xref: /llvm-project/llvm/utils/UnicodeData/UnicodeNameMappingGenerator.cpp (revision 380a1b204c36cc8c6fdf28618d8a838aed8b7197)
1 //===--- UnicodeNameMappingGenerator.cpp - Unicode name data generator ---===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file is used to generate lib/Support/UnicodeNameToCodepointGenerated.cpp
10 // using UnicodeData.txt and NameAliases.txt available at
11 // https://unicode.org/Public/14.0.0/ucd/
12 //===----------------------------------------------------------------------===//
13 
14 #include "llvm/ADT/Optional.h"
15 #include "llvm/ADT/STLExtras.h"
16 #include "llvm/ADT/StringExtras.h"
17 #include "llvm/ADT/StringRef.h"
18 #include <algorithm>
19 #include <array>
20 #include <deque>
21 #include <fstream>
22 #include <memory>
23 #include <set>
24 #include <string>
25 #include <unordered_map>
26 #include <utility>
27 #include <vector>
28 
29 static const llvm::StringRef Letters =
30     " _-ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789";
31 
32 // Collect names UnicodeData.txt and AliasNames.txt
33 // There may be multiple names per code points.
34 static std::unordered_multimap<char32_t, std::string>
35 loadDataFiles(const std::string &NamesFile, const std::string &AliasesFile) {
36   std::unordered_multimap<char32_t, std::string> CollectedCharacters;
37   auto FromFile = [&](const std::string &File, bool IsAliasFile = false) {
38     std::ifstream InputFile(File);
39     for (std::string Line; getline(InputFile, Line);) {
40       if (Line.empty() || !isxdigit(Line[0]))
41         continue;
42       auto FirstSemiPos = Line.find(';');
43       if (FirstSemiPos == std::string::npos)
44         continue;
45       auto SecondSemiPos = Line.find(';', FirstSemiPos + 1);
46       if (FirstSemiPos == std::string::npos)
47         continue;
48       unsigned long long CodePoint;
49       if (llvm::getAsUnsignedInteger(
50               llvm::StringRef(Line.c_str(), FirstSemiPos), 16, CodePoint)) {
51         continue;
52       }
53 
54       std::string Name =
55           Line.substr(FirstSemiPos + 1, SecondSemiPos - FirstSemiPos - 1);
56 
57       if (!Name.empty() && Name[0] == '<') {
58         // Ignore ranges of characters, as their name is either absent or
59         // generated.
60         continue;
61       }
62 
63       // Some aliases are ignored for compatibility with C++
64       if (IsAliasFile) {
65         std::string Kind = Line.substr(SecondSemiPos + 1);
66         if (Kind != "control" && Kind != "correction" && Kind != "alternate")
67           continue;
68       }
69 
70       auto InsertUnique = [&](char32_t CP, std::string Name) {
71         auto It = CollectedCharacters.find(CP);
72         while (It != std::end(CollectedCharacters) && It->first == CP) {
73           if (It->second == Name)
74             return;
75           ++It;
76         }
77         CollectedCharacters.insert({CP, std::move(Name)});
78       };
79       InsertUnique(CodePoint, std::move(Name));
80     }
81   };
82 
83   FromFile(NamesFile);
84   FromFile(AliasesFile, true);
85   return CollectedCharacters;
86 }
87 
88 class Trie {
89   struct Node;
90 
91 public:
92   // When inserting named codepoint
93   // We create a node per character in the name.
94   // SPARKLE becomes S <- P <- A <- R <- K <- L <- E
95   // Once all  characters are inserted, the tree is compacted
96   void insert(llvm::StringRef Name, char32_t Codepoint) {
97     Node *N = Root.get();
98     for (auto Ch : Name) {
99       std::string Label(1, Ch);
100       auto It = std::find_if(N->Children.begin(), N->Children.end(),
101                              [&](const auto &C) { return C->Name == Label; });
102       if (It == N->Children.end()) {
103         It = N->Children.insert(It, std::make_unique<Node>(Label, N));
104       }
105       N = It->get();
106     }
107     N->Value = Codepoint;
108   }
109 
110   void compact() { compact(Root.get()); }
111 
112   // This creates 2 arrays of bytes from the tree:
113   // A serialized dictionary of node labels,
114   // And the nodes themselves.
