1
|
|
|
<?php |
2
|
|
|
|
3
|
|
|
declare(strict_types=1); |
4
|
|
|
|
5
|
|
|
namespace Antlr\Antlr4\Runtime\Atn; |
6
|
|
|
|
7
|
|
|
use Antlr\Antlr4\Runtime\Atn\SemanticContexts\SemanticContext; |
8
|
|
|
use Antlr\Antlr4\Runtime\Atn\States\ATNState; |
9
|
|
|
use Antlr\Antlr4\Runtime\Atn\States\BlockEndState; |
10
|
|
|
use Antlr\Antlr4\Runtime\Atn\States\BlockStartState; |
11
|
|
|
use Antlr\Antlr4\Runtime\Atn\States\DecisionState; |
12
|
|
|
use Antlr\Antlr4\Runtime\Atn\States\RuleStopState; |
13
|
|
|
use Antlr\Antlr4\Runtime\Atn\States\StarLoopEntryState; |
14
|
|
|
use Antlr\Antlr4\Runtime\Atn\Transitions\ActionTransition; |
15
|
|
|
use Antlr\Antlr4\Runtime\Atn\Transitions\EpsilonTransition; |
16
|
|
|
use Antlr\Antlr4\Runtime\Atn\Transitions\PrecedencePredicateTransition; |
17
|
|
|
use Antlr\Antlr4\Runtime\Atn\Transitions\PredicateTransition; |
18
|
|
|
use Antlr\Antlr4\Runtime\Atn\Transitions\RuleTransition; |
19
|
|
|
use Antlr\Antlr4\Runtime\Atn\Transitions\Transition; |
20
|
|
|
use Antlr\Antlr4\Runtime\Dfa\DFA; |
21
|
|
|
use Antlr\Antlr4\Runtime\Dfa\DFAState; |
22
|
|
|
use Antlr\Antlr4\Runtime\Dfa\PredPrediction; |
23
|
|
|
use Antlr\Antlr4\Runtime\Error\Exceptions\NoViableAltException; |
24
|
|
|
use Antlr\Antlr4\Runtime\Interval; |
25
|
|
|
use Antlr\Antlr4\Runtime\IntervalSet; |
26
|
|
|
use Antlr\Antlr4\Runtime\IntStream; |
27
|
|
|
use Antlr\Antlr4\Runtime\Parser; |
28
|
|
|
use Antlr\Antlr4\Runtime\ParserRuleContext; |
29
|
|
|
use Antlr\Antlr4\Runtime\PredictionContexts\PredictionContext; |
30
|
|
|
use Antlr\Antlr4\Runtime\PredictionContexts\PredictionContextCache; |
31
|
|
|
use Antlr\Antlr4\Runtime\PredictionContexts\SingletonPredictionContext; |
32
|
|
|
use Antlr\Antlr4\Runtime\RuleContext; |
33
|
|
|
use Antlr\Antlr4\Runtime\Token; |
34
|
|
|
use Antlr\Antlr4\Runtime\TokenStream; |
35
|
|
|
use Antlr\Antlr4\Runtime\Utils\BitSet; |
36
|
|
|
use Antlr\Antlr4\Runtime\Utils\DoubleKeyMap; |
37
|
|
|
use Antlr\Antlr4\Runtime\Utils\Set; |
38
|
|
|
use Antlr\Antlr4\Runtime\VocabularyImpl; |
39
|
|
|
|
40
|
|
|
/** |
41
|
|
|
* The embodiment of the adaptive LL(*), ALL(*), parsing strategy. |
42
|
|
|
* |
43
|
|
|
* The basic complexity of the adaptive strategy makes it harder to understand. |
44
|
|
|
* We begin with ATN simulation to build paths in a DFA. Subsequent prediction |
45
|
|
|
* requests go through the DFA first. If they reach a state without an edge for |
46
|
|
|
* the current symbol, the algorithm fails over to the ATN simulation to |
47
|
|
|
* complete the DFA path for the current input (until it finds a conflict state |
48
|
|
|
* or uniquely predicting state). |
49
|
|
|
* |
50
|
|
|
* All of that is done without using the outer context because we want to create |
51
|
|
|
* a DFA that is not dependent upon the rule invocation stack when we do a |
52
|
|
|
* prediction. One DFA works in all contexts. We avoid using context not |
53
|
|
|
* necessarily because it's slower, although it can be, but because of the DFA |
54
|
|
|
* caching problem. The closure routine only considers the rule invocation stack |
55
|
|
|
* created during prediction beginning in the decision rule. For example, if |
56
|
|
|
* prediction occurs without invoking another rule's ATN, there are no context |
57
|
|
|
* stacks in the configurations. When lack of context leads to a conflict, we |
58
|
|
|
* don't know if it's an ambiguity or a weakness in the strong LL(*) parsing |
59
|
|
|
* strategy (versus full LL(*)). |
60
|
|
|
* |
61
|
|
|
* When SLL yields a configuration set with conflict, we rewind the input and |
62
|
|
|
* retry the ATN simulation, this time using full outer context without adding |
63
|
|
|
* to the DFA. Configuration context stacks will be the full invocation stacks |
64
|
|
|
* from the start rule. If we get a conflict using full context, then we can |
65
|
|
|
* definitively say we have a true ambiguity for that input sequence. If we |
66
|
|
|
* don't get a conflict, it implies that the decision is sensitive to the outer |
67
|
|
|
* context. (It is not context-sensitive in the sense of context-sensitive |
68
|
|
|
* grammars.) |
69
|
|
|
* |
70
|
|
|
* The next time we reach this DFA state with an SLL conflict, through DFA |
71
|
|
|
* simulation, we will again retry the ATN simulation using full context mode. |
72
|
|
|
* This is slow because we can't save the results and have to "interpret" the |
73
|
|
|
* ATN each time we get that input. |
74
|
|
|
* |
75
|
|
|
* CACHING FULL CONTEXT PREDICTIONS |
76
|
|
|
* |
77
|
|
|
* We could cache results from full context to predicted alternative easily and |
78
|
|
|
* that saves a lot of time but doesn't work in presence of predicates. The set |
79
|
|
|
* of visible predicates from the ATN start state changes depending on the |
80
|
|
|
* context, because closure can fall off the end of a rule. I tried to cache |
81
|
|
|
* tuples (stack context, semantic context, predicted alt) but it was slower |
82
|
|
|
* than interpreting and much more complicated. Also required a huge amount of |
83
|
|
|
* memory. The goal is not to create the world's fastest parser anyway. I'd like |
84
|
|
|
* to keep this algorithm simple. By launching multiple threads, we can improve |
85
|
|
|
* the speed of parsing across a large number of files. |
86
|
|
|
* |
87
|
|
|
* There is no strict ordering between the amount of input used by SLL vs LL, |
88
|
|
|
* which makes it really hard to build a cache for full context. Let's say that |
89
|
|
|
* we have input A B C that leads to an SLL conflict with full context X. That |
90
|
|
|
* implies that using X we might only use A B but we could also use A B C D to |
91
|
|
|
* resolve conflict. Input A B C D could predict alternative 1 in one position |
92
|
|
|
* in the input and A B C E could predict alternative 2 in another position in |
93
|
|
|
* input. The conflicting SLL configurations could still be non-unique in the |
94
|
|
|
* full context prediction, which would lead us to requiring more input than the |
95
|
|
|
* original A B C. To make a prediction cache work, we have to track the exact |
96
|
|
|
* input used during the previous prediction. That amounts to a cache that maps |
97
|
|
|
* X to a specific DFA for that context. |
98
|
|
|
* |
99
|
|
|
* Something should be done for left-recursive expression predictions. They are |
100
|
|
|
* likely LL(1) + pred eval. Easier to do the whole SLL unless error and retry |
101
|
|
|
* with full LL thing Sam does. |
102
|
|
|
* |
103
|
|
|
* AVOIDING FULL CONTEXT PREDICTION |
104
|
|
|
* |
105
|
|
|
* We avoid doing full context retry when the outer context is empty, we did not |
106
|
|
|
* dip into the outer context by falling off the end of the decision state rule, |
107
|
|
|
* or when we force SLL mode. |
108
|
|
|
* |
109
|
|
|
* As an example of the not dip into outer context case, consider as super |
110
|
|
|
* constructor calls versus function calls. One grammar might look like |
111
|
|
|
* this: |
112
|
|
|
* |
113
|
|
|
* ctorBody |
114
|
|
|
* : '{' superCall? stat* '}' |
115
|
|
|
* ; |
116
|
|
|
* |
117
|
|
|
* |
118
|
|
|
* Or, you might see something like |
119
|
|
|
* |
120
|
|
|
* stat |
121
|
|
|
* : superCall ';' |
122
|
|
|
* | expression ';' |
123
|
|
|
* | ... |
124
|
|
|
* ; |
125
|
|
|
* |
126
|
|
|
* |
127
|
|
|
* In both cases I believe that no closure operations will dip into the outer |
128
|
|
|
* context. In the first case ctorBody in the worst case will stop at the '}'. |
129
|
|
|
* In the 2nd case it should stop at the ';'. Both cases should stay within the |
130
|
|
|
* entry rule and not dip into the outer context. |
131
|
|
|
* |
132
|
|
|
* PREDICATES |
133
|
|
|
* |
134
|
|
|
* Predicates are always evaluated if present in either SLL or LL both. SLL and |
135
|
|
|
* LL simulation deals with predicates differently. SLL collects predicates as |
136
|
|
|
* it performs closure operations like ANTLR v3 did. It delays predicate |
137
|
|
|
* evaluation until it reaches and accept state. This allows us to cache the SLL |
138
|
|
|
* ATN simulation whereas, if we had evaluated predicates on-the-fly during |
139
|
|
|
* closure, the DFA state configuration sets would be different and we couldn't |
140
|
|
|
* build up a suitable DFA. |
141
|
|
|
* |
142
|
|
|
* When building a DFA accept state during ATN simulation, we evaluate any |
143
|
|
|
* predicates and return the sole semantically valid alternative. If there is |
144
|
|
|
* more than 1 alternative, we report an ambiguity. If there are 0 alternatives, |
145
|
|
|
* we throw an exception. Alternatives without predicates act like they have |
146
|
|
|
* true predicates. The simple way to think about it is to strip away all |
147
|
|
|
* alternatives with false predicates and choose the minimum alternative that |
148
|
|
|
* remains. |
149
|
|
|
* |
150
|
|
|
* When we start in the DFA and reach an accept state that's predicated, we test |
151
|
|
|
* those and return the minimum semantically viable alternative. If no |
152
|
|
|
* alternatives are viable, we throw an exception. |
153
|
|
|
* |
154
|
|
|
* During full LL ATN simulation, closure always evaluates predicates and |
155
|
|
|
* on-the-fly. This is crucial to reducing the configuration set size during |
156
|
|
|
* closure. It hits a landmine when parsing with the Java grammar, for example, |
157
|
|
|
* without this on-the-fly evaluation. |
158
|
|
|
* |
159
|
|
|
* SHARING DFA |
160
|
|
|
* |
161
|
|
|
* All instances of the same parser share the same decision DFAs through a |
162
|
|
|
* static field. Each instance gets its own ATN simulator but they share the |
163
|
|
|
* same {@see ParserATNSimulator::$decisionToDFA} field. They also share a |
164
|
|
|
* {@see PredictionContextCache} object that makes sure that all |
165
|
|
|
* {@see PredictionContext} objects are shared among the DFA states. This makes |
166
|
|
|
* a big size difference. |
167
|
|
|
* |
168
|
|
|
* THREAD SAFETY |
169
|
|
|
* |
170
|
|
|
* The {@see ParserATNSimulator} locks on the {@see ParserATNSimulator::$decisionToDFA} |
171
|
|
|
* field when it adds a new DFA object to that array. |
172
|
|
|
* {@see ParserATNSimulator::$addDFAEdge} locks on the DFA for the current |
173
|
|
|
* decision when setting the {@see DFAState::$edges} field. |
174
|
|
|
* {@see ParserATNSimulator::addDFAState()} locks on the DFA for the current |
175
|
|
|
* decision when looking up a DFA state to see if it already exists. We must |
176
|
|
|
* make sure that all requests to add DFA states that are equivalent result in |
177
|
|
|
* the same shared DFA object. This is because lots of threads will be trying |
178
|
|
|
* to update the DFA at once. {@see ParserATNSimulator::addDFAState()} also |
179
|
|
|
* locks inside the DFA lock but this time on the shared context cache when it |
180
|
|
|
* rebuilds the configurations' {@see PredictionContext} objects using cached |
181
|
|
|
* subgraphs/nodes. No other locking occurs, even during DFA simulation. This is |
182
|
|
|
* safe as long as we can guarantee that all threads referencing |
183
|
|
|
* `s.edge[t]` get the same physical target {@see DFAState}, or `null`. Once |
184
|
|
|
* into the DFA, the DFA simulation does not reference the {@see DFA::$states} map. |
185
|
|
|
* It follows the {@see DFAState::$edges} field to new targets. The DFA simulator |
186
|
|
|
* will either find {@see DFAState::$edges} to be `null`, to be non-`null` and |
