1/*
2 * Copyright (C) 2015 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 *      http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17#ifndef ART_CMDLINE_TOKEN_RANGE_H_
18#define ART_CMDLINE_TOKEN_RANGE_H_
19
20#include <assert.h>
21#include <vector>
22#include <string>
23#include <algorithm>
24#include <memory>
25
26namespace art {
27// A range of tokens to make token matching algorithms easier.
28//
29// We try really hard to avoid copying and store only a pointer and iterators to the
30// interiors of the vector, so a typical copy constructor never ends up doing a deep copy.
31// It is up to the user to play nice and not to mutate the strings in-place.
32//
33// Tokens are only copied if a mutating operation is performed (and even then only
34// if it *actually* mutates the token).
35struct TokenRange {
36  // Short-hand for a vector of strings. A single string and a token is synonymous.
37  using TokenList = std::vector<std::string>;
38
39  // Copying-from-vector constructor.
40  explicit TokenRange(const TokenList& token_list)
41    : token_list_(new TokenList(token_list)),
42      begin_(token_list_->begin()),
43      end_(token_list_->end())
44  {}
45
46  // Copying-from-iterator constructor
47  template <typename ForwardIterator>
48  TokenRange(ForwardIterator it_begin, ForwardIterator it_end)
49    : token_list_(new TokenList(it_begin, it_end)),
50      begin_(token_list_->begin()),
51      end_(token_list_->end())
52  {}
53
54#if 0
55  // Copying-from-vector constructor.
56  TokenRange(const TokenList& token_list ATTRIBUTE_UNUSED,
57             TokenList::const_iterator it_begin,
58             TokenList::const_iterator it_end)
59    : token_list_(new TokenList(it_begin, it_end)),
60      begin_(token_list_->begin()),
61      end_(token_list_->end()) {
62    assert(it_begin >= token_list.begin());
63    assert(it_end <= token_list.end());
64  }
65#endif
66
67  // Copying from char array constructor, convertings into tokens (strings) along the way.
68  TokenRange(const char* token_list[], size_t length)
69    : token_list_(new TokenList(&token_list[0], &token_list[length])),
70      begin_(token_list_->begin()),
71      end_(token_list_->end())
72  {}
73
74  // Non-copying move-from-vector constructor. Takes over the token vector.
75  explicit TokenRange(TokenList&& token_list)
76    : token_list_(new TokenList(std::forward<TokenList>(token_list))),
77      begin_(token_list_->begin()),
78      end_(token_list_->end())
79  {}
80
81  // Non-copying constructor. Retain reference to existing list of tokens.
82  TokenRange(std::shared_ptr<TokenList> token_list,
83             TokenList::const_iterator it_begin,
84             TokenList::const_iterator it_end)
85    : token_list_(token_list),
86      begin_(it_begin),
87      end_(it_end) {
88    assert(it_begin >= token_list->begin());
89    assert(it_end <= token_list->end());
90  }
91
92  // Non-copying copy constructor.
93  TokenRange(const TokenRange&) = default;
94
95  // Non-copying move constructor.
96  TokenRange(TokenRange&&) = default;
97
98  // Non-copying constructor. Retains reference to an existing list of tokens, with offset.
99  explicit TokenRange(std::shared_ptr<TokenList> token_list)
100    : token_list_(token_list),
101      begin_(token_list_->begin()),
102      end_(token_list_->end())
103  {}
104
105  // Iterator type for begin() and end(). Guaranteed to be a RandomAccessIterator.
106  using iterator = TokenList::const_iterator;
107
108  // Iterator type for const begin() and const end(). Guaranteed to be a RandomAccessIterator.
109  using const_iterator = iterator;
110
111  // Create a token range by splitting a string. Each separator gets their own token.
112  // Since the separator are retained as tokens, it might be useful to call
113  // RemoveToken afterwards.
114  static TokenRange Split(const std::string& string, std::initializer_list<char> separators) {
115    TokenList new_token_list;
116
117    std::string tok;
118    for (auto&& c : string) {
119      for (char sep : separators) {
120        if (c == sep) {
121          // We spotted a separator character.
122          // Push back everything before the last separator as a new token.
123          // Push back the separator as a token.
124          if (!tok.empty()) {
125            new_token_list.push_back(tok);
126            tok = "";
127          }
128          new_token_list.push_back(std::string() + sep);
129        } else {
130          // Build up the token with another character.
131          tok += c;
132        }
133      }
134    }
135
136    if (!tok.empty()) {
137      new_token_list.push_back(tok);
138    }
139
140    return TokenRange(std::move(new_token_list));
141  }
142
143  // A RandomAccessIterator to the first element in this range.
