start = cur_end = letter
198 cat_out.append((cur_start, cur_end))
201 def ungroup_cat(cat):
209 def gen_unassigned(assigned):
210 assigned = set(assigned)
211 return ([i for i in range(0, 0xd800) if i not in assigned] +
212 [i for i in range(0xe000, 0x110000) if i not in assigned])
214 def to_combines(combs):
217 for (lo, hi) in combs[comb]:
218 combs_out.append((lo, hi, comb))
219 combs_out.sort(key=lambda comb: comb[0])
222 def format_table_content(f, content, indent):
225 for chunk in content.split(","):
226 if len(line) + len(chunk) < 98:
233 f.write(line + ",\n")
234 line = " "*indent + chunk
237 def load_properties(f, interestingprops):
240 re1 = re.compile("^ *([0-9A-F]+) *; *(\w+)")
241 re2 = re.compile("^ *([0-9A-F]+)\.\.([0-9A-F]+) *; *(\w+)")
243 for line in fileinput.input(os.path.basename(f)):
260 if interestingprops and prop not in interestingprops:
264 if prop not in props:
266 props[prop].append((d_lo, d_hi))
268 # optimize if possible
270 props[prop] = group_cat(ungroup_cat(props[prop]))
275 return "'\\u{%x}'" % c if c != 0 else "'\\0'"
277 def emit_bsearch_range_table(f):
279 fn bsearch_range_table(c: char, r: &'static [(char, char)]) -> bool {
280 use core::cmp::Ordering::{Equal, Less, Greater};
281 r.binary_search_by(|&(lo, hi)| {
294 def emit_table(f, name, t_data, t_type = "&'static [(char, char)]", is_pub=True,
295 pfun=lambda x: "(%s,%s)" % (escape_char(x[0]), escape_char(x[1]))):
299 f.write(" %sconst %s: %s = &[\n" % (pub_string, name, t_type))
307 format_table_content(f, data, 8)
310 def emit_trie_lookup_range_table(f):
313 // BoolTrie is a trie for representing a set of Unicode codepoints. It is
314 // implemented with postfix compression (sharing of identical child nodes),
315 // which gives both compact size and fast lookup.
317 // The space of Unicode codepoints is divided into 3 subareas, each
318 // represented by a trie with different depth. In the first (0..0x800), there
319 // is no trie structure at all; each u64 entry corresponds to a bitvector
320 // effectively holding 64 bool values.
322 // In the second (0x800..0x10000), each child of the root node represents a
323 // 64-wide subrange, but instead of storing the full 64-bit value of the leaf,
324 // the trie stores an 8-bit index into a shared table of leaf values. This
325 // exploits the fact that in reasonable sets, many such leaves can be shared.
327 // In the third (0x10000..0x110000), each child of the root node represents a
328 // 4096-wide subrange, and the trie stores an 8-bit index into a 64-byte slice
329 // of a child tree. Each of these 64 bytes represents an index into the table
330 // of shared 64-bit leaf values. This exploits the sparse structure in the
331 // non-BMP range of most Unicode sets.