187
|
|
|
* `dfa.edges[t]` null, or `dfa.edges[t]` to be non-null. The |
188
|
|
|
* {@see ParserATNSimulator::addDFAEdge()} method could be racing to set the field |
189
|
|
|
* but in either case the DFA simulator works; if `null`, and requests ATN |
190
|
|
|
* simulation. It could also race trying to get `dfa.edges[t]`, but either |
191
|
|
|
* way it will work because it's not doing a test and set operation. |
192
|
|
|
* |
193
|
|
|
* tarting with SLL then failing to combined SLL/LL (Two-Stage Parsing) |
194
|
|
|
* |
195
|
|
|
* Sam pointed out that if SLL does not give a syntax error, then there is no |
196
|
|
|
* point in doing full LL, which is slower. We only have to try LL if we get a |
197
|
|
|
* syntax error. For maximum speed, Sam starts the parser set to pure SLL |
198
|
|
|
* mode with the {@see BailErrorStrategy}: |
199
|
|
|
* |
200
|
|
|
* parser.{@see Parser::getInterpreter()}. |
201
|
|
|
* {@see ParserATNSimulator::setPredictionMode()}`(}{@see PredictionMode::$SLL})`; |
202
|
|
|
* parser->{@see Parser::setErrorHandler()}(new {@see BailErrorStrategy}()); |
203
|
|
|
* |
204
|
|
|
* If it does not get a syntax error, then we're done. If it does get a syntax |
205
|
|
|
* error, we need to retry with the combined SLL/LL strategy. |
206
|
|
|
* |
207
|
|
|
* The reason this works is as follows. If there are no SLL conflicts, then the |
208
|
|
|
* grammar is SLL (at least for that input set). If there is an SLL conflict, |
209
|
|
|
* the full LL analysis must yield a set of viable alternatives which is a |
210
|
|
|
* subset of the alternatives reported by SLL. If the LL set is a singleton, |
211
|
|
|
* then the grammar is LL but not SLL. If the LL set is the same size as the SLL |
212
|
|
|
* set, the decision is SLL. If the LL set has size > 1, then that decision |
213
|
|
|
* is truly ambiguous on the current input. If the LL set is smaller, then the |
214
|
|
|
* SLL conflict resolution might choose an alternative that the full LL would |
215
|
|
|
* rule out as a possibility based upon better context information. If that's |
216
|
|
|
* the case, then the SLL parse will definitely get an error because the full LL |
217
|
|
|
* analysis says it's not viable. If SLL conflict resolution chooses an |
218
|
|
|
* alternative within the LL set, them both SLL and LL would choose the same |
219
|
|
|
* alternative because they both choose the minimum of multiple conflicting |
220
|
|
|
* alternatives. |
221
|
|
|
* |
222
|
|
|
* Let's say we have a set of SLL conflicting alternatives `{1, 2, 3}` and |
223
|
|
|
* a smaller LL set called s. If s is `{2, 3}`, then SLL parsing will |
224
|
|
|
* get an error because SLL will pursue alternative 1. If s is `{1, 2}` or |
225
|
|
|
* `{1, 3}` then both SLL and LL will choose the same alternative because |
226
|
|
|
* alternative one is the minimum of either set. If s is `{2}` or `{3}` then |
227
|
|
|
* SLL will get a syntax error. If s is `{1}` then SLL will succeed. |
228
|
|
|
* |
229
|
|
|
* Of course, if the input is invalid, then we will get an error for sure in |
230
|
|
|
* both SLL and LL parsing. Erroneous input will therefore require 2 passes over |
231
|
|
|
* the input. |
232
|
|
|
*/ |
233
|
|
|
final class ParserATNSimulator extends ATNSimulator |
234
|
|
|
{ |
235
|
|
|
/** @var bool */ |
236
|
|
|
public static $debug = false; |
237
|
|
|
|
238
|
|
|
/** @var bool */ |
239
|
|
|
public static $debug_closure = false; |
240
|
|
|
|
241
|
|
|
/** @var bool */ |
242
|
|
|
public static $debug_add = false; |
243
|
|
|
|
244
|
|
|
/** @var bool */ |
245
|
|
|
public static $debug_list_atn_decisions = false; |
246
|
|
|
|
247
|
|
|
/** @var bool */ |
248
|
|
|
public static $dfa_debug = false; |
249
|
|
|
|
250
|
|
|
/** @var bool */ |
251
|
|
|
public static $retry_debug = false; |
252
|
|
|
|
253
|
|
|
/** @var array<string> */ |
254
|
|
|
public $log = []; |
255
|
|
|
|
256
|
|
|
/** @var Parser */ |
257
|
|
|
protected $parser; |
258
|
|
|
|
259
|
|
|
/** @var array<DFA> */ |
260
|
|
|
public $decisionToDFA = []; |
261
|
|
|
|
262
|
|
|
/** @var int */ |
263
|
|
|
private $mode = PredictionMode::LL; |
264
|
|
|
|
265
|
|
|
/** |
266
|
|
|
* Each prediction operation uses a cache for merge of prediction contexts. |
267
|
|
|
* Don't keep around as it wastes huge amounts of memory. DoubleKeyMap |
268
|
|
|
* isn't synchronized but we're ok since two threads shouldn't reuse same |
269
|
|
|
* parser/atnsim object because it can only handle one input at a time. |
270
|
|
|
* This maps graphs a and b to merged result c. (a,b)→c. We can avoid |
271
|
|
|
* the merge if we ever see a and b again. Note that (b,a)→c should |
272
|
|
|
* also be examined during cache lookup. |
273
|
|
|
* |
274
|
|
|
* @var DoubleKeyMap|null |
275
|
|
|
*/ |
276
|
|
|
protected $mergeCache; |
277
|
|
|
|
278
|
|
|
/** |
279
|
|
|
* LAME globals to avoid parameters!!!!! I need these down deep in predTransition. |
280
|
|
|
* |
281
|
|
|
* @var TokenStream |
282
|
|
|
*/ |
283
|
|
|
protected $input; |
284
|
|
|
|
285
|
|
|
/** @var int */ |
286
|
|
|
protected $startIndex = 0; |
287
|
|
|
|
288
|
|
|
/** @var ParserRuleContext|null */ |
289
|
|
|
protected $outerContext; |
290
|
|
|
|
291
|
|
|
/** @var DFA|null */ |
292
|
|
|
protected $dfa; |
293
|
|
|
|
294
|
|
|
/** |
295
|
|
|
* @param array<DFA> $decisionToDFA |
296
|
|
|
*/ |
297
|
7 |
|
public function __construct( |
298
|
|
|
Parser $parser, |
299
|
|
|
ATN $atn, |
300
|
|
|
array $decisionToDFA, |
301
|
|
|
PredictionContextCache $sharedContextCache |
302
|
|
|
) { |
303
|
7 |
|
parent::__construct($atn, $sharedContextCache); |
304
|
|
|
|
305
|
7 |
|
$this->parser = $parser; |
306
|
7 |
|
$this->decisionToDFA = $decisionToDFA; |
307
|
7 |
|
} |
308
|
|
|
|
309
|
|
|
public function reset() : void |
310
|
|
|
{ |
311
|
|
|
} |
312
|
|
|
|
313
|
|
|
public function clearDFA() : void |
314
|
|
|
{ |
315
|
|
|
for ($d = 0, $count = \count($this->decisionToDFA); $d < $count; $d++) { |
316
|
|
|
$decisionState = $this->atn->getDecisionState($d); |
317
|
|
|
|
318
|
|
|
if ($decisionState !== null) { |
319
|
|
|
$this->decisionToDFA[$d] = new DFA($decisionState, $d); |
320
|
|
|
} |
321
|
|
|
} |
322
|
|
|
} |
323
|
|
|
|
324
|
4 |
|
public function adaptivePredict(TokenStream $input, int $decision, ParserRuleContext $outerContext) : int |
325
|
|
|
{ |
326
|
4 |
|
if (self::$debug || self::$debug_list_atn_decisions) { |
327
|
|
|
$token = $input->LT(1); |
328
|
|
|
|
329
|
|
|
$this->log[] = \sprintf( |
330
|
|
|
'adaptivePredict decision %d exec LA(1)==%s line %d:%d', |
331
|
|
|
$decision, |
332
|
|
|
$this->getLookaheadName($input), |
333
|
|
|
$token === null ? '' : $token->getLine(), |
334
|
|
|
$token === null ? '' : $token->getCharPositionInLine() |
335
|
|
|
); |
336
|
|
|
} |
337
|
|
|
|
338
|
4 |
|
$this->input = $input; |
339
|
4 |
|
$this->startIndex = $input->getIndex(); |
340
|
4 |
|
$this->outerContext = $outerContext; |
341
|
|
|
|
342
|
4 |
|
$dfa = $this->decisionToDFA[$decision]; |
343
|
4 |
|
$this->dfa = $dfa; |
344
|
|
|
|
345
|
4 |
|
$m = $input->mark(); |
346
|
4 |
|
$index = $this->startIndex; |
347
|
|
|
|
348
|
|
|
// Now we are certain to have a specific decision's DFA, but do we still need an initial state? |
349
|
|
|
try { |
350
|
4 |
|
if ($dfa->isPrecedenceDfa()) { |
351
|
|
|
// The start state for a precedence DFA depends on the current |
352
|
|
|
// parser precedence, and is provided by a DFA method. |
353
|
|
|
|
354
|
4 |
|
$s0 = $dfa->getPrecedenceStartState($this->parser->getPrecedence()); |
355
|
|
|
} else { |
356
|
|
|
// The start state for a "regular" DFA is just s0. |
357
|
4 |
|
$s0 = $dfa->s0; |
358
|
|
|
} |
359
|
|
|
|
360
|
4 |
|
if ($s0 === null) { |
361
|
1 |
|
if (self::$debug || self::$debug_list_atn_decisions) { |
362
|
|
|
$this->log[] = \sprintf( |
363
|
|
|
'predictATN decision %d exec LA(1)==%s, outerContext=%s', |
364
|
|
|
$dfa->decision, |
365
|
|
|
$this->getLookaheadName($input), |
366
|
|
|
$outerContext->toString($this->parser->getRuleNames()) |
367
|
|
|
); |
368
|
|
|
} |
369
|
|
|
|
370
|
1 |
|
$fullCtx = false; |
371
|
|
|
|
372
|
1 |
|
if ($dfa->atnStartState === null) { |
373
|
|
|
throw new \RuntimeException('ATN Start State cannot be null.'); |
374
|
|
|
} |
375
|
|
|
|
376
|
1 |
|
$s0_closure = $this->computeStartState( |
377
|
1 |
|
$dfa->atnStartState, |
378
|
1 |
|
ParserRuleContext::emptyContext(), |
379
|
|
|
$fullCtx |
380
|
|
|
); |
381
|
|
|
|
382
|
1 |
|
if ($dfa->isPrecedenceDfa()) { |
383
|
|
|
/* |
384
|
|
|
* If this is a precedence DFA, we use applyPrecedenceFilter |
385
|
|
|
* to convert the computed start state to a precedence start |
386
|
|
|
* state. We then use DFA.setPrecedenceStartState to set the |
387
|
|
|
* appropriate start state for the precedence level rather |
388
|
|
|
* than simply setting DFA.s0. |
389
|
|
|
*/ |
390
|
|
|
|
391
|
1 |
|
if ($dfa->s0 === null) { |
392
|
|
|
throw new \RuntimeException('DFA.s0 cannot be null.'); |
393
|
|
|
} |
394
|
|
|
|
395
|
1 |
|
$dfa->s0->configs = $s0_closure; // not used for prediction but useful to know start configs anyway |
396
|
|
|
|
397
|
1 |
|
$s0_closure = $this->applyPrecedenceFilter($s0_closure); |
398
|
|
|
|
399
|
1 |
|
$s0 = $this->addDFAState($dfa, new DFAState($s0_closure)); |
400
|
|
|
|
401
|
1 |
|
$dfa->setPrecedenceStartState($this->parser->getPrecedence(), $s0); |
402
|
|
|
} else { |
403
|
1 |
|
$s0 = $this->addDFAState($dfa, new DFAState($s0_closure)); |
404
|
1 |
|
$dfa->s0 = $s0; |
405
|
|
|
} |
406
|
|
|
} |
407
|
|
|
|
408
|
4 |
|
$alt = $this->execATN($dfa, $s0, $input, $index, $outerContext); |
409
|
|
|
|
410
|
4 |
|
if (self::$debug) { |
411
|
|
|
$this->log[] = \sprintf('DFA after predictATN: %s', $dfa->toString($this->parser->getVocabulary())); |
412
|
|
|
} |
413
|
|
|
|
414
|
4 |
|
return $alt ?? 0; |
415
|
|
|
} finally { |
416
|
4 |
|
$this->mergeCache = null; // wack cache after each prediction |
417
|
4 |
|
$this->dfa = null; |
418
|
4 |
|
$input->seek($index); |
419
|
4 |
|
$input->release($m); |
420
|
|
|
} |
421
|
|
|
} |
422
|
|
|
|
423
|
|
|
/** |
424
|
|
|
* Performs ATN simulation to compute a predicted alternative based |
425
|
|
|
* upon the remaining input, but also updates the DFA cache to avoid |
426
|
|
|
* having to traverse the ATN again for the same input sequence. |
427
|
|
|
* |
428
|
|
|
* There are some key conditions we're looking for after computing a new |
429
|
|
|
* set of ATN configs (proposed DFA state): |
430
|
|
|
* if the set is empty, there is no viable alternative for current symbol |
431
|
|
|
* does the state uniquely predict an alternative? |
432
|
|
|
* does the state have a conflict that would prevent us from |
433
|
|
|
* putting it on the work list? |
434
|
|
|
* |
435
|
|
|
* We also have some key operations to do: |
436
|
|
|
* - add an edge from previous DFA state to potentially new DFA state, D, |
437
|
|
|
* upon current symbol but only if adding to work list, which means in all |
438
|
|
|
* cases except no viable alternative (and possibly non-greedy decisions?) |
439
|
|
|
* - collecting predicates and adding semantic context to DFA accept states |
440
|
|
|
* - adding rule context to context-sensitive DFA accept states |
441
|
|
|
* - consuming an input symbol |
442
|
|
|
* - reporting a conflict |
443
|
|
|
* - reporting an ambiguity |
444
|
|
|