144  iterator begin() const {
145    return begin_;
146  }
147
148  // A RandomAccessIterator to one past the last element in this range.
149  iterator end() const {
150    return end_;
151  }
152
153  // The size of the range, i.e. how many tokens are in it.
154  size_t Size() const {
155    return std::distance(begin_, end_);
156  }
157
158  // Are there 0 tokens in this range?
159  bool IsEmpty() const {
160    return Size() > 0;
161  }
162
163  // Look up a token by it's offset.
164  const std::string& GetToken(size_t offset) const {
165    assert(offset < Size());
166    return *(begin_ + offset);
167  }
168
169  // Does this token range equal the other range?
170  // Equality is defined as having both the same size, and
171  // each corresponding token being equal.
172  bool operator==(const TokenRange& other) const {
173    if (this == &other) {
174      return true;
175    }
176
177    if (Size() != other.Size()) {
178      return false;
179    }
180
181    return std::equal(begin(), end(), other.begin());
182  }
183
184  // Look up the token at the requested index.
185  const std::string& operator[](int index) const {
186    assert(index >= 0 && static_cast<size_t>(index) < Size());
187    return *(begin() + index);
188  }
189
190  // Does this current range start with the other range?
191  bool StartsWith(const TokenRange& other) const {
192    if (this == &other) {
193      return true;
194    }
195
196    if (Size() < other.Size()) {
197      return false;
198    }
199
200    auto& smaller = Size() < other.Size() ? *this : other;
201    auto& greater = Size() < other.Size() ? other : *this;
202
203    return std::equal(smaller.begin(), smaller.end(), greater.begin());
204  }
205
206  // Remove all characters 'c' from each token, potentially copying the underlying tokens.
207  TokenRange RemoveCharacter(char c) const {
208    TokenList new_token_list(begin(), end());
209
210    bool changed = false;
211    for (auto&& token : new_token_list) {
212      auto it = std::remove_if(token.begin(), token.end(), [&](char ch) {
213        if (ch == c) {
214          changed = true;
215          return true;
216        }
217        return false;
218      });
219      token.erase(it, token.end());
220    }
221
222    if (!changed) {
223      return *this;
224    }
225
226    return TokenRange(std::move(new_token_list));
227  }
228
229  // Remove all tokens matching this one, potentially copying the underlying tokens.
230  TokenRange RemoveToken(const std::string& token) {
231    return RemoveIf([&](const std::string& tok) { return tok == token; });
232  }
233
234  // Discard all empty tokens, potentially copying the underlying tokens.
235  TokenRange DiscardEmpty() const {
236    return RemoveIf([](const std::string& token) { return token.empty(); });
237  }
238
239  // Create a non-copying subset of this range.
240  // Length is trimmed so that the Slice does not go out of range.
241  TokenRange Slice(size_t offset, size_t length = std::string::npos) const {
242    assert(offset < Size());
243
244    if (length != std::string::npos && offset + length > Size()) {
245      length = Size() - offset;
246    }
247
248    iterator it_end;
249    if (length == std::string::npos) {
250      it_end = end();
251    } else {
252      it_end = begin() + offset + length;
253    }
254
255    return TokenRange(token_list_, begin() + offset, it_end);
256  }
257
258  // Try to match the string with tokens from this range.
259  // Each token is used to match exactly once (after which the next token is used, and so on).
260  // The matching happens from left-to-right in a non-greedy fashion.
261  // If the currently-matched token is the wildcard, then the new outputted token will
262  // contain as much as possible until the next token is matched.
263  //
264  // For example, if this == ["a:", "_", "b:] and "_" is the match string, then
265  // MatchSubstrings on "a:foob:" will yield: ["a:", "foo", "b:"]
266  //
267  // Since the string matching can fail (e.g. ["foo"] against "bar"), then this
268  // function can fail, in which cause it will return null.
269  std::unique_ptr<TokenRange> MatchSubstrings(const std::string& string,
270                                              const std::string& wildcard) const {
271    TokenList new_token_list;
272
273    size_t wildcard_idx = std::string::npos;
274    size_t string_idx = 0;
275
276    // Function to push all the characters matched as a wildcard so far
277    // as a brand new token. It resets the wildcard matching.
278    // Empty wildcards are possible and ok, but only if wildcard matching was on.