115   // The name of each label is found by indexing into the dictionary.
116   // The longest names are inserted first into the dictionary,
117   // in the hope it will contain shorter labels as substring,
118   // thereby reducing duplication.
119   // We could theorically be more clever by trying to minimizing the size
120   // of the dictionary.
121   std::pair<std::string, std::vector<uint8_t>> serialize() {
122     std::set<std::string> Names = this->getNameFragments();
123     std::vector<std::string> Sorted(Names.begin(), Names.end());
124     std::sort(Sorted.begin(), Sorted.end(),
125               [](const auto &a, const auto &b) { return a.size() > b.size(); });
126     std::string Dict(Letters.begin(), Letters.end());
127     Dict.reserve(50000);
128     for (const std::string &Name : Sorted) {
129       if (Name.size() <= 1)
130         continue;
131       if (Dict.find(Name) != std::string::npos)
132         continue;
133       Dict += Name;
134     }
135 
136     if (Dict.size() >= std::numeric_limits<uint16_t>::max()) {
137       fprintf(stderr, "Dictionary too big  to be serialized");
138       exit(1);
139     }
140 
141     auto Bytes = dumpIndex(Dict);
142     return {Dict, Bytes};
143   }
144 
145   std::set<std::string> getNameFragments() {
146     std::set<std::string> Keys;
147     collectKeys(Root.get(), Keys);
148     return Keys;
149   }
150 
151   // Maps a valid char in an Unicode character name
152   // To a 6 bits index.
153   static uint8_t letter(char C) {
154     auto Pos = Letters.find(C);
155     assert(Pos != std::string::npos &&
156            "Invalid letter in Unicode character name");
157     return Pos;
158   }
159 
160   // clang-format off
161   // +================+============+======================+=============+========+===+==============+===============+
162   // | 0          | 1             | 2-7 (6)              | 8-23        | 24-44  |    | 46           | 47            |
163   // +================+============+======================+=============+========+===+==============+===============+
164   // | Has Value |  Has Long Name | Letter OR Name Size  | Dict Index  | Value  |    | Has Sibling  | Has Children  |
165   // +----------------+------------+----------------------+-------------+--------+---+--------------+---------------+
166   // clang-format on
167 
168   std::vector<uint8_t> dumpIndex(const std::string &Dict) {
169     struct ChildrenOffset {
170       Node *FirstChild;
171       std::size_t Offset;
172       bool HasValue;
173     };
174 
175     // Keep track of the start of each node
176     // position in the serialized data.
177     std::unordered_map<Node *, int32_t> Offsets;
178 
179     // Keep track of where to write the index
180     // of the first children
181     std::vector<ChildrenOffset> ChildrenOffsets;
182     std::unordered_map<Node *, bool> SiblingTracker;
183     std::deque<Node *> AllNodes;
184     std::vector<uint8_t> Bytes;
185     Bytes.reserve(250'000);
186     // This leading byte is used by the reading code to detect the root node.