* - reporting a context sensitivity |
445
|
|
|
* - reporting insufficient predicates |
446
|
|
|
* |
447
|
|
|
* cover these cases: |
448
|
|
|
* - dead end |
449
|
|
|
* - single alt |
450
|
|
|
* - single alt + preds |
451
|
|
|
* - conflict |
452
|
|
|
* - conflict + preds |
453
|
|
|
*/ |
454
|
4 |
|
public function execATN( |
455
|
|
|
DFA $dfa, |
456
|
|
|
DFAState $s0, |
457
|
|
|
TokenStream $input, |
458
|
|
|
int $startIndex, |
459
|
|
|
ParserRuleContext $outerContext |
460
|
|
|
) : ?int { |
461
|
4 |
|
if (self::$debug || self::$debug_list_atn_decisions) { |
462
|
|
|
$token = $input->LT(1); |
463
|
|
|
|
464
|
|
|
$this->log[] = \sprintf( |
465
|
|
|
'execATN decision %d exec LA(1)==%s line %d:%d', |
466
|
|
|
$dfa->decision, |
467
|
|
|
$this->getLookaheadName($input), |
468
|
|
|
$token === null ? '' : $token->getLine(), |
469
|
|
|
$token === null ? '' : $token->getCharPositionInLine() |
470
|
|
|
); |
471
|
|
|
} |
472
|
|
|
|
473
|
4 |
|
$previousD = $s0; |
474
|
|
|
|
475
|
4 |
|
if (self::$debug) { |
476
|
|
|
$this->log[] = 's0 = ' . $s0; |
477
|
|
|
} |
478
|
|
|
|
479
|
4 |
|
$t = $input->LA(1); |
480
|
|
|
|
481
|
4 |
|
while (true) { |
482
|
4 |
|
$D = $this->getExistingTargetState($previousD, $t); |
483
|
|
|
|
484
|
4 |
|
if ($D === null) { |
485
|
3 |
|
$D = $this->computeTargetState($dfa, $previousD, $t); |
486
|
|
|
} |
487
|
|
|
|
488
|
4 |
|
if ($D === null) { |
489
|
|
|
throw new \RuntimeException('DFA State cannot be null.'); |
490
|
|
|
} |
491
|
|
|
|
492
|
4 |
|
if ($D === self::error()) { |
493
|
|
|
/* If any configs in previous dipped into outer context, that |
494
|
|
|
* means that input up to t actually finished entry rule |
495
|
|
|
* at least for SLL decision. Full LL doesn't dip into outer |
496
|
|
|
* so don't need special case. |
497
|
|
|
* We will get an error no matter what so delay until after |
498
|
|
|
* decision; better error message. Also, no reachable target |
499
|
|
|
* ATN states in SLL implies LL will also get nowhere. |
500
|
|
|
* If conflict in states that dip out, choose min since we |
501
|
|
|
* will get error no matter what. |
502
|
|
|
*/ |
503
|
|
|
|
504
|
4 |
|
$e = $this->noViableAlt($input, $outerContext, $previousD->configs, $startIndex); |
505
|
|
|
|
506
|
4 |
|
$input->seek($startIndex); |
507
|
|
|
|
508
|
4 |
|
$alt = $this->getSynValidOrSemInvalidAltThatFinishedDecisionEntryRule( |
509
|
4 |
|
$previousD->configs, |
510
|
|
|
$outerContext |
511
|
|
|
); |
512
|
|
|
|
513
|
4 |
|
if ($alt !== ATN::INVALID_ALT_NUMBER) { |
514
|
4 |
|
return $alt; |
515
|
|
|
} |
516
|
|
|
|
517
|
|
|
throw $e; |
518
|
|
|
} |
519
|
|
|
|
520
|
4 |
|
if ($D->requiresFullContext && $this->mode !== PredictionMode::SLL) { |
521
|
|
|
// IF PREDS, MIGHT RESOLVE TO SINGLE ALT => SLL (or syntax error) |
522
|
|
|
|
523
|
|
|
$conflictingAlts = $D->configs->getConflictingAlts(); |
524
|
|
|
|
525
|
|
|
if ($D->predicates !== null) { |
526
|
|
|
if (self::$debug) { |
527
|
|
|
$this->log[] = 'DFA state has preds in DFA sim LL failover'; |
528
|
|
|
} |
529
|
|
|
|
530
|
|
|
$conflictIndex = $input->getIndex(); |
531
|
|
|
|
532
|
|
|
if ($conflictIndex !== $startIndex) { |
533
|
|
|
$input->seek($startIndex); |
534
|
|
|
} |
535
|
|
|
|
536
|
|
|
$conflictingAlts = $this->evalSemanticContextMany($D->predicates, $outerContext, true); |
537
|
|
|
|
538
|
|
|
if ($conflictingAlts->length() === 1) { |
539
|
|
|
if (self::$debug) { |
540
|
|
|
$this->log[] = 'Full LL avoided'; |
541
|
|
|
} |
542
|
|
|
|
543
|
|
|
return $conflictingAlts->minValue(); |
544
|
|
|
} |
545
|
|
|
|
546
|
|
|
if ($conflictIndex !== $startIndex) { |
547
|
|
|
// Restore the index so reporting the fallback to full |
548
|
|
|
// context occurs with the index at the correct spot |
549
|
|
|
|
550
|
|
|
$input->seek($conflictIndex); |
551
|
|
|
} |
552
|
|
|
} |
553
|
|
|
|
554
|
|
|
if (self::$dfa_debug) { |
555
|
|
|
$this->log[] = \sprintf( |
556
|
|
|
'Ctx sensitive state %s in %s', |
557
|
|
|
(string) $outerContext, |
558
|
|
|
(string) $D |
559
|
|
|
); |
560
|
|
|
} |
561
|
|
|
|
562
|
|
|
if ($dfa->atnStartState === null) { |
563
|
|
|
throw new \RuntimeException('ATN Start State cannot be null.'); |
564
|
|
|
} |
565
|
|
|
|
566
|
|
|
$s0_closure = $this->computeStartState($dfa->atnStartState, $outerContext, true); |
567
|
|
|
|
568
|
|
|
$this->reportAttemptingFullContext( |
569
|
|
|
$dfa, |
570
|
|
|
$conflictingAlts, |
571
|
|
|
$D->configs, |
572
|
|
|
$startIndex, |
573
|
|
|
$input->getIndex() |
574
|
|
|
); |
575
|
|
|
|
576
|
|
|
return $this->execATNWithFullContext($dfa, $D, $s0_closure, $input, $startIndex, $outerContext); |
577
|
|
|
} |
578
|
|
|
|
579
|
4 |
|
if ($D->isAcceptState) { |
580
|
4 |
|
if ($D->predicates === null) { |
581
|
4 |
|
return $D->prediction; |
582
|
|
|
} |
583
|
|
|
|
584
|
3 |
|
$stopIndex = $input->getIndex(); |
585
|
3 |
|
$input->seek($startIndex); |
586
|
3 |
|
$alts = $this->evalSemanticContextMany($D->predicates, $outerContext, true); |
587
|
|
|
|
588
|
3 |
|
switch ($alts->length()) { |
589
|
3 |
|
case 0: |
590
|
|
|
throw $this->noViableAlt($input, $outerContext, $D->configs, $startIndex); |
591
|
|
|
|
592
|
3 |
|
case 1: |
593
|
3 |
|
return $alts->minValue(); |
594
|
|
|
|
595
|
|
|
default: |
596
|
|
|
// Report ambiguity after predicate evaluation to make sure |
597
|
|
|
// the correct set of ambig alts is reported. |
598
|
|
|
$this->reportAmbiguity($dfa, $D, $startIndex, $stopIndex, false, $alts, $D->configs); |
599
|
|
|
|
600
|
|
|
return $alts->minValue(); |
601
|
|
|
} |
602
|
|
|
} |
603
|
|
|
|
604
|
2 |
|
$previousD = $D; |
605
|
|
|
|
606
|
2 |
|
if ($t !== IntStream::EOF) { |
607
|
2 |
|
$input->consume(); |
608
|
2 |
|
$t = $input->LA(1); |
609
|
|
|
} |
610
|
|
|
} |
611
|
|
|
} |
612
|
|
|
|
613
|
|
|
/** |
614
|
|
|
* Get an existing target state for an edge in the DFA. If the target state |
615
|
|
|
* for the edge has not yet been computed or is otherwise not available, |
616
|
|
|
* this method returns `null`. |
617
|
|
|
* |
618
|
|
|
* @param DFAState $previousD The current DFA state |
619
|
|
|
* @param int $t The next input symbol |
620
|
|
|
* |
621
|
|
|
* @return DFAState|null The existing target DFA state for the given input |
622
|
|
|
* symbol `t`, or `null` if the target state for |
623
|
|
|
* this edge is not already cached. |
624
|
|
|
*/ |
625
|
4 |
|
public function getExistingTargetState(DFAState $previousD, int $t) : ?DFAState |
626
|
|
|
{ |
627
|
4 |
|
$edges = $previousD->edges; |
628
|
|
|
|
629
|
4 |
|
if ($edges === null || $t + 1 < 0 || $t + 1 >= $edges->count()) { |
630
|
1 |
|
return null; |
631
|
|
|
} |
632
|
|
|
|
633
|
4 |
|
return $edges[$t + 1]; |
634
|
|
|
} |
635
|
|
|
|
636
|
|
|
/** |
637
|
|
|
* Compute a target state for an edge in the DFA, and attempt to add |
638
|
|
|
* the computed state and corresponding edge to the DFA. |
639
|
|
|
* |
640
|
|
|
* @param DFA $dfa The DFA |
641
|
|
|
* @param DFAState $previousD The current DFA state |
642
|
|
|
* @param int $t The next input symbol |
643
|
|
|
* |
644
|
|
|
* @return DFAState|null The computed target DFA state for the given input |
645
|
|
|
* symbol `t`. If `t` does not lead to a valid DFA |
646
|
|
|
* state, this method returns |
647
|
|
|
* {@see ParserATNSimulator::error()}. |
648
|
|
|
*/ |
649
|
3 |
|
public function computeTargetState(DFA $dfa, DFAState $previousD, int $t) : ?DFAState |
650
|
|
|
{ |
651
|
3 |
|
$reach = $this->computeReachSet($previousD->configs, $t, false); |
652
|
|
|
|
653
|
3 |
|
if ($reach === null) { |
654
|
1 |
|
$this->addDFAEdge($dfa, $previousD, $t, self::error()); |
655
|
|
|
|
656
|
1 |
|
return self::error(); |
657
|
|
|
} |
658
|
|
|
|
659
|
|
|
// Create new target state; we'll add to DFA after it's complete |
660
|
3 |
|
$D = new DFAState($reach); |
661
|
|
|
|
662
|
3 |
|
$predictedAlt = self::getUniqueAlt($reach); |
663
|
|
|
|
664
|
3 |
|
if (self::$debug) { |
665
|
|
|
$altSubSets = PredictionMode::getConflictingAltSubsets($reach); |
666
|
|
|
|
667
|
|
|
$this->log[] = \sprintf( |
668
|
|
|
'SLL altSubSets=[%s], previous=%s, configs=%s, predict=%d, allSubsetsConflict=%s, conflictingAlts=%s', |
669
|
|
|
\implode(', ', $altSubSets), |
670
|
|
|
(string) $previousD->configs, |
671
|
|
|
(string) $reach, |
672
|
|
|
$predictedAlt, |
673
|
|
|
PredictionMode::allSubsetsConflict($altSubSets), |
674
|
|
|
$this->getConflictingAlts($reach) |
675
|
|
|
); |
676
|
|
|
} |
677
|
|
|
|
678
|
3 |
|
if ($predictedAlt !== ATN::INVALID_ALT_NUMBER) { |
679
|
|
|
// NO CONFLICT, UNIQUELY PREDICTED ALT |
680
|
|
|
|
681
|
3 |
|
$D->isAcceptState = true; |
682
|
3 |
|
$D->configs->uniqueAlt = $predictedAlt; |
683
|
3 |
|
$D->prediction = $predictedAlt; |
684
|
1 |
|
} elseif (PredictionMode::hasSLLConflictTerminatingPrediction($this->mode, $reach)) { |
685
|
|
|
// MORE THAN ONE VIABLE ALTERNATIVE |
686
|
|
|
|
687
|
|
|
$D->configs->setConflictingAlts($this->getConflictingAlts($reach)); |
688
|
|
|
$D->requiresFullContext = true; |
689
|
|
|
|
690
|
|
|
// in SLL-only mode, we will stop at this state and return the minimum alt |
691
|
|
|
$D->isAcceptState = true; |
692
|
|
|
|
693
|
|
|
$conflictingAlts = $D->configs->getConflictingAlts(); |
694
|
|
|
|
695
|
|
|
if ($conflictingAlts === null) { |
696
|
|
|
throw new \RuntimeException('Unexpected null conflicting alternatives.'); |
697
|
|
|
} |
698
|
|
|
|
699
|
|
|
$D->prediction = $conflictingAlts->minValue(); |
700
|
|
|
} |
701
|
|
|
|
702
|
3 |
|
if ($D->isAcceptState && $D->configs->hasSemanticContext) { |
703
|
2 |
|
$decisionState = $this->atn->getDecisionState($dfa->decision); |
704
|
|
|
|
705
|
2 |
|
if ($decisionState !== null) { |
706
|
2 |
|
$this->predicateDFAState($D, $decisionState); |
707
|
|
|
} |
708
|
|
|
|
709
|
2 |
|
if ($D->predicates !== null) { |
710
|
2 |
|
$D->prediction = ATN::INVALID_ALT_NUMBER; |
711
|
|
|
} |
712
|
|
|
} |
713
|
|
|
|
714
|
|
|
// All adds to dfa are done after we've created full D state |
715
|
3 |
|
$D = $this->addDFAEdge($dfa, $previousD, $t, $D); |
716
|
|
|
|
717
|
3 |
|
return $D; |
718
|
|
|
} |
719
|
|
|
|
720
|
2 |
|
protected function predicateDFAState(DFAState $dfaState, DecisionState $decisionState) : void |
721
|
|
|
{ |
722
|
|
|
// We need to test all predicates, even in DFA states that uniquely predict alternative. |
723
|
2 |
|
$nalts = $decisionState->getNumberOfTransitions(); |
724
|
|
|
|
725
|
|
|
// Update DFA so reach becomes accept state with (predicate,alt) pairs |
726
|
|
|
// if preds found for conflicting alts. |
727
|
|
|
|
728
|
2 |
|
$altsToCollectPredsFrom = $this->getConflictingAltsOrUniqueAlt($dfaState->configs); |
729
|
2 |
|
$altToPred = $altsToCollectPredsFrom === null ? |
730
|
|
|
null : |
731
|
2 |
|
$this->getPredsForAmbigAlts($altsToCollectPredsFrom, $dfaState->configs, $nalts); |
|
|
|
|
732
|
|
|
|
733
|
2 |
|
if ($altToPred !== null) { |
|
|
|
|
734
|
2 |
|
$dfaState->predicates = $this->getPredicatePredictions($altsToCollectPredsFrom, $altToPred); |
735
|
2 |
|
$dfaState->prediction = ATN::INVALID_ALT_NUMBER; // make sure we use preds |
736
|
|
|
} else { |
737
|
|
|
// There are preds in configs but they might go away when |
738
|
|
|
// OR'd together like {p}? || NONE == NONE. If neither alt has preds, |
739
|
|
|
// resolve to min alt. |
740
|
|
|
|
741
|
|
|
if ($altsToCollectPredsFrom === null) { |
742
|
|
|
throw new \RuntimeException('Unexpected null alternatives to collect predicates'); |