332 pub struct BoolTrie {
333 // 0..0x800 (corresponding to 1 and 2 byte utf-8 sequences)
334 r1: [u64; 32], // leaves
336 // 0x800..0x10000 (corresponding to 3 byte utf-8 sequences)
337 r2: [u8; 992], // first level
338 r3: &'static [u64], // leaves
340 // 0x10000..0x110000 (corresponding to 4 byte utf-8 sequences)
341 r4: [u8; 256], // first level
342 r5: &'static [u8], // second level
343 r6: &'static [u64], // leaves
346 fn trie_range_leaf(c: usize, bitmap_chunk: u64) -> bool {
347 ((bitmap_chunk >> (c & 63)) & 1) != 0
350 fn trie_lookup_range_table(c: char, r: &'static BoolTrie) -> bool {
353 trie_range_leaf(c, r.r1[c >> 6])
354 } else if c < 0x10000 {
355 let child = r.r2[(c >> 6) - 0x20];
356 trie_range_leaf(c, r.r3[child as usize])
358 let child = r.r4[(c >> 12) - 0x10];
359 let leaf = r.r5[((child as usize) << 6) + ((c >> 6) & 0x3f)];
360 trie_range_leaf(c, r.r6[leaf as usize])
364 pub struct SmallBoolTrie {
365 r1: &'static [u8], // first level
366 r2: &'static [u64], // leaves
370 fn lookup(&self, c: char) -> bool {
372 match self.r1.get(c >> 6) {
373 Some(&child) => trie_range_leaf(c, self.r2[child as usize]),
381 def compute_trie(rawdata, chunksize):
385 for i in range(len(rawdata) / chunksize):
386 data = rawdata[i * chunksize: (i + 1) * chunksize]
387 child = '|'.join(map(str, data))
388 if child not in childmap:
389 childmap[child] = len(childmap)
390 child_data.extend(data)
391 root.append(childmap[child])
392 return (root, child_data)
394 def emit_bool_trie(f, name, t_data, is_pub=True):
396 rawdata = [False] * 0x110000
397 for (lo, hi) in t_data:
398 for cp in range(lo, hi + 1):
401 # convert to bitmap chunks of 64 bits each
403 for i in range(0x110000 / CHUNK):
406 if rawdata[i * 64 + j]:
413 f.write(" %sconst %s: &'static super::BoolTrie = &super::BoolTrie {\n" % (pub_string, name))
415 data = ','.join('0x%016x' % chunk for chunk in chunks[0:0x800 / CHUNK])
416 format_table_content(f, data, 12)
419 # 0x800..0x10000 trie
420 (r2, r3) = compute_trie(chunks[0x800 / CHUNK : 0x10000 / CHUNK], 64 / CHUNK)
422 data = ','.join(str(node) for node in r2)
423 format_table_content(f, data, 12)
426 data = ','.join('0x%016x' % chunk for chunk in r3)
427 format_table_content(f, data, 12)
430 # 0x10000..0x110000 trie
431 (mid, r6) = compute_trie(chunks[0x10000 / CHUNK : 0x110000 / CHUNK], 64 / CHUNK)
432 (r4, r5) = compute_trie(mid, 64)
434 data = ','.join(str(node) for node in r4)
435 format_table_content(f, data, 12)
438 data = ','.join(str(node) for node in r5)
439 format_table_content(f, data, 12)
442 data = ','.join('0x%016x' % chunk for chunk in r6)
443 format_table_content(f, data, 12)
448 def emit_small_bool_trie(f, name, t_data, is_pub=True):
449 last_chunk = max(int(hi / 64) for (lo, hi) in t_data)
450 n_chunks = last_chunk + 1
451 chunks = [0] * n_chunks
452 for (lo, hi) in t_data:
453 for cp in range(lo, hi + 1):
454 if int(cp / 64) >= len(chunks):
455 print(cp, int(cp / 64), len(chunks), lo, hi)
456 chunks[int(cp / 64)] |= 1 << (cp & 63)
461 f.write(" %sconst %s: &'static super::SmallBoolTrie = &super::SmallBoolTrie {\n"
462 % (pub_string, name))
464 (r1, r2) = compute_trie(chunks, 1)
467 data = ','.join(str(node) for node in r1)
468 format_table_content(f, data, 12)
472 data = ','.join('0x%016x' % node for node in r2)
473 format_table_content(f, data, 12)
478 def emit_property_module(f, mod, tbl, emit):
479 f.write("pub mod %s {\n" % mod)
480 for cat in sorted(emit):
481 if cat in ["Cc", "White_Space", "Pattern_White_Space"]:
482 emit_small_bool_trie(f, "%s_table" % cat, tbl[cat])
483 f.write(" pub fn %s(c: char) -> bool {\n" % cat)