187     Bytes.push_back(0);
188 
189     auto CollectChildren = [&SiblingTracker, &AllNodes](const auto &Children) {
190       for (std::size_t Index = 0; Index < Children.size(); Index++) {
191         const std::unique_ptr<Node> &Child = Children[Index];
192         AllNodes.push_back(Child.get());
193         if (Index != Children.size() - 1)
194           SiblingTracker[Child.get()] = true;
195       }
196     };
197     CollectChildren(Root->Children);
198 
199     while (!AllNodes.empty()) {
200       const std::size_t Offset = Bytes.size();
201       Node *const N = AllNodes.front();
202       AllNodes.pop_front();
203 
204       assert(!N->Name.empty());
205       Offsets[N] = Offset;
206 
207       uint8_t FirstByte = (!!N->Value) ? 0x80 : 0;
208       // Single letter node are indexed in 6 bits
209       if (N->Name.size() == 1) {
210         FirstByte |= letter(N->Name[0]);
211         Bytes.push_back(FirstByte);
212       } else {
213         // Otherwise we use a 16 bits index
214         FirstByte = FirstByte | uint8_t(N->Name.size()) | 0x40;
215         Bytes.push_back(FirstByte);
216         auto PosInDict = Dict.find(N->Name);
217         assert(PosInDict != std::string::npos);
218         uint8_t Low = PosInDict;
219         uint8_t High = ((PosInDict >> 8) & 0xFF);
220         Bytes.push_back(High);
221         Bytes.push_back(Low);
222       }
223 
224       const bool HasSibling = SiblingTracker.count(N) != 0;
225       const bool HasChildren = N->Children.size() != 0;
226 
227       if (!!N->Value) {
228         uint32_t Value = (*(N->Value) << 3);
229         uint8_t H = ((Value >> 16) & 0xFF);
230         uint8_t M = ((Value >> 8) & 0xFF);
231         uint8_t L = (Value & 0xFF) | uint8_t(HasSibling ? 0x01 : 0) |
232                     uint8_t(HasChildren ? 0x02 : 0);
233 
234         Bytes.push_back(H);
235         Bytes.push_back(M);
236         Bytes.push_back(L);
237 
238         if (HasChildren) {
239           ChildrenOffsets.push_back(
240               ChildrenOffset{N->Children[0].get(), Bytes.size(), true});
241           // index of the first children
242           Bytes.push_back(0x00);
243           Bytes.push_back(0x00);
244           Bytes.push_back(0x00);
245         }
246       } else {
247         // When there is no value (that's most intermediate nodes)
248         // Dispense of the 3 values bytes, and only store
249         // 1 byte to track whether the node has sibling and chidren
250         // + 2 bytes for the index of the first children if necessary.
251         // That index also uses bytes 0-6 of the previous byte.
252         uint8_t Byte =
253             uint8_t(HasSibling ? 0x80 : 0) | uint8_t(HasChildren ? 0x40 : 0);
254         Bytes.push_back(Byte);
255         if (HasChildren) {
256           ChildrenOffsets.emplace_back(
257               ChildrenOffset{N->Children[0].get(), Bytes.size() - 1, false});
258           Bytes.push_back(0x00);
259           Bytes.push_back(0x00);
260         }
261       }
262       CollectChildren(N->Children);
263     }
264 
265     // Once all the nodes are in the inndex
266     // Fill the bytes we left to indicate the position
267     // of the children
268     for (const ChildrenOffset &Parent : ChildrenOffsets) {
269       const auto It = Offsets.find(Parent.FirstChild);
270       assert(It != Offsets.end());
271       std::size_t Pos = It->second;
272       if (Parent.HasValue) {
273         Bytes[Parent.Offset] = ((Pos >> 16) & 0xFF);
274       } else {
275         Bytes[Parent.Offset] =
276             Bytes[Parent.Offset] | uint8_t((Pos >> 16) & 0xFF);
277       }
278       Bytes[Parent.Offset + 1] = ((Pos >> 8) & 0xFF);
279       Bytes[Parent.Offset + 2] = Pos & 0xFF;
280     }
281 
282     // Add some padding so that the deserialization code
283     // doesn't try to read past the enf of the array.
284     Bytes.push_back(0);
285     Bytes.push_back(0);
286     Bytes.push_back(0);
287     Bytes.push_back(0);
288     Bytes.push_back(0);
289     Bytes.push_back(0);
290 
291     return Bytes;
292   }
293 
294 private:
295   void collectKeys(Node *N, std::set<std::string> &Keys) {
296     Keys.insert(N->Name);
297     for (const std::unique_ptr<Node> &Child : N->Children) {
298       collectKeys(Child.get(), Keys);
299     }
300   }
301 
302   // Merge sequences of 1-character nodes
303   // This greatly reduce the total number of nodes,
304   // and therefore the size of the index.