743
|
|
|
} |
744
|
|
|
|
745
|
|
|
$dfaState->prediction = $altsToCollectPredsFrom->minValue(); |
746
|
|
|
} |
747
|
2 |
|
} |
748
|
|
|
|
749
|
|
|
/** |
750
|
|
|
* Comes back with reach.uniqueAlt set to a valid alt. |
751
|
|
|
*/ |
752
|
|
|
protected function execATNWithFullContext( |
753
|
|
|
DFA $dfa, |
754
|
|
|
DFAState $D, // how far we got before failing over |
755
|
|
|
ATNConfigSet $s0, |
756
|
|
|
TokenStream $input, |
757
|
|
|
int $startIndex, |
758
|
|
|
ParserRuleContext $outerContext |
759
|
|
|
) : int { |
760
|
|
|
if (self::$debug || self::$debug_list_atn_decisions) { |
761
|
|
|
$this->log[] = \sprintf('ExecATNWithFullContext %s', (string) $s0); |
762
|
|
|
} |
763
|
|
|
|
764
|
|
|
$fullCtx = true; |
765
|
|
|
$foundExactAmbig = false; |
766
|
|
|
$reach = null; |
767
|
|
|
$previous = $s0; |
768
|
|
|
$input->seek($startIndex); |
769
|
|
|
$t = $input->LA(1); |
770
|
|
|
$predictedAlt = 0; |
771
|
|
|
|
772
|
|
|
while (true) { // while more work |
773
|
|
|
$reach = $this->computeReachSet($previous, $t, $fullCtx); |
774
|
|
|
|
775
|
|
|
if ($reach === null) { |
776
|
|
|
/* I any configs in previous dipped into outer context, that |
777
|
|
|
* means that input up to t actually finished entry rule |
778
|
|
|
* at least for LL decision. Full LL doesn't dip into outer |
779
|
|
|
* so don't need special case. |
780
|
|
|
* We will get an error no matter what so delay until after |
781
|
|
|
* decision; better error message. Also, no reachable target |
782
|
|
|
* ATN states in SLL implies LL will also get nowhere. |
783
|
|
|
* If conflict in states that dip out, choose min since we |
784
|
|
|
* will get error no matter what. |
785
|
|
|
*/ |
786
|
|
|
|
787
|
|
|
$e = $this->noViableAlt($input, $outerContext, $previous, $startIndex); |
788
|
|
|
|
789
|
|
|
$input->seek($startIndex); |
790
|
|
|
|
791
|
|
|
$alt = $this->getSynValidOrSemInvalidAltThatFinishedDecisionEntryRule($previous, $outerContext); |
792
|
|
|
|
793
|
|
|
if ($alt !== ATN::INVALID_ALT_NUMBER) { |
794
|
|
|
return $alt; |
795
|
|
|
} |
796
|
|
|
|
797
|
|
|
throw $e; |
798
|
|
|
} |
799
|
|
|
|
800
|
|
|
$altSubSets = PredictionMode::getConflictingAltSubsets($reach); |
801
|
|
|
|
802
|
|
|
if (self::$debug) { |
803
|
|
|
$this->log[] = \sprintf( |
804
|
|
|
'LL altSubSets=%s, predict=%d, resolvesToJustOneViableAlt=%d', |
805
|
|
|
\implode(',', $altSubSets), |
806
|
|
|
PredictionMode::getUniqueAlt($altSubSets), |
807
|
|
|
PredictionMode::resolvesToJustOneViableAlt($altSubSets) |
808
|
|
|
); |
809
|
|
|
} |
810
|
|
|
|
811
|
|
|
$reach->uniqueAlt = self::getUniqueAlt($reach); |
812
|
|
|
|
813
|
|
|
// unique prediction? |
814
|
|
|
if ($reach->uniqueAlt !== ATN::INVALID_ALT_NUMBER) { |
815
|
|
|
$predictedAlt = $reach->uniqueAlt; |
816
|
|
|
|
817
|
|
|
break; |
818
|
|
|
} |
819
|
|
|
|
820
|
|
|
if ($this->mode !== PredictionMode::LL_EXACT_AMBIG_DETECTION) { |
821
|
|
|
$predictedAlt = PredictionMode::resolvesToJustOneViableAlt($altSubSets); |
822
|
|
|
|
823
|
|
|
if ($predictedAlt !== ATN::INVALID_ALT_NUMBER) { |
824
|
|
|
break; |
825
|
|
|
} |
826
|
|
|
} else { |
827
|
|
|
// In exact ambiguity mode, we never try to terminate early. |
828
|
|
|
// Just keeps scarfing until we know what the conflict is |
829
|
|
|
if (PredictionMode::allSubsetsConflict($altSubSets) && PredictionMode::allSubsetsEqual($altSubSets)) { |
830
|
|
|
$foundExactAmbig = true; |
831
|
|
|
$predictedAlt = PredictionMode::getSingleViableAlt($altSubSets); |
832
|
|
|
|
833
|
|
|
break; |
834
|
|
|
} |
835
|
|
|
|
836
|
|
|
// Else there are multiple non-conflicting subsets or |
837
|
|
|
// we're not sure what the ambiguity is yet. |
838
|
|
|
// So, keep going. |
839
|
|
|
} |
840
|
|
|
|
841
|
|
|
$previous = $reach; |
842
|
|
|
|
843
|
|
|
if ($t !== IntStream::EOF) { |
844
|
|
|
$input->consume(); |
845
|
|
|
$t = $input->LA(1); |
846
|
|
|
} |
847
|
|
|
} |
848
|
|
|
|
849
|
|
|
// If the configuration set uniquely predicts an alternative, |
850
|
|
|
// without conflict, then we know that it's a full LL decision not SLL. |
851
|
|
|
if ($reach->uniqueAlt !== ATN::INVALID_ALT_NUMBER) { |
852
|
|
|
$this->reportContextSensitivity($dfa, $predictedAlt, $reach, $startIndex, $input->getIndex()); |
|
|
|
|
853
|
|
|
|
854
|
|
|
return $predictedAlt; |
855
|
|
|
} |
856
|
|
|
|
857
|
|
|
/* We do not check predicates here because we have checked them on-the-fly |
858
|
|
|
* when doing full context prediction. |
859
|
|
|
* |
860
|
|
|
* In non-exact ambiguity detection mode, we might actually be able to |
861
|
|
|
* detect an exact ambiguity, but I'm not going to spend the cycles |
862
|
|
|
* needed to check. We only emit ambiguity warnings in exact ambiguity |
863
|
|
|
* mode. |
864
|
|
|
* |
865
|
|
|
* For example, we might know that we have conflicting configurations. |
866
|
|
|
* But, that does not mean that there is no way forward without a |
867
|
|
|
* conflict. It's possible to have nonconflicting alt subsets as in: |
868
|
|
|
* |
869
|
|
|
* altSubSets=[{1, 2}, {1, 2}, {1}, {1, 2}] |
870
|
|
|
* |
871
|
|
|
* from |
872
|
|
|
* [ |
873
|
|
|
* (17,1,[5 $]), (13,1,[5 10 $]), (21,1,[5 10 $]), (11,1,[$]), |
874
|
|
|
* (13,2,[5 10 $]), (21,2,[5 10 $]), (11,2,[$]) |
875
|
|
|
* ] |
876
|
|
|
* |
877
|
|
|
* In this case, (17,1,[5 $]) indicates there is some next sequence that |
878
|
|
|
* would resolve this without conflict to alternative 1. Any other viable |
879
|
|
|
* next sequence, however, is associated with a conflict. We stop |
880
|
|
|
* looking for input because no amount of further lookahead will alter |
881
|
|
|
* the fact that we should predict alternative 1. We just can't say for |
882
|
|
|
* sure that there is an ambiguity without looking further. |
883
|
|
|
*/ |
884
|
|
|
|
885
|
|
|
$this->reportAmbiguity($dfa, $D, $startIndex, $input->getIndex(), $foundExactAmbig, null, $reach); |
|
|
|
|
886
|
|
|
|
887
|
|
|
return $predictedAlt; |
888
|
|
|
} |
889
|
|
|
|
890
|
3 |
|
protected function computeReachSet(ATNConfigSet $closure, int $t, bool $fullCtx) : ?ATNConfigSet |
891
|
|
|
{ |
892
|
3 |
|
if (self::$debug) { |
893
|
|
|
$this->log[] = \sprintf('In computeReachSet, starting closure: %s', (string) $closure); |
894
|
|
|
} |
895
|
|
|
|
896
|
3 |
|
if ($this->mergeCache === null) { |
897
|
3 |
|
$this->mergeCache = new DoubleKeyMap(); |
898
|
|
|
} |
899
|
|
|
|
900
|
3 |
|
$intermediate = new ATNConfigSet($fullCtx); |
901
|
|
|
|
902
|
|
|
/* |
903
|
|
|
* Configurations already in a rule stop state indicate reaching the end |
904
|
|
|
* of the decision rule (local context) or end of the start rule (full |
905
|
|
|
* context). Once reached, these configurations are never updated by a |
906
|
|
|
* closure operation, so they are handled separately for the performance |
907
|
|
|
* advantage of having a smaller intermediate set when calling closure. |
908
|
|
|
* |
909
|
|
|
* For full-context reach operations, separate handling is required to |
910
|
|
|
* ensure that the alternative matching the longest overall sequence is |
911
|
|
|
* chosen when multiple such configurations can match the input. |
912
|
|
|
*/ |
913
|
3 |
|
$skippedStopStates = null; |
914
|
|
|
|
915
|
|
|
// First figure out where we can reach on input t |
916
|
3 |
|
foreach ($closure->elements() as $c) { |
917
|
3 |
|
if (self::$debug_add) { |
918
|
|
|
$this->log[] = \sprintf('Testing %s at %s', $this->getTokenName($t), (string) $c); |
919
|
|
|
} |
920
|
|
|
|
921
|
3 |
|
if ($c->state instanceof RuleStopState) { |
922
|
|
|
if ($c->context !== null && !$c->context->isEmpty()) { |
923
|
|
|
throw new \RuntimeException('Context cannot be empty.'); |
924
|
|
|
} |
925
|
|
|
|
926
|
|
|
if ($fullCtx || $t === IntStream::EOF) { |
927
|
|
|
if ($skippedStopStates === null) { |
928
|
|
|
$skippedStopStates = []; |
929
|
|
|
} |
930
|
|
|
|
931
|
|
|
$skippedStopStates[] = $c; |
932
|
|
|
|
933
|
|
|
if (self::$debug_add) { |
934
|
|
|
$this->log[] = \sprintf('Added %s to skippedStopStates', (string) $c); |
935
|
|
|
} |
936
|
|
|
} |
937
|
|
|
|
938
|
|
|
continue; |
939
|
|
|
} |
940
|
|
|
|
941
|
3 |
|
foreach ($c->state->getTransitions() as $trans) { |
942
|
3 |
|
$target = $this->getReachableTarget($trans, $t); |
943
|
|
|
|
944
|
3 |
|
if ($target !== null) { |
945
|
3 |
|
$cfg = new ATNConfig($c, $target); |
946
|
3 |
|
$intermediate->add($cfg, $this->mergeCache); |
947
|
|
|
|
948
|
3 |
|
if (self::$debug_add) { |
949
|
|
|
$this->log[] = \sprintf('Added %s to intermediate', (string) $cfg); |
950
|
|
|
} |
951
|
|
|
} |
952
|
|
|
} |
953
|
|
|
} |
954
|
|
|
|
955
|
|
|
// Now figure out where the reach operation can take us... |
956
|
|
|
|
957
|
3 |
|
$reach = null; |
958
|
|
|
|
959
|
|
|
/* This block optimizes the reach operation for intermediate sets which |
960
|
|
|
* trivially indicate a termination state for the overall |
961
|
|
|
* adaptivePredict operation. |
962
|
|
|
* |
963
|
|
|
* The conditions assume that intermediate contains all configurations |
964
|
|
|
* relevant to the reach set, but this condition is not true when one |
965
|
|
|
* or more configurations have been withheld in skippedStopStates, or |
966
|
|
|
* when the current symbol is EOF. |
967
|
|
|
*/ |
968
|
3 |
|
if ($skippedStopStates === null && $t !== Token::EOF) { |
969
|
3 |
|
if (\count($intermediate->elements()) === 1) { |
970
|
|
|
// Don't pursue the closure if there is just one state. |
971
|
|
|
// It can only have one alternative; just add to result |
972
|
|
|
// Also don't pursue the closure if there is unique alternative |
973
|
|
|
// among the configurations. |
974
|
|
|
|
975
|
3 |
|
$reach = $intermediate; |
976
|
2 |
|
} elseif (self::getUniqueAlt($intermediate) !== ATN::INVALID_ALT_NUMBER) { |
977
|
|
|
// Also don't pursue the closure if there is unique alternative among the configurations. |
978
|
2 |
|
$reach = $intermediate; |
979
|
|
|
} |
980
|
|
|
} |
981
|
|
|
|
982
|
|
|
// If the reach set could not be trivially determined, perform a closure |
983
|
|
|
// operation on the intermediate set to compute its initial value. |
984
|
3 |
|
if ($reach === null) { |
985
|
1 |
|
$reach = new ATNConfigSet($fullCtx); |
986
|
1 |
|
$closureBusy = new Set(); |
987
|
1 |
|
$treatEofAsEpsilon = $t === Token::EOF; |
988
|
|
|
|
989
|
1 |
|
foreach ($intermediate->elements() as $item) { |
990
|
1 |
|
$this->closure($item, $reach, $closureBusy, false, $fullCtx, $treatEofAsEpsilon); |
991
|
|
|
} |
992
|
|
|
} |
993
|
|
|
|
994
|
3 |
|
if ($t === IntStream::EOF) { |
995
|
|
|
/* After consuming EOF no additional input is possible, so we are |
996
|
|
|
* only interested in configurations which reached the end of the |
997
|
|
|
* decision rule (local context) or end of the start rule (full |
998
|
|
|
* context). Update reach to contain only these configurations. This |
999
|
|
|
* handles both explicit EOF transitions in the grammar and implicit |
1000
|
|
|
* EOF transitions following the end of the decision or start rule. |
1001
|
|
|
* |
1002
|
|
|
* When reach==intermediate, no closure operation was performed. In |
1003
|
|
|
* this case, removeAllConfigsNotInRuleStopState needs to check for |
1004
|
|
|
* reachable rule stop states as well as configurations already in |
1005
|
|
|
* a rule stop state. |
1006
|
|
|
* |
1007
|
|
|
* This is handled before the configurations in skippedStopStates, |
1008
|
|
|