484 f.write(" %s_table.lookup(c)\n" % cat)
487 emit_bool_trie(f, "%s_table" % cat, tbl[cat])
488 f.write(" pub fn %s(c: char) -> bool {\n" % cat)
489 f.write(" super::trie_lookup_range_table(c, %s_table)\n" % cat)
493 def emit_conversions_module(f, to_upper, to_lower, to_title):
494 f.write("pub mod conversions {")
496 use core::option::Option;
497 use core::option::Option::{Some, None};
499 pub fn to_lower(c: char) -> [char; 3] {
500 match bsearch_case_table(c, to_lowercase_table) {
501 None => [c, '\\0', '\\0'],
502 Some(index) => to_lowercase_table[index].1,
506 pub fn to_upper(c: char) -> [char; 3] {
507 match bsearch_case_table(c, to_uppercase_table) {
508 None => [c, '\\0', '\\0'],
509 Some(index) => to_uppercase_table[index].1,
513 fn bsearch_case_table(c: char, table: &'static [(char, [char; 3])]) -> Option<usize> {
514 table.binary_search_by(|&(key, _)| key.cmp(&c)).ok()
518 t_type = "&'static [(char, [char; 3])]"
519 pfun = lambda x: "(%s,[%s,%s,%s])" % (
520 escape_char(x[0]), escape_char(x[1][0]), escape_char(x[1][1]), escape_char(x[1][2]))
521 emit_table(f, "to_lowercase_table",
522 sorted(to_lower.iteritems(), key=operator.itemgetter(0)),
523 is_pub=False, t_type = t_type, pfun=pfun)
524 emit_table(f, "to_uppercase_table",
525 sorted(to_upper.iteritems(), key=operator.itemgetter(0)),
526 is_pub=False, t_type = t_type, pfun=pfun)
529 def emit_norm_module(f, canon, compat, combine, norm_props):
530 canon_keys = canon.keys()
533 compat_keys = compat.keys()
537 comp_exclusions = norm_props["Full_Composition_Exclusion"]
538 for char in canon_keys:
539 if True in map(lambda (lo, hi): lo <= char <= hi, comp_exclusions):
543 if not canon_comp.has_key(decomp[0]):
544 canon_comp[decomp[0]] = []
545 canon_comp[decomp[0]].append( (decomp[1], char) )
546 canon_comp_keys = canon_comp.keys()
547 canon_comp_keys.sort()
549 if __name__ == "__main__":
551 if os.path.exists(r):
553 with open(r, "w") as rf:
554 # write the file's preamble
557 # download and parse all the data
559 with open("ReadMe.txt") as readme:
560 pattern = "for Version (\d+)\.(\d+)\.(\d+) of the Unicode"
561 unicode_version = re.search(pattern, readme.read()).groups()
563 /// The version of [Unicode](http://www.unicode.org/)
564 /// that the unicode parts of `CharExt` and `UnicodeStrPrelude` traits are based on.
565 pub const UNICODE_VERSION: (u64, u64, u64) = (%s, %s, %s);
566 """ % unicode_version)
567 (canon_decomp, compat_decomp, gencats, combines,
568 to_upper, to_lower, to_title) = load_unicode_data("UnicodeData.txt")
569 load_special_casing("SpecialCasing.txt", to_upper, to_lower, to_title)
570 want_derived = ["XID_Start", "XID_Continue", "Alphabetic", "Lowercase", "Uppercase",
571 "Cased", "Case_Ignorable"]
572 derived = load_properties("DerivedCoreProperties.txt", want_derived)
573 scripts = load_properties("Scripts.txt", [])
574 props = load_properties("PropList.txt",
575 ["White_Space", "Join_Control", "Noncharacter_Code_Point", "Pattern_White_Space"])
576 norm_props = load_properties("DerivedNormalizationProps.txt",
577 ["Full_Composition_Exclusion"])
579 # trie_lookup_table is used in all the property modules below
580 emit_trie_lookup_range_table(rf)
581 # emit_bsearch_range_table(rf)
584 for (name, cat, pfuns) in ("general_category", gencats, ["N", "Cc"]), \
585 ("derived_property", derived, want_derived), \
586 ("property", props, ["White_Space", "Pattern_White_Space"]):
587 emit_property_module(rf, name, cat, pfuns)
589 # normalizations and conversions module
590 emit_norm_module(rf, canon_decomp, compat_decomp, combines, norm_props)
591 emit_conversions_module(rf, to_upper, to_lower, to_title)