305   // When the tree gets serialized, we only have 5 bytes to store the
306   // size of a name. Overlong names (>32 characters) are therefore
307   // kep into separate nodes
308   void compact(Node *N) {
309     for (auto &&Child : N->Children) {
310       compact(Child.get());
311     }
312     if (N->Parent && N->Parent->Children.size() == 1 && !N->Parent->Value &&
313         (N->Parent->Name.size() + N->Name.size() <= 32)) {
314       N->Parent->Value = N->Value;
315       N->Parent->Name += N->Name;
316       N->Parent->Children = std::move(N->Children);
317       for (std::unique_ptr<Node> &c : N->Parent->Children) {
318         c->Parent = N->Parent;
319       }
320     }
321   }
322   struct Node {
323     Node(std::string Name, Node *Parent = nullptr)
324         : Name(Name), Parent(Parent) {}
325 
326     std::vector<std::unique_ptr<Node>> Children;
327     std::string Name;
328     Node *Parent = nullptr;
329     llvm::Optional<char32_t> Value;
330   };
331 
332   std::unique_ptr<Node> Root = std::make_unique<Node>("");
333 };
334 
335 extern const char *UnicodeLicense;
336 
337 int main(int argc, char **argv) {
338   printf("Unicode name -> codepoint mapping generator\n"
339          "Usage: %s UnicodeData.txt NameAliases.txt output\n\n",
340          argv[0]);
341   printf("NameAliases.txt can be found at "
342          "https://unicode.org/Public/14.0.0/ucd/NameAliases.txt\n"
343          "UnicodeData.txt can be found at "
344          "https://unicode.org/Public/14.0.0/ucd/UnicodeData.txt\n\n");
345 
346   if (argc != 4)
347     return EXIT_FAILURE;
348 
349   FILE *Out = fopen(argv[3], "w");
350   if (!Out) {
351     printf("Error creating output file.\n");
352     return EXIT_FAILURE;
353   }
354 
355   Trie T;
356   uint32_t NameCount = 0;
357   std::size_t LongestName = 0;
358   auto Entries = loadDataFiles(argv[1], argv[2]);
359   for (const std::pair<const char32_t, std::string> &Entry : Entries) {
360     char32_t Codepoint = Entry.first;
361     const std::string &Name = Entry.second;
362     // Ignore names which are not valid.
363     if (Name.empty() || !std::all_of(Name.begin(), Name.end(), [](char C) {
364           return llvm::is_contained(Letters, C);
365         })) {
366       continue;
367     }
368     printf("%06x: %s\n", static_cast<unsigned int>(Codepoint), Name.c_str());
369     T.insert(Name, Codepoint);
370     LongestName =
371         std::max(LongestName, std::size_t(llvm::count_if(Name, llvm::isAlnum)));
372     NameCount++;
373   }
374   T.compact();
375 
376   std::pair<std::string, std::vector<uint8_t>> Data = T.serialize();
377   const std::string &Dict = Data.first;
378   const std::vector<uint8_t> &Tree = Data.second;
379 
380   fprintf(Out, R"(
381 //===------------- Support/UnicodeNameToCodepointGenerated.cpp ------------===//
382 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
383 // See https://llvm.org/LICENSE.txt for license information.
384 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
385 //
386 //===----------------------------------------------------------------------===//
387 //
388 // This file implements mapping the name of a unicode code point to its value.
389 //
390 // This file was generated using %s.
391 // Do not edit manually.