279    auto maybe_push_wildcard_token = [&]() {
280      if (wildcard_idx != std::string::npos) {
281        size_t wildcard_length = string_idx - wildcard_idx;
282        std::string wildcard_substr = string.substr(wildcard_idx, wildcard_length);
283        new_token_list.push_back(std::move(wildcard_substr));
284
285        wildcard_idx = std::string::npos;
286      }
287    };
288
289    for (iterator it = begin(); it != end(); ++it) {
290      const std::string& tok = *it;
291
292      if (tok == wildcard) {
293        maybe_push_wildcard_token();
294        wildcard_idx = string_idx;
295        continue;
296      }
297
298      size_t next_token_idx = string.find(tok);
299      if (next_token_idx == std::string::npos) {
300        // Could not find token at all
301        return nullptr;
302      } else if (next_token_idx != string_idx && wildcard_idx == std::string::npos) {
303        // Found the token at a non-starting location, and we weren't
304        // trying to parse the wildcard.
305        return nullptr;
306      }
307
308      new_token_list.push_back(string.substr(next_token_idx, tok.size()));
309      maybe_push_wildcard_token();
310      string_idx += tok.size();
311    }
312
313    size_t remaining = string.size() - string_idx;
314    if (remaining > 0) {
315      if (wildcard_idx == std::string::npos) {
316        // Some characters were still remaining in the string,
317        // but it wasn't trying to match a wildcard.
318        return nullptr;
319      }
320    }
321
322    // If some characters are remaining, the rest must be a wildcard.
323    string_idx += remaining;
324    maybe_push_wildcard_token();
325
326    return std::unique_ptr<TokenRange>(new TokenRange(std::move(new_token_list)));
327  }
328
329  // Do a quick match token-by-token, and see if they match.
330  // Any tokens with a wildcard in them are only matched up until the wildcard.
331  // If this is true, then the wildcard matching later on can still fail, so this is not
332  // a guarantee that the argument is correct, it's more of a strong hint that the
333  // user-provided input *probably* was trying to match this argument.
334  //
335  // Returns how many tokens were either matched (or ignored because there was a
336  // wildcard present). 0 means no match. If the size() tokens are returned.
337  size_t MaybeMatches(const TokenRange& token_list, const std::string& wildcard) const {
338    auto token_it = token_list.begin();
339    auto token_end = token_list.end();
340    auto name_it = begin();
341    auto name_end = end();
342
343    size_t matched_tokens = 0;
344
345    while (token_it != token_end && name_it != name_end) {
346      // Skip token matching when the corresponding name has a wildcard in it.
347      const std::string& name = *name_it;
348
349      size_t wildcard_idx = name.find(wildcard);
350      if (wildcard_idx == std::string::npos) {  // No wildcard present
351        // Did the definition token match the user token?
352        if (name != *token_it) {
353          return matched_tokens;
354        }
355      } else {
356        std::string name_prefix = name.substr(0, wildcard_idx);
357
358        // Did the user token start with the up-to-the-wildcard prefix?
359        if (!StartsWith(*token_it, name_prefix)) {
360          return matched_tokens;
361        }
362      }
363
364      ++token_it;
365      ++name_it;
366      ++matched_tokens;
367    }
368
369    // If we got this far, it's either a full match or the token list was too short.
370    return matched_tokens;
371  }
372
373  // Flatten the token range by joining every adjacent token with the separator character.
374  // e.g. ["hello", "world"].join('$') == "hello$world"
375  std::string Join(char separator) const {
376    TokenList tmp(begin(), end());
377    return art::Join(tmp, separator);
378    // TODO: Join should probably take an offset or iterators
379  }
380
381 private:
382  static bool StartsWith(const std::string& larger, const std::string& smaller) {
383    if (larger.size() >= smaller.size()) {
384      return std::equal(smaller.begin(), smaller.end(), larger.begin());
385    }
386
387    return false;
388  }
389
390  template <typename TPredicate>
391  TokenRange RemoveIf(const TPredicate& predicate) const {
392    // If any of the tokens in the token lists are empty, then
393    // we need to remove them and compress the token list into a smaller one.
394    bool remove = false;
395    for (auto it = begin_; it != end_; ++it) {
396      auto&& token = *it;
397
398      if (predicate(token)) {
399        remove = true;
400        break;
401      }
402    }
403
404    // Actually copy the token list and remove the tokens that don't match our predicate.
405    if (remove) {
406      auto token_list = std::make_shared<TokenList>(begin(), end());
407      TokenList::iterator new_end =
408          std::remove_if(token_list->begin(), token_list->end(), predicate);
409      token_list->erase(new_end, token_list->end());
410
411      assert(token_list_->size() > token_list->size() && "Nothing was actually removed!");
412
413      return TokenRange(token_list);
414    }
415
416    return *this;
417  }
418
419  const std::shared_ptr<std::vector<std::string>> token_list_;
420  const iterator begin_;
421  const iterator end_;
422};
423}  // namespace art
424
425#endif  // ART_CMDLINE_TOKEN_RANGE_H_
426