* because any configurations potentially added from that list are |
1009
|
|
|
* already guaranteed to meet this condition whether or not it's |
1010
|
|
|
* required. |
1011
|
|
|
*/ |
1012
|
|
|
|
1013
|
1 |
|
$reach = $this->removeAllConfigsNotInRuleStopState($reach, $reach->equals($intermediate)); |
1014
|
|
|
} |
1015
|
|
|
|
1016
|
|
|
/* If `skippedStopStates !== null`, then it contains at least one |
1017
|
|
|
* configuration. For full-context reach operations, these |
1018
|
|
|
* configurations reached the end of the start rule, in which case we |
1019
|
|
|
* only add them back to reach if no configuration during the current |
1020
|
|
|
* closure operation reached such a state. This ensures adaptivePredict |
1021
|
|
|
* chooses an alternative matching the longest overall sequence when |
1022
|
|
|
* multiple alternatives are viable.*/ |
1023
|
|
|
|
1024
|
3 |
|
if ($skippedStopStates !== null && (!$fullCtx || !PredictionMode::hasConfigInRuleStopState($reach))) { |
|
|
|
|
1025
|
|
|
if (\count($skippedStopStates) === 0) { |
1026
|
|
|
throw new \RuntimeException('Skipped stop states cannot be empty.'); |
1027
|
|
|
} |
1028
|
|
|
|
1029
|
|
|
foreach ($skippedStopStates as $lValue) { |
1030
|
|
|
$reach->add($lValue, $this->mergeCache); |
1031
|
|
|
} |
1032
|
|
|
} |
1033
|
|
|
|
1034
|
3 |
|
if ($reach->isEmpty()) { |
1035
|
1 |
|
return null; |
1036
|
|
|
} |
1037
|
|
|
|
1038
|
3 |
|
return $reach; |
1039
|
|
|
} |
1040
|
|
|
|
1041
|
|
|
/** |
1042
|
|
|
* Return a configuration set containing only the configurations from |
1043
|
|
|
* `configs` which are in a {@see RuleStopState}. If all configurations in |
1044
|
|
|
* `configs` are already in a rule stop state, this method simply returns |
1045
|
|
|
* `configs`. |
1046
|
|
|
* |
1047
|
|
|
* When `lookToEndOfRule` is true, this method uses {@see ATN::nextTokens()} |
1048
|
|
|
* for each configuration in `configs` which is not already in a rule stop |
1049
|
|
|
* state to see if a rule stop state is reachable from the configuration via |
1050
|
|
|
* epsilon-only transitions. |
1051
|
|
|
* |
1052
|
|
|
* @param ATNConfigSet $configs The configuration set to update |
1053
|
|
|
* @param bool $lookToEndOfRule When true, this method checks for |
1054
|
|
|
* rule stop states reachable by |
1055
|
|
|
* epsilon-only transitions from each |
1056
|
|
|
* configuration in `configs`. |
1057
|
|
|
* |
1058
|
|
|
* @return ATNConfigSet `configs` if all configurations in `configs` are |
1059
|
|
|
* in a rule stop state, otherwise return a new |
1060
|
|
|
* configuration set containing only the configurations |
1061
|
|
|
* from `configs` which are in a rule stop state. |
1062
|
|
|
* |
1063
|
|
|
* @throws \InvalidArgumentException |
1064
|
|
|
*/ |
1065
|
1 |
|
protected function removeAllConfigsNotInRuleStopState(ATNConfigSet $configs, bool $lookToEndOfRule) : ATNConfigSet |
1066
|
|
|
{ |
1067
|
1 |
|
if (PredictionMode::allConfigsInRuleStopStates($configs)) { |
1068
|
1 |
|
return $configs; |
1069
|
|
|
} |
1070
|
|
|
|
1071
|
|
|
$result = new ATNConfigSet($configs->fullCtx); |
1072
|
|
|
|
1073
|
|
|
foreach ($configs->elements() as $config) { |
1074
|
|
|
if ($config->state instanceof RuleStopState) { |
1075
|
|
|
$result->add($config, $this->mergeCache); |
1076
|
|
|
|
1077
|
|
|
continue; |
1078
|
|
|
} |
1079
|
|
|
|
1080
|
|
|
if ($lookToEndOfRule && $config->state->onlyHasEpsilonTransitions()) { |
1081
|
|
|
$nextTokens = $this->atn->nextTokens($config->state); |
1082
|
|
|
|
1083
|
|
|
if ($nextTokens->contains(Token::EPSILON)) { |
1084
|
|
|
$endOfRuleState = $this->atn->ruleToStopState[$config->state->ruleIndex]; |
1085
|
|
|
$result->add(new ATNConfig($config, $endOfRuleState), $this->mergeCache); |
1086
|
|
|
} |
1087
|
|
|
} |
1088
|
|
|
} |
1089
|
|
|
|
1090
|
|
|
return $result; |
1091
|
|
|
} |
1092
|
|
|
|
1093
|
1 |
|
protected function computeStartState(ATNState $p, RuleContext $ctx, bool $fullCtx) : ATNConfigSet |
1094
|
|
|
{ |
1095
|
|
|
// Always at least the implicit call to start rule |
1096
|
1 |
|
$initialContext = PredictionContext::fromRuleContext($this->atn, $ctx); |
1097
|
1 |
|
$configs = new ATNConfigSet($fullCtx); |
1098
|
|
|
|
1099
|
1 |
|
foreach ($p->getTransitions() as $i => $t) { |
1100
|
1 |
|
$c = new ATNConfig(null, $t->target, $initialContext, null, $i + 1); |
1101
|
1 |
|
$closureBusy = new Set(); |
1102
|
|
|
|
1103
|
1 |
|
$this->closure($c, $configs, $closureBusy, true, $fullCtx, false); |
1104
|
|
|
} |
1105
|
|
|
|
1106
|
1 |
|
return $configs; |
1107
|
|
|
} |
1108
|
|
|
|
1109
|
|
|
/* |
1110
|
|
|
* parrt internal source braindump that doesn't mess up external API spec. |
1111
|
|
|
* |
1112
|
|
|
* Context-sensitive in that they can only be properly evaluated in the context |
1113
|
|
|
* of the proper prec argument. Without pruning, these predicates are normal |
1114
|
|
|
* predicates evaluated when we reach conflict state (or unique prediction). |
1115
|
|
|
* As we cannot evaluate these predicates out of context, the resulting |
1116
|
|
|
* conflict leads to full LL evaluation and nonlinear prediction which |
1117
|
|
|
* shows up very clearly with fairly large expressions. |
1118
|
|
|
* |
1119
|
|
|
* Example grammar: |
1120
|
|
|
* e |
1121
|
|
|
* : e '*' e |
1122
|
|
|
* | e '+' e |
1123
|
|
|
* | INT |
1124
|
|
|
* ; |
1125
|
|
|
* |
1126
|
|
|
* We convert that to the following: |
1127
|
|
|
* |
1128
|
|
|
* e[int prec] |
1129
|
|
|
* : INT ( {3>=prec}? '*' e[4] | {2>=prec}? '+' e[3] )* |
1130
|
|
|
* ; |
1131
|
|
|
* |
1132
|
|
|
* The (..)* loop has a decision for the inner block as well as an enter |
1133
|
|
|
* or exit decision, which is what concerns us here. At the 1st + of input |
1134
|
|
|
* 1+2+3, the loop entry sees both predicates and the loop exit also sees |
1135
|
|
|
* both predicates by falling off the edge of e. This is because we have |
1136
|
|
|
* no stack information with SLL and find the follow of e, which will |
1137
|
|
|
* hit the return states inside the loop after e[4] and e[3], which brings |
1138
|
|
|
* it back to the enter or exit decision. In this case, we know that we |
1139
|
|
|
* cannot evaluate those predicates because we have fallen off the edge |
1140
|
|
|
* of the stack and will in general not know which prec parameter is |
1141
|
|
|
* the right one to use in the predicate. |
1142
|
|
|
* |
1143
|
|
|
* Because we have special information, that these are precedence predicates, |
1144
|
|
|
* we can resolve them without failing over to full LL despite their context |
1145
|
|
|
* sensitive nature. We make an assumption that prec[-1] <= prec[0], meaning |
1146
|
|
|
* that the current precedence level is greater than or equal to the precedence |
1147
|
|
|
* level of recursive invocations above us in the stack. For example, if |
1148
|
|
|
* predicate {3>=prec}? is true of the current prec, then one option is to |
1149
|
|
|
* enter the loop to match it now. The other option is to exit the loop and |
1150
|
|
|
* the left recursive rule to match the current operator in rule invocation |
1151
|
|
|
* further up the stack. But, we know that all of those prec are lower or |
1152
|
|
|
* the same value and so we can decide to enter the loop instead of matching |
1153
|
|
|
* it later. That means we can strip out the other configuration for the exit branch. |
1154
|
|
|
* |
1155
|
|
|
* So imagine we have (14,1,$,{2>=prec}?) and then |
1156
|
|
|
* (14,2,$-dipsIntoOuterContext,{2>=prec}?). The optimization allows us to |
1157
|
|
|
* collapse these two configurations. We know that if {2>=prec}? is true |
1158
|
|
|
* for the current prec parameter, it will also be true for any precfrom |
1159
|
|
|
* an invoking e call, indicated by dipsIntoOuterContext. As the predicates |
1160
|
|
|
* are both true, we have the option to evaluate them early in the decision |
1161
|
|
|
* start state. We do this by stripping both predicates and choosing to |
1162
|
|
|
* enter the loop as it is consistent with the notion of operator precedence. |
1163
|
|
|
* It's also how the full LL conflict resolution would work. |
1164
|
|
|
* |
1165
|
|
|
* The solution requires a different DFA start state for each precedence level. |
1166
|
|
|
* |
1167
|
|
|
* The basic filter mechanism is to remove configurations of the form (p, 2, pi) |
1168
|
|
|
* if (p, 1, pi) exists for the same p and pi. In other words, for the same |
1169
|
|
|
* ATN state and predicate context, remove any configuration associated with |
1170
|
|
|
* an exit branch if there is a configuration associated with the enter branch. |
1171
|
|
|
* |
1172
|
|
|
* It's also the case that the filter evaluates precedence predicates and |
1173
|
|
|
* resolves conflicts according to precedence levels. For example, for input |
1174
|
|
|
* 1+2+3 at the first +, we see prediction filtering. |
1175
|
|
|
* |
1176
|
|
|
* [(11,1,[$],{3>=prec}?), (14,1,[$],{2>=prec}?), (5,2,[$],up=1), |
1177
|
|
|
* (11,2,[$],up=1), (14,2,[$],up=1)], hasSemanticContext=true,dipsIntoOuterContext |
1178
|
|
|
* |
1179
|
|
|
* to |
1180
|
|
|
* |
1181
|
|
|
* [(11,1,[$]), (14,1,[$]), (5,2,[$],up=1)],dipsIntoOuterContext |
1182
|
|
|
* |
1183
|
|
|
* This filters because {3>=prec}? evals to true and collapses |
1184
|
|
|
* (11,1,[$],{3>=prec}?) and (11,2,[$],up=1) since early conflict |
1185
|
|
|
* resolution based upon rules of operator precedence fits with our |
1186
|
|
|
* usual match first alt upon conflict. |
1187
|
|
|
* |
1188
|
|
|
* We noticed a problem where a recursive call resets precedence to 0. |
1189
|
|
|
* Sam's fix: each config has flag indicating if it has returned from |
1190
|
|
|
* an expr[0] call. then just don't filter any config with that flag set. |
1191
|
|
|
* flag is carried along in closure(). so to avoid adding field, set bit |
1192
|
|
|
* just under sign bit of dipsIntoOuterContext (SUPPRESS_PRECEDENCE_FILTER). |
1193
|
|
|
* With the change you filter "unless (p, 2, pi) was reached after leaving |
1194
|
|
|
* the rule stop state of the LR rule containing state p, corresponding |
1195
|
|
|
* to a rule invocation with precedence level 0". |
1196
|
|
|
*/ |
1197
|
|
|
|
1198
|
|
|
/** |
1199
|
|
|
* This method transforms the start state computed by |
1200
|
|
|
* {@see ParserATNSimulator::computeStartState()} to the special start state |
1201
|
|
|
* used by a precedence DFA for a particular precedence value. The transformation |
1202
|
|
|
* process applies the following changes to the start state's configuration |
1203
|
|
|
* set. |
1204
|
|
|
* |
1205
|
|
|
* - Evaluate the precedence predicates for each configuration using |
1206
|
|
|
* {@see SemanticContext//evalPrecedence}. |
1207
|
|
|
* - Remove all configurations which predict an alternative greater than |
1208
|
|
|
* 1, for which another configuration that predicts alternative 1 is in the |
1209
|
|
|
* same ATN state with the same prediction context. This transformation is |
1210
|
|
|
* valid for the following reasons: |
1211
|
|
|
* - The closure block cannot contain any epsilon transitions which bypass |
1212
|
|
|
* the body of the closure, so all states reachable via alternative 1 are |
1213
|
|
|
* part of the precedence alternatives of the transformed left-recursive |
1214
|
|
|
* rule. |
1215
|
|
|
* - The "primary" portion of a left recursive rule cannot contain an |
1216
|
|
|
* epsilon transition, so the only way an alternative other than 1 can exist |
1217
|
|
|
* in a state that is also reachable via alternative 1 is by nesting calls |
1218
|
|
|
* to the left-recursive rule, with the outer calls not being at the |
1219
|
|
|
* preferred precedence level. |
1220
|
|
|
* |
1221
|
|
|
* The prediction context must be considered by this filter to address |
1222
|
|
|
* situations like the following. |
1223
|
|
|
* |
1224
|
|
|
* grammar TA; |
1225
|
|
|
* prog: statement* EOF; |
1226
|
|
|
* statement: letterA | statement letterA 'b' ; |
1227
|
|
|
* letterA: 'a'; |
1228
|
|
|
* |
1229
|
|
|
* If the above grammar, the ATN state immediately before the token |
1230
|
|
|
* reference `'a'` in `letterA` is reachable from the left edge |
1231
|
|
|
* of both the primary and closure blocks of the left-recursive rule |
1232
|
|
|
* `statement`. The prediction context associated with each of these |
1233
|
|
|
* configurations distinguishes between them, and prevents the alternative |
1234
|
|
|
* which stepped out to `prog` (and then back in to `statement` from being |
1235
|
|
|
* eliminated by the filter. |
1236
|
|
|
* |
1237
|
|
|
* @param ATNConfigSet $configs The configuration set computed by |
1238
|
|
|
* {@see ParserATNSimulator::computeStartState()} |
1239
|
|
|
* as the start state for the DFA. |
1240
|
|
|
* |
1241
|
|
|
* @return ATNConfigSet The transformed configuration set representing the start state |
1242
|
|
|
* for a precedence DFA at a particular precedence level |
1243
|
|
|
* (determined by calling {@see Parser::getPrecedence()}). |
1244
|
|
|
* |
1245
|
|
|
* @throws \InvalidArgumentException |
1246
|
|
|
*/ |
1247
|
1 |
|
protected function applyPrecedenceFilter(ATNConfigSet $configs) : ATNConfigSet |
1248
|
|
|
{ |
1249
|
|
|
/** @var array<PredictionContext> $statesFromAlt1 */ |
1250
|
1 |
|
$statesFromAlt1 = []; |
1251
|
1 |
|
$configSet = new ATNConfigSet($configs->fullCtx); |
1252
|
|
|
|
1253
|
1 |
|
foreach ($configs->elements() as $config) { |
1254
|
|
|
// handle alt 1 first |
1255
|
1 |
|
if ($config->alt !== 1) { |
1256
|
1 |
|
continue; |
1257
|
|
|
} |
1258
|
|
|
|
1259
|
1 |
|
$updatedContext = $this->outerContext !== null ? |
1260
|
1 |
|
$config->semanticContext->evalPrecedence($this->parser, $this->outerContext) : |
1261
|
|
|
null; |
1262
|
|
|
|
1263
|
1 |
|
if ($updatedContext === null) { |
1264
|
|
|
continue; |
1265
|
|
|
} |
1266
|
|
|
|
1267
|
1 |
|
$statesFromAlt1[$config->state->stateNumber] = $config->context; |
1268
|
|
|
|
1269
|
1 |
|
if (!$updatedContext->equals($config->semanticContext)) { |
1270
|
1 |
|
$configSet->add( |
1271
|
1 |
|
new ATNConfig($config, null, null, $updatedContext), |
1272
|
1 |
|
$this->mergeCache |
1273
|
|
|
); |
1274
|
|
|
} else { |
1275
|
|
|
$configSet->add($config, $this->mergeCache); |
1276
|
|
|
} |
1277
|
|
|
} |
1278
|
|
|
|
1279
|
1 |
|
foreach ($configs->elements() as $config) { |
1280
|
1 |
|
if ($config->alt === 1) { |
1281
|
1 |
|
continue; // already handled |
1282
|
|
|
} |
1283
|
|
|
|
1284
|
|
|
/* In the future, this elimination step could be updated to also |
1285
|
|
|
* filter the prediction context for alternatives predicting alt>1 |
1286
|
|
|
* (basically a graph subtraction algorithm). |
1287
|
|
|
*/ |
1288
|
1 |
|
if (!$config->isPrecedenceFilterSuppressed()) { |
1289
|
|
|
$context = $statesFromAlt1[$config->state->stateNumber] ?? null; |
1290
|
|
|
|
1291
|
|
|
if ($context !== null && $config->context !== null && $context->equals($config->context)) { |
1292
|
|
|
continue; // eliminated |
1293
|
|
|
} |
1294
|
|
|
} |
1295
|
|
|
|
1296
|
1 |
|
$configSet->add($config, $this->mergeCache); |
1297
|
|
|
} |
1298
|
|
|
|
1299
|
1 |
|
return $configSet; |
1300
|
|
|
} |
1301
|
|
|
|
1302
|
3 |
|
protected function getReachableTarget(Transition $trans, int $ttype) : ?ATNState |
1303
|
|
|
{ |
1304
|
3 |
|
return $trans->matches($ttype, 0, $this->atn->maxTokenType) ? $trans->target : null; |
1305
|
|
|
} |
1306
|
|
|
|
1307
|
|
|
/** |
1308
|
|
|
* @return array<SemanticContext>|null |
1309
|
|
|
*/ |
1310
|
2 |
|
protected function getPredsForAmbigAlts(BitSet $ambigAlts, ATNConfigSet $configs, int $nalts) : ?array |
1311
|
|
|
{ |
1312
|
|
|
/* REACH=[1|1|[]|0:0, 1|2|[]|0:1] |
1313
|
|
|
* altToPred starts as an array of all null contexts. The entry at index i |
1314
|
|
|
* corresponds to alternative i. altToPred[i] may have one of three values: |
1315
|
|
|
* 1. null: no ATNConfig c is found such that c.alt==i |
1316
|
|
|
* 2. SemanticContext.NONE: At least one ATNConfig c exists such that |
1317
|
|
|
* c.alt==i and c.semanticContext==SemanticContext.NONE. In other words, |
1318
|
|
|
* alt i has at least one unpredicated config. |
1319
|
|
|
* 3. Non-NONE Semantic Context: There exists at least one, and for all |
1320
|
|
|
* ATNConfig c such that c.alt==i, c.semanticContext!=SemanticContext.NONE. |
1321
|
|
|
* |
1322
|
|
|
* From this, it is clear that NONE||anything==NONE. |
1323
|
|
|
*/ |
1324
|
|
|
|
1325
|
2 |
|
$altToPred = new \SplFixedArray($nalts + 1); |
1326
|
|
|
|
1327
|
2 |
|
foreach ($configs->elements() as $c) { |
1328
|
2 |
|
if ($ambigAlts->contains($c->alt)) { |
1329
|
2 |
|
$altToPred[$c->alt] = SemanticContext::orContext($altToPred[$c->alt] ?? null, $c->semanticContext); |
1330
|
|
|
} |
1331
|
|
|
} |
1332
|
|
|
|
1333
|
2 |
|
$nPredAlts = 0; |
1334
|
|
|
|
1335
|
2 |
|
for ($i = 1; $i <= $nalts; $i++) { |
1336
|
2 |
|
$pred = $altToPred[$i]; |
1337
|
|
|
|
1338
|
2 |
|
if ($pred === null) { |
1339
|
2 |
|
$altToPred[$i] = SemanticContext::none(); |
1340
|
2 |
|
} elseif ($pred !== SemanticContext::none()) { |
1341
|
2 |
|
$nPredAlts++; |
1342
|
|
|
} |
1343
|
|
|
} |
1344
|
|
|
|
1345
|
|
|
// nonambig alts are null in altToPred |
1346
|
2 |
|
if ($nPredAlts === 0) { |
|
|
|
|
1347
|
|
|
$altToPred = null; |
1348
|
|
|
} else { |
1349
|
2 |
|
$altToPred = $altToPred->toArray(); |
1350
|
|
|
} |
1351
|
|
|
|
1352
|
2 |
|
if (self::$debug) { |
1353
|
|
|
$this->log[] = \sprintf( |
1354
|
|
|
'getPredsForAmbigAlts result [%s]', |
1355
|
|
|
$altToPred === null ? '' : \implode(', ', $altToPred) |
|
|
|
|
1356
|
|
|
); |
1357
|
|
|
} |
1358
|
|
|
|
1359
|
2 |
|
return $altToPred; |
1360
|
|
|
} |
1361
|
|
|
|
1362
|
|
|
/** |
1363
|
|
|
* @param array<SemanticContext> $altToPred |
1364
|
|
|
* |
1365
|
|
|
* @return array<PredPrediction>|null |
1366
|
|
|
*/ |
1367
|
2 |
|
protected function getPredicatePredictions(?BitSet $ambigAlts, array $altToPred) : ?array |
1368
|
|
|
{ |
1369
|
2 |
|
$pairs = []; |
1370
|
2 |
|
$containsPredicate = false; |
1371
|
2 |
|
$count = \count($altToPred); |
1372
|
|
|
|
1373
|
2 |
|
for ($i = 1; $i < $count; $i++) { |
1374
|
2 |
|
$pred = $altToPred[$i]; |
1375
|
|
|
|
1376
|
|
|
// unpredicated is indicated by SemanticContext.NONE |
1377
|
2 |
|
if ($ambigAlts !== null && $ambigAlts->contains($i)) { |
1378
|
2 |
|
$pairs[] = new PredPrediction($pred, $i); |
1379
|
|
|
} |
1380
|
|
|
|
1381
|
2 |
|
if ($pred !== SemanticContext::none()) { |
1382
|
2 |
|
$containsPredicate = true; |
1383
|
|
|
} |
1384
|
|
|
} |
1385
|
|
|
|
1386
|
2 |
|
if (!$containsPredicate) { |
1387
|
|
|
return null; |
1388
|
|
|
} |
1389
|
|
|
|
1390
|
2 |
|
return $pairs; |
1391
|
|
|
} |
1392
|
|
|
|
1393
|
|
|
/** |
1394
|
|
|
* This method is used to improve the localization of error messages by |
1395
|
|
|
* choosing an alternative rather than throwing a |
1396
|
|
|
* {@see NoViableAltException} in particular prediction scenarios where the |
1397
|
|
|
* {@see ParserATNSimulator::error()} state was reached during ATN simulation. |
1398
|
|
|
* |
1399
|
|
|
* The default implementation of this method uses the following |
1400
|
|
|
* algorithm to identify an ATN configuration which successfully parsed the |
1401
|
|
|
* decision entry rule. Choosing such an alternative ensures that the |
1402
|
|
|
* {@see ParserRuleContext} returned by the calling rule will be complete |
1403
|
|
|
* and valid, and the syntax error will be reported later at a more |
1404
|
|
|
* localized location. |
1405
|
|
|
* |
1406
|
|
|
* - If a syntactically valid path or paths reach the end of the decision rule and |
1407
|
|
|
* they are semantically valid if predicated, return the min associated alt. |
1408
|
|
|
* - Else, if a semantically invalid but syntactically valid path exist |
1409
|
|
|
* or paths exist, return the minimum associated alt. |
1410
|
|
|
* - Otherwise, return {@see ATN::INVALID_ALT_NUMBER}. |
1411
|
|
|
* |
1412
|
|
|
* In some scenarios, the algorithm described above could predict an |
1413
|
|
|
* alternative which will result in a {@see FailedPredicateException} in |
1414
|
|
|
* the parser. Specifically, this could occur if the only configuration |
1415
|
|
|
* capable of successfully parsing to the end of the decision rule is |
1416
|
|
|
* blocked by a semantic predicate. By choosing this alternative within |
1417
|
|
|
* {@see ParserATNSimulator::adaptivePredict()} instead of throwing a |
1418
|
|
|
* {@see NoViableAltException}, the resulting {@see FailedPredicateException} |
1419
|
|
|
* in the parser will identify the specific predicate which is preventing |
1420
|
|
|
* the parser from successfully parsing the decision rule, which helps |
1421
|
|
|
* developers identify and correct logic errors in semantic predicates. |
1422
|
|
|
* |
1423
|
|
|
* @param ATNConfigSet $configs The ATN configurations which were valid |
1424
|
|
|
* immediately before the |
1425
|
|
|
* {@see ParserATNSimulator::error()} |
1426
|
|
|
* state was reached. |
1427
|
|
|
* @param ParserRuleContext $outerContext The \gamma_0 initial parser context |
1428
|
|
|
* from the paper or the parser stack |
1429
|
|
|
* at the instant before prediction commences. |
1430
|
|
|
* |
1431
|
|
|
* @return int The value to return from {@see ParserATNSimulator::adaptivePredict()}, |
1432
|
|
|
* or {@see ATN::INVALID_ALT_NUMBER} if a suitable alternative |
1433
|
|
|
* was not identified and {@see ParserATNSimulator::::adaptivePredict()} |
1434
|
|
|
* should report an error instead. |
1435
|
|
|
* |
1436
|
|
|
* @throws \Exception |
1437
|
|
|
*/ |
1438
|
4 |
|
protected function getSynValidOrSemInvalidAltThatFinishedDecisionEntryRule( |
1439
|
|
|
ATNConfigSet $configs, |
1440
|
|
|
ParserRuleContext $outerContext |
1441
|
|
|
) : int { |
1442
|
|
|
/** @var ATNConfigSet $semValidConfigs */ |
1443
|
|
|
/** @var ATNConfigSet $semInvalidConfigs */ |
1444
|
4 |
|
[$semValidConfigs, $semInvalidConfigs] = $this->splitAccordingToSemanticValidity($configs, $outerContext); |
1445
|
|
|
|
1446
|
4 |
|
$alt = $this->getAltThatFinishedDecisionEntryRule($semValidConfigs); |
1447
|
|
|
|
1448
|
4 |
|
if ($alt !== ATN::INVALID_ALT_NUMBER) { |
1449
|
|
|
// semantically/syntactically viable path exists |
1450
|
|
|
|
1451
|
4 |
|
return $alt; |
1452
|
|
|
} |
1453
|
|
|
|
1454
|
|
|
// Is there a syntactically valid path with a failed pred? |
1455
|
|
|
if ($semInvalidConfigs->getLength() > 0) { |
1456
|
|
|
$alt = $this->getAltThatFinishedDecisionEntryRule($semInvalidConfigs); |
1457
|
|
|
|
1458
|
|
|
if ($alt !== ATN::INVALID_ALT_NUMBER) { |
1459
|
|
|
// syntactically viable path exists |
1460
|
|
|
|
1461
|
|
|
return $alt; |
1462
|
|
|
} |
1463
|
|
|
} |
1464
|
|
|
|
1465
|
|
|
return ATN::INVALID_ALT_NUMBER; |
1466
|
|
|
} |
1467
|
|
|
|
1468
|
4 |
|
protected function getAltThatFinishedDecisionEntryRule(ATNConfigSet $configs) : int |
1469
|
|
|
{ |
1470
|
4 |
|
$alts = new IntervalSet(); |
1471
|
|
|
/** @var ATNConfig $c */ |
1472
|
4 |
|
foreach ($configs->elements() as $c) { |
1473
|
4 |
|
if ($c->getOuterContextDepth() > 0 |
1474
|
4 |
|
|| ($c->state instanceof RuleStopState && $c->context !== null && $c->context->hasEmptyPath())) { |
1475
|
4 |
|
$alts->addOne($c->alt); |
1476
|
|
|
} |
1477
|
|
|
} |
1478
|
|
|
|
1479
|
4 |
|
return $alts->length() === 0 ? ATN::INVALID_ALT_NUMBER : $alts->getMinElement(); |
1480
|
|
|
} |
1481
|
|
|
|
1482
|
|
|
/** |
1483
|
|
|
* Walk the list of configurations and split them according to |
1484
|
|
|
* those that have preds evaluating to true/false. If no pred, assume |
1485
|
|
|
* true pred and include in succeeded set. Returns Pair of sets. |
1486
|
|
|
* |
1487
|
|
|
* Create a new set so as not to alter the incoming parameter. |
1488
|
|
|
* |
1489
|
|
|
* Assumption: the input stream has been restored to the starting point |
1490
|
|
|
* prediction, which is where predicates need to evaluate. |
1491
|
|
|
* |
1492
|
|
|
* @return array<ATNConfigSet> |
1493
|
|
|
|
1494
|
|
|
* @throws \Exception |
1495
|
|
|
*/ |
1496
|
4 |
|
protected function splitAccordingToSemanticValidity(ATNConfigSet $configs, ParserRuleContext $outerContext) : array |
1497
|
|
|
{ |
1498
|
4 |
|
$succeeded = new ATNConfigSet($configs->fullCtx); |
1499
|
4 |
|
$failed = new ATNConfigSet($configs->fullCtx); |
1500
|
|
|
|
1501
|
4 |
|
foreach ($configs->elements() as $c) { |
1502
|
4 |
|
if ($c->semanticContext !== SemanticContext::none()) { |
1503
|
|
|
$predicateEvaluationResult = $c->semanticContext->eval($this->parser, $outerContext); |
1504
|
|
|
|
1505
|
|
|
if ($predicateEvaluationResult) { |
1506
|
|
|
$succeeded->add($c); |
1507
|
|
|
} else { |
1508
|
|
|
$failed->add($c); |
1509
|
|
|
} |
1510
|
|
|
} else { |
1511
|
4 |
|
$succeeded->add($c); |
1512
|
|
|
} |
1513
|
|
|
} |
1514
|
|
|
|
1515
|
4 |
|
return [$succeeded, $failed]; |
1516
|
|
|
} |
1517
|
|
|
|
1518
|
|
|
/** |
1519
|
|
|
* Look through a list of predicate/alt pairs, returning alts for the |
1520
|
|
|
* pairs that win. A `NONE` predicate indicates an alt containing an |
1521
|
|
|
* unpredicated config which behaves as "always true." If !complete |
1522
|
|
|
* then we stop at the first predicate that evaluates to true. This |
1523
|
|
|
* includes pairs with null predicates. |
1524
|
|
|
* |
1525
|
|
|
* @param array<PredPrediction> $predPredictions |
1526
|
|
|
*/ |
1527
|
3 |
|
protected function evalSemanticContextMany( |
1528
|
|
|
array $predPredictions, |
1529
|
|
|
ParserRuleContext $outerContext, |
1530
|
|
|
bool $complete |
1531
|
|
|
) : BitSet { |
1532
|
3 |
|
$predictions = new BitSet(); |
1533
|
|
|
|
1534
|
3 |
|
foreach ($predPredictions as $pair) { |
1535
|
3 |
|
if ($pair->pred === SemanticContext::none()) { |
1536
|
|
|
$predictions->add($pair->alt); |
1537
|
|
|
|
1538
|
|
|
if (!$complete) { |
1539
|
|
|
break; |
1540
|
|
|
} |
1541
|
|
|
|
1542
|
|
|
continue; |
1543
|
|
|
} |
1544
|
|
|
|
1545
|
3 |
|
$fullCtx = false; // in dfa |
1546
|
|
|
|
1547
|
3 |
|
$predicateEvaluationResult = $this->evalSemanticContextOne( |
1548
|
3 |
|
$pair->pred, |
1549
|
|
|
$outerContext, |
1550
|
3 |
|
$pair->alt, |
1551
|
|
|
$fullCtx |
1552
|
|
|
); |
1553
|
|
|
|
1554
|
3 |
|
if (self::$debug || self::$dfa_debug) { |
1555
|
|
|
$this->log[] = \sprintf('eval pred $pair = %s"', $predicateEvaluationResult); |
1556
|
|
|
} |
1557
|
|
|
|
1558
|
3 |
|
if ($predicateEvaluationResult) { |
1559
|
3 |
|
if (self::$debug || self::$dfa_debug) { |
1560
|
|
|
$this->log[] = \sprintf('PREDICT %d', $pair->alt); |
1561
|
|
|
} |
1562
|
|
|
|
1563
|
3 |
|
$predictions->add($pair->alt); |
1564
|
|
|
|
1565
|
3 |
|
if (!$complete) { |
1566
|
|
|
break; |
1567
|
|
|
} |
1568
|
|
|
} |
1569
|
|
|
} |
1570
|
|
|
|
1571
|
3 |
|
return $predictions; |
1572
|
|
|
} |
1573
|
|
|
|
1574
|
3 |
|
protected function evalSemanticContextOne( |
1575
|
|
|
SemanticContext $pred, |
1576
|
|
|
ParserRuleContext $parserCallStack, |
1577
|
|
|
int $alt, |
|
|
|
|
1578
|
|
|
bool $fullCtx |
|
|
|
|
1579
|
|
|
) : bool { |
1580
|
3 |
|
return $pred->eval($this->parser, $parserCallStack); |
1581
|
|
|
} |
1582
|
|
|
|
1583
|
|
|
/** |
1584
|
|
|
* TODO: If we are doing predicates, there is no point in pursuing |
1585
|
|
|
* closure operations if we reach a DFA state that uniquely predicts |
1586
|
|
|
* alternative. We will not be caching that DFA state and it is a |
1587
|
|
|
* waste to pursue the closure. Might have to advance when we do |
1588
|
|
|
* ambig detection thought :( |
1589
|
|
|
*/ |
1590
|
1 |
|
protected function closure( |
1591
|
|
|
ATNConfig $config, |
1592
|
|
|
ATNConfigSet $configs, |
1593
|
|
|
Set $closureBusy, |
1594
|
|
|
bool $collectPredicates, |
1595
|
|
|
bool $fullCtx, |
1596
|
|
|
bool $treatEofAsEpsilon |
1597
|
|
|
) : void { |
1598
|
1 |
|
$initialDepth = 0; |
1599
|
|
|
|
1600
|
1 |
|
$this->closureCheckingStopState( |
1601
|
1 |
|
$config, |
1602
|
|
|
$configs, |
1603
|
|
|
$closureBusy, |
1604
|
|
|
$collectPredicates, |
1605
|
|
|
$fullCtx, |
1606
|
|
|
$initialDepth, |
1607
|
|
|
$treatEofAsEpsilon |
1608
|
|
|
); |
1609
|
|
|
|
1610
|
1 |
|
if ($fullCtx && $configs->dipsIntoOuterContext) { |
1611
|
|
|
throw new \RuntimeException('Error.'); |
1612
|
|
|
} |
1613
|
1 |
|
} |
1614
|
|
|
|
1615
|
1 |
|
protected function closureCheckingStopState( |
1616
|
|
|
ATNConfig $config, |
1617
|
|
|
ATNConfigSet $configs, |
1618
|
|
|
Set $closureBusy, |
1619
|
|
|
bool $collectPredicates, |
1620
|
|
|
bool $fullCtx, |
1621
|
|
|
int $depth, |
1622
|
|
|
bool $treatEofAsEpsilon |
1623
|
|
|
) : void { |
1624
|
1 |
|
if (self::$debug || self::$debug_closure) { |
1625
|
|
|
$this->log[] = \sprintf('closure(%s)', $config->toString(true)); |
1626
|
|
|
|
1627
|
|
|
if ($config->reachesIntoOuterContext > 50) { |
1628
|
|
|
throw new \RuntimeException('problem'); |
1629
|
|
|
} |
1630
|
|
|
} |
1631
|
|
|
|
1632
|
1 |
|
if ($config->state instanceof RuleStopState) { |
1633
|
|
|
// We hit rule end. If we have context info, use it run thru all possible stack tops in ctx |
1634
|
|
|
|
1635
|
1 |
|
if ($config->context !== null && !$config->context->isEmpty()) { |
1636
|
1 |
|
for ($i =0; $i < $config->context->getLength(); $i++) { |
1637
|
1 |
|
if ($config->context->getReturnState($i) === PredictionContext::EMPTY_RETURN_STATE) { |
1638
|
|
|
if ($fullCtx) { |
1639
|
|
|
$configs->add( |
1640
|
|
|
new ATNConfig($config, $config->state, PredictionContext::empty(), null, null), |
1641
|
|
|
$this->mergeCache |
1642
|
|
|
); |
1643
|
|
|
} else { |
1644
|
|
|
// we have no context info, just chase follow links (if greedy) |
1645
|
|
|
if (self::$debug) { |
1646
|
|
|
$this->log[] = \sprintf( |
1647
|
|
|
'FALLING off rule %s', |
1648
|
|
|
$this->getRuleName($config->state->ruleIndex) |
1649
|
|
|
); |
1650
|
|
|
} |
1651
|
|
|
|
1652
|
|
|
$this->closure_( |
1653
|
|
|
$config, |
1654
|
|
|
$configs, |
1655
|
|
|
$closureBusy, |
1656
|
|
|
$collectPredicates, |
1657
|
|
|
$fullCtx, |
1658
|
|
|
$depth, |
1659
|
|
|
$treatEofAsEpsilon |
1660
|
|
|
); |
1661
|
|
|
} |
1662
|
|
|
|
1663
|
|
|
continue; |
1664
|
|
|
} |
1665
|
|
|
|
1666
|
1 |
|
$returnState = $this->atn->states[$config->context->getReturnState($i)]; |
1667
|
1 |
|
$newContext = $config->context->getParent($i);// "pop" return state |
1668
|
|
|
|
1669
|
1 |
|
$c = new ATNConfig(null, $returnState, $newContext, $config->semanticContext, $config->alt); |
1670
|
|
|
|
1671
|
|
|
// While we have context to pop back from, we may have |
1672
|
|
|
// gotten that context AFTER having falling off a rule. |
1673
|
|
|
// Make sure we track that we are now out of context. |
1674
|
|
|
// |
1675
|
|
|
// This assignment also propagates the |
1676
|
|
|
// isPrecedenceFilterSuppressed() value to the new |
1677
|
|
|
// configuration. |
1678
|
1 |
|
$c->reachesIntoOuterContext = $config->reachesIntoOuterContext; |
1679
|
|
|
|
1680
|
1 |
|
$this->closureCheckingStopState( |
1681
|
1 |
|
$c, |
1682
|
|
|
$configs, |
1683
|
|
|
$closureBusy, |
1684
|
|
|
$collectPredicates, |
1685
|
|
|
$fullCtx, |
1686
|
1 |
|
$depth - 1, |
1687
|
|
|
$treatEofAsEpsilon |
1688
|
|
|
); |
1689
|
|
|
} |
1690
|
|
|
|
1691
|
1 |
|
return; |
1692
|
1 |
|
} elseif ($fullCtx) { |
1693
|
|
|
// Reached end of start rule |
1694
|
|
|
$configs->add($config, $this->mergeCache); |
1695
|
|
|
|
1696
|
|
|
return; |
1697
|
|
|
} else { |
1698
|
|
|
// Else if we have no context info, just chase follow links (if greedy) |
1699
|
1 |
|
if (self::$debug) { |
1700
|
|
|
$this->log[] = \sprintf('FALLING off rule %s.', $this->getRuleName($config->state->ruleIndex)); |
1701
|
|
|
} |
1702
|
|
|
} |
1703
|
|
|
} |
1704
|
|
|
|
1705
|
1 |
|
$this->closure_($config, $configs, $closureBusy, $collectPredicates, $fullCtx, $depth, $treatEofAsEpsilon); |
1706
|
1 |
|
} |
1707
|
|
|
|
1708
|
|
|
/** |
1709
|
|
|
* Do the actual work of walking epsilon edges. |
1710
|
|
|
*/ |
1711
|
1 |
|
protected function closure_( |
1712
|
|
|
ATNConfig $config, |
1713
|
|
|
ATNConfigSet $configs, |
1714
|
|
|
Set $closureBusy, |
1715
|
|
|
bool $collectPredicates, |
1716
|
|
|
bool $fullCtx, |
1717
|
|
|
int $depth, |
1718
|
|
|
bool $treatEofAsEpsilon |
1719
|
|
|
) : void { |
1720
|
1 |
|
$p = $config->state; |
1721
|
|
|
|
1722
|
|
|
// optimization |
1723
|
1 |
|
if (!$p->onlyHasEpsilonTransitions()) { |
1724
|
|
|
// make sure to not return here, because EOF transitions can act as |
1725
|
|
|
// both epsilon transitions and non-epsilon transitions. |
1726
|
|
|
|
1727
|
1 |
|
$configs->add($config, $this->mergeCache); |
1728
|
|
|
} |
1729
|
|
|
|
1730
|
1 |
|
foreach ($p->getTransitions() as $i => $t) { |
1731
|
1 |
|
if ($i === 0 && $this->canDropLoopEntryEdgeInLeftRecursiveRule($config)) { |
1732
|
|
|
continue; |
1733
|
|
|
} |
1734
|
|
|
|
1735
|
1 |
|
$continueCollecting = $collectPredicates && !$t instanceof ActionTransition; |
1736
|
1 |
|
$c = $this->getEpsilonTarget($config, $t, $continueCollecting, $depth === 0, $fullCtx, $treatEofAsEpsilon); |
1737
|
|
|
|
1738
|
1 |
|
if ($c !== null) { |
1739
|
1 |
|
$newDepth = $depth; |
1740
|
|
|
|
1741
|
1 |
|
if ($config->state instanceof RuleStopState) { |
1742
|
1 |
|
if ($fullCtx) { |
1743
|
|
|
throw new \RuntimeException('Error.'); |
1744
|
|
|
} |
1745
|
|
|
|
1746
|
|
|
// Target fell off end of rule; mark resulting c as having dipped into outer context |
1747
|
|
|
// We can't get here if incoming config was rule stop and we had context |
1748
|
|
|
// track how far we dip into outer context. Might |
1749
|
|
|
// come in handy and we avoid evaluating context dependent |
1750
|
|
|
// preds if this is > 0. |
1751
|
|
|
|
1752
|
1 |
|
if ($this->dfa && $this->dfa->isPrecedenceDfa()) { |
1753
|
1 |
|
if ($t instanceof EpsilonTransition |
1754
|
1 |
|
&& $this->dfa->atnStartState !== null |
1755
|
1 |
|
&& $t->getOutermostPrecedenceReturn() === $this->dfa->atnStartState->ruleIndex) { |
1756
|
1 |
|
$c->setPrecedenceFilterSuppressed(true); |
1757
|
|
|
} |
1758
|
|
|
} |
1759
|
|
|
|
1760
|
1 |
|
$c->reachesIntoOuterContext++; |
1761
|
|
|
|
1762
|
1 |
|
if (!$closureBusy->add($c)) { |
1763
|
|
|
// avoid infinite recursion for right-recursive rules |
1764
|
|
|
|
1765
|
1 |
|
continue; |
1766
|
|
|
} |
1767
|
|
|
|
1768
|
|
|
// TODO: can remove? only care when we add to set per middle of this method |
1769
|
1 |
|
$configs->dipsIntoOuterContext = true; |
1770
|
1 |
|
$newDepth--; |
1771
|
|
|
|
1772
|
1 |
|
if (self::$debug) { |
1773
|
1 |
|
$this->log[] = \sprintf('dips into outer ctx: %s', $c); |
1774
|
|
|
} |
1775
|
|
|
} else { |
1776
|
1 |
|
if (!$t->isEpsilon() && !$closureBusy->add($c)) { |
1777
|
|
|
// avoid infinite recursion for EOF* and EOF+ |
1778
|
|
|
|
1779
|
|
|
continue; |
1780
|
|
|
} |
1781
|
|
|
|
1782
|
1 |
|
if ($t instanceof RuleTransition) { |
1783
|
|
|
// latch when newDepth goes negative - once we step out of the entry context we can't return |
1784
|
|
|
|
1785
|
1 |
|
if ($newDepth >= 0) { |
1786
|
1 |
|
$newDepth++; |
1787
|
|
|
} |
1788
|
|
|
} |
1789
|
|
|
} |
1790
|
|
|
|
1791
|
1 |
|
$this->closureCheckingStopState( |
1792
|
1 |
|
$c, |
1793
|
|
|
$configs, |
1794
|
|
|
$closureBusy, |
1795
|
|
|
$continueCollecting, |
1796
|
|
|
$fullCtx, |
1797
|
|
|
$newDepth, |
1798
|
|
|
$treatEofAsEpsilon |
1799
|
|
|
); |
1800
|
|
|
} |
1801
|
|
|
} |
1802
|
1 |
|
} |
1803
|
|
|
|
1804
|
|
|
/** |
1805
|
|
|
* Implements first-edge (loop entry) elimination as an optimization |
1806
|
|
|
* during closure operations. See antlr/antlr4#1398. |
1807
|
|
|
* |
1808
|
|
|
* The optimization is to avoid adding the loop entry config when |
1809
|
|
|
* the exit path can only lead back to the same |
1810
|
|
|
* StarLoopEntryState after popping context at the rule end state |
1811
|
|
|
* (traversing only epsilon edges, so we're still in closure, in |
1812
|
|
|
* this same rule). |
1813
|
|
|
* |
1814
|
|
|
* We need to detect any state that can reach loop entry on |
1815
|
|
|
* epsilon w/o exiting rule. We don't have to look at FOLLOW |
1816
|
|
|
* links, just ensure that all stack tops for config refer to key |
1817
|
|
|
* states in LR rule. |
1818
|
|
|
* |
1819
|
|
|
* To verify we are in the right situation we must first check |
1820
|
|
|
* closure is at a StarLoopEntryState generated during LR removal. |
1821
|
|
|
* Then we check that each stack top of context is a return state |
1822
|
|
|
* from one of these cases: |
1823
|
|
|
* |
1824
|
|
|
* 1. 'not' expr, '(' type ')' expr. The return state points at loop entry state |
1825
|
|
|
* 2. expr op expr. The return state is the block end of internal block of (...)* |
1826
|
|
|
* 3. 'between' expr 'and' expr. The return state of 2nd expr reference. |
1827
|
|
|
* That state points at block end of internal block of (...)*. |
1828
|
|
|
* 4. expr '?' expr ':' expr. The return state points at block end, |
1829
|
|
|
* which points at loop entry state. |
1830
|
|
|
* |
1831
|
|
|
* If any is true for each stack top, then closure does not add a |
1832
|
|
|
* config to the current config set for edge[0], the loop entry branch. |
1833
|
|
|
* |
1834
|
|
|
* Conditions fail if any context for the current config is: |
1835
|
|
|
* |
1836
|
|
|
* a. empty (we'd fall out of expr to do a global FOLLOW which could |
1837
|
|
|
* even be to some weird spot in expr) or, |
1838
|
|
|
* b. lies outside of expr or, |
1839
|
|
|
* c. lies within expr but at a state not the BlockEndState |
1840
|
|
|
* generated during LR removal |
1841
|
|
|
* |
1842
|
|
|
* Do we need to evaluate predicates ever in closure for this case? |
1843
|
|
|
* |
1844
|
|
|
* No. Predicates, including precedence predicates, are only |
1845
|
|
|
* evaluated when computing a DFA start state. I.e., only before |
1846
|
|
|
* the lookahead (but not parser) consumes a token. |
1847
|
|
|
* |
1848
|
|
|
* There are no epsilon edges allowed in LR rule alt blocks or in |
1849
|
|
|
* the "primary" part (ID here). If closure is in |
1850
|
|
|
* StarLoopEntryState any lookahead operation will have consumed a |
1851
|
|
|
* token as there are no epsilon-paths that lead to |
1852
|
|
|
* StarLoopEntryState. We do not have to evaluate predicates |
1853
|
|
|
* therefore if we are in the generated StarLoopEntryState of a LR |
1854
|
|
|
* rule. Note that when making a prediction starting at that |
1855
|
|
|
* decision point, decision d=2, compute-start-state performs |
1856
|
|
|
* closure starting at edges[0], edges[1] emanating from |
1857
|
|
|
* StarLoopEntryState. That means it is not performing closure on |
1858
|
|
|
* StarLoopEntryState during compute-start-state. |
1859
|
|
|
* |
1860
|
|
|
* How do we know this always gives same prediction answer? |
1861
|
|
|
* |
1862
|
|
|
* Without predicates, loop entry and exit paths are ambiguous |
1863
|
|
|
* upon remaining input +b (in, say, a+b). Either paths lead to |
1864
|
|
|
* valid parses. Closure can lead to consuming + immediately or by |
1865
|
|
|
* falling out of this call to expr back into expr and loop back |
1866
|
|
|
* again to StarLoopEntryState to match +b. In this special case, |
1867
|
|
|
* we choose the more efficient path, which is to take the bypass |
1868
|
|
|
* path. |
1869
|
|
|
* |
1870
|
|
|
* The lookahead language has not changed because closure chooses |
1871
|
|
|
* one path over the other. Both paths lead to consuming the same |
1872
|
|
|
* remaining input during a lookahead operation. If the next token |
1873
|
|
|
* is an operator, lookahead will enter the choice block with |
1874
|
|
|
* operators. If it is not, lookahead will exit expr. Same as if |
1875
|
|
|
* closure had chosen to enter the choice block immediately. |
1876
|
|
|
* |
1877
|
|
|
* Closure is examining one config (some loopentrystate, some alt, |
1878
|
|
|
* context) which means it is considering exactly one alt. Closure |
1879
|
|
|
* always copies the same alt to any derived configs. |
1880
|
|
|
* |
1881
|
|
|
* How do we know this optimization doesn't mess up precedence in |
1882
|
|
|
* our parse trees? |
1883
|
|
|
* |
1884
|
|
|
* Looking through expr from left edge of stat only has to confirm |
1885
|
|
|
* that an input, say, a+b+c; begins with any valid interpretation |
1886
|
|
|
* of an expression. The precedence actually doesn't matter when |
1887
|
|
|
* making a decision in stat seeing through expr. It is only when |
1888
|
|
|
* parsing rule expr that we must use the precedence to get the |
1889
|
|
|
* right interpretation and, hence, parse tree. |
1890
|
|
|
* |
1891
|
|
|
* @since 4.6 |
1892
|
|
|
*/ |
1893
|
1 |
|
protected function canDropLoopEntryEdgeInLeftRecursiveRule(ATNConfig $config) : bool |
1894
|
|
|
{ |
1895
|
1 |
|
$p = $config->state; |
1896
|
|
|
|
1897
|
|
|
/* First check to see if we are in StarLoopEntryState generated during |
1898
|
|
|
* left-recursion elimination. For efficiency, also check if |
1899
|
|
|
* the context has an empty stack case. If so, it would mean |
1900
|
|
|
* global FOLLOW so we can't perform optimization |
1901
|
|
|
* Are we the special loop entry/exit state? or SLL wildcard |
1902
|
|
|
*/ |
1903
|
|
|
|
1904
|
1 |
|
if ($config->context === null) { |
1905
|
|
|
throw new \RuntimeException('Prediction context cannot be null.'); |
1906
|
|
|
} |
1907
|
|
|
|
1908
|
1 |
|
if ($p->getStateType() !== ATNState::STAR_LOOP_ENTRY |
1909
|
1 |
|
|| ($p instanceof StarLoopEntryState && !$p->isPrecedenceDecision) |
1910
|
1 |
|
|| $config->context->isEmpty() |
1911
|
1 |
|
|| $config->context->hasEmptyPath()) { |
1912
|
1 |
|
return false; |
1913
|
|
|
} |
1914
|
|
|
|
1915
|
|
|
// Require all return states to return back to the same rule that p is in. |
1916
|
1 |
|
$numCtxs = $config->context->getLength(); |
1917
|
|
|
|
1918
|
1 |
|
for ($i = 0; $i < $numCtxs; $i++) { |
1919
|
|
|
// For each stack context |
1920
|
1 |
|
$returnState = $this->atn->states[$config->context->getReturnState($i)]; |
1921
|
|
|
|
1922
|
1 |
|
if ($returnState->ruleIndex !== $p->ruleIndex) { |
1923
|
1 |
|
return false; |
1924
|
|
|
} |
1925
|
|
|
} |
1926
|
|
|
|
1927
|
|
|
$decisionStartState = $p->getTransition(0)->target; |
1928
|
|
|
|
1929
|
|
|
if (!$decisionStartState instanceof BlockStartState || $decisionStartState->endState === null) { |
1930
|
|
|
throw new \RuntimeException('Unexpected transition type.'); |
1931
|
|
|
} |
1932
|
|
|
|
1933
|
|
|
$blockEndStateNum = $decisionStartState->endState->stateNumber; |
1934
|
|
|
$blockEndState = $this->atn->states[$blockEndStateNum]; |
1935
|
|
|
|
1936
|
|
|
if (!$blockEndState instanceof BlockEndState) { |
1937
|
|
|
throw new \RuntimeException('Unexpected transition type.'); |
1938
|
|
|
} |
1939
|
|
|
|
1940
|
|
|
// Verify that the top of each stack context leads to loop entry/exit |
1941
|
|
|
// state through epsilon edges and w/o leaving rule. |
1942
|
|
|
for ($i = 0; $i < $numCtxs; $i++) { |
1943
|
|
|
// For each stack context |
1944
|
|
|
|
1945
|
|
|
$returnStateNumber = $config->context->getReturnState($i); |
1946
|
|
|
$returnState = $this->atn->states[$returnStateNumber]; |
1947
|
|
|
|
1948
|
|
|
// All states must have single outgoing epsilon edge |
1949
|
|
|
if ($returnState->getNumberOfTransitions() !== 1 || !$returnState->getTransition(0)->isEpsilon()) { |
1950
|
|
|
return false; |
1951
|
|
|
} |
1952
|
|
|
|
1953
|
|
|
// Look for prefix op case like 'not expr', (' type ')' expr |
1954
|
|
|
$returnStateTarget = $returnState->getTransition(0)->target; |
1955
|
|
|
|
1956
|
|
|
if ($returnState->getStateType() === ATNState::BLOCK_END && $returnStateTarget->equals($p)) { |
1957
|
|
|
continue; |
1958
|
|
|
} |
1959
|
|
|
|
1960
|
|
|
// Look for 'expr op expr' or case where expr's return state is block end |
1961
|
|
|
// of (...)* internal block; the block end points to loop back |
1962
|
|
|
// which points to p but we don't need to check that |
1963
|
|
|
if ($returnState->equals($blockEndState)) { |
1964
|
|
|
continue; |
1965
|
|
|
} |
1966
|
|
|
|
1967
|
|
|
// Look for ternary expr ? expr : expr. The return state points at block end, |
1968
|
|
|
// which points at loop entry state |
1969
|
|
|
if ($returnStateTarget->equals($blockEndState)) { |
1970
|
|
|
continue; |
1971
|
|
|
} |
1972
|
|
|
|
1973
|
|
|
// Look for complex prefix 'between expr and expr' case where 2nd expr's |
1974
|
|
|
// return state points at block end state of (...)* internal block |
1975
|
|
|
if ($returnStateTarget->getStateType() === ATNState::BLOCK_END |
1976
|
|
|
&& $returnStateTarget->getNumberOfTransitions() === 1 |
1977
|
|
|
&& $returnStateTarget->getTransition(0)->isEpsilon() |
1978
|
|
|
&& $returnStateTarget->getTransition(0)->target->equals($p)) { |
1979
|
|
|
continue; |
1980
|
|
|
} |
1981
|
|
|
|
1982
|
|
|
// anything else ain't conforming |
1983
|
|
|
return false; |
1984
|
|
|
} |
1985
|
|
|
|
1986
|
|
|
return true; |
1987
|
|
|
} |
1988
|
|
|
|
1989
|
|
|
public function getRuleName(int $index) : string |
1990
|
|
|
{ |
1991
|
|
|
if ($this->parser !== null && $index >= 0) { |
1992
|
|
|
return $this->parser->getRuleNames()[$index]; |
1993
|
|
|
} |
1994
|
|
|
|
1995
|
|
|
return '<rule $index>'; |
1996
|
|
|
} |
1997
|
|
|
|
1998
|
1 |
|
protected function getEpsilonTarget( |
1999
|
|
|
ATNConfig $config, |
2000
|
|
|
Transition $t, |
2001
|
|
|
bool $collectPredicates, |
2002
|
|
|
bool $inContext, |
2003
|
|
|
bool $fullCtx, |
2004
|
|
|
bool $treatEofAsEpsilon |
2005
|
|
|
) : ?ATNConfig { |
2006
|
1 |
|
switch ($t->getSerializationType()) { |
2007
|
|
|
case Transition::RULE: |
2008
|
1 |
|
if (!$t instanceof RuleTransition) { |
2009
|
|
|
throw new \RuntimeException('Unexpected transition type.'); |
2010
|
|
|
} |
2011
|
|
|
|
2012
|
1 |
|
return $this->ruleTransition($config, $t); |
2013
|
|
|
|
2014
|
|
|
case Transition::PRECEDENCE: |
2015
|
1 |
|
if (!$t instanceof