392 //
393 //===----------------------------------------------------------------------===//
394 %s
395 
396 
397 
398 #include "llvm/Support/Compiler.h"
399 #include <cstddef>
400 #include <cstdint>
401 )",
402           argv[0], UnicodeLicense);
403 
404   fprintf(Out,
405           "namespace llvm { namespace sys { namespace unicode { \n"
406           "extern const char *UnicodeNameToCodepointDict;\n"
407           "extern const uint8_t *UnicodeNameToCodepointIndex;\n"
408           "extern const std::size_t UnicodeNameToCodepointIndexSize;\n"
409           "extern const std::size_t UnicodeNameToCodepointLargestNameSize;\n");
410 
411   fprintf(Out, "const char* UnicodeNameToCodepointDict = \"%s\";\n",
412           Dict.c_str());
413 
414   fprintf(Out, "uint8_t UnicodeNameToCodepointIndex_[%zu] = {\n",
415           Tree.size() + 1);
416 
417   for (auto Byte : Tree) {
418     fprintf(Out, "0x%02x,", Byte);
419   }
420 
421   fprintf(Out, "0};");
422   fprintf(Out, "const uint8_t* UnicodeNameToCodepointIndex = "
423                "UnicodeNameToCodepointIndex_; \n");
424   fprintf(Out, "const std::size_t UnicodeNameToCodepointIndexSize = %zu;\n",
425           Tree.size() + 1);
426   fprintf(Out,
427           "const std::size_t UnicodeNameToCodepointLargestNameSize = %zu;\n",
428           LongestName);
429   fprintf(Out, "\n}}}\n");
430   fclose(Out);
431   printf("Generated %s: %u Files.\nIndex: %f kB, Dictionary: %f kB.\nDone\n\n",
432          argv[3], NameCount, Tree.size() / 1024.0, Dict.size() / 1024.0);
433 }
434 
435 const char *UnicodeLicense = R"(
436 /*
437 UNICODE, INC. LICENSE AGREEMENT - DATA FILES AND SOFTWARE
438 
439 See Terms of Use <https://www.unicode.org/copyright.html>
440 for definitions of Unicode Inc.’s Data Files and Software.
441 
442 NOTICE TO USER: Carefully read the following legal agreement.
443 BY DOWNLOADING, INSTALLING, COPYING OR OTHERWISE USING UNICODE INC.'S
444 DATA FILES ("DATA FILES"), AND/OR SOFTWARE ("SOFTWARE"),
445 YOU UNEQUIVOCALLY ACCEPT, AND AGREE TO BE BOUND BY, ALL OF THE
446 TERMS AND CONDITIONS OF THIS AGREEMENT.
447 IF YOU DO NOT AGREE, DO NOT DOWNLOAD, INSTALL, COPY, DISTRIBUTE OR USE
448 THE DATA FILES OR SOFTWARE.
449 
450 COPYRIGHT AND PERMISSION NOTICE
451 
452 Copyright © 1991-2022 Unicode, Inc. All rights reserved.
453 Distributed under the Terms of Use in https://www.unicode.org/copyright.html.
454 
455 Permission is hereby granted, free of charge, to any person obtaining
456 a copy of the Unicode data files and any associated documentation
457 (the "Data Files") or Unicode software and any associated documentation
458 (the "Software") to deal in the Data Files or Software
459 without restriction, including without limitation the rights to use,
460 copy, modify, merge, publish, distribute, and/or sell copies of
461 the Data Files or Software, and to permit persons to whom the Data Files
462 or Software are furnished to do so, provided that either
463 (a) this copyright and permission notice appear with all copies
464 of the Data Files or Software, or
465 (b) this copyright and permission notice appear in associated
466 Documentation.
467 
468 THE DATA FILES AND SOFTWARE ARE PROVIDED "AS IS", WITHOUT WARRANTY OF
469 ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
470 WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
471 NONINFRINGEMENT OF THIRD PARTY RIGHTS.
472 IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN THIS
473 NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSEQUENTIAL
474 DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
475 DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
476 TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
477 PERFORMANCE OF THE DATA FILES OR SOFTWARE.
478 
479 Except as contained in this notice, the name of a copyright holder
480 shall not be used in advertising or otherwise to promote the sale,
481 use or other dealings in these Data Files or Software without prior
482 written authorization of the copyright holder.
483 */
484 )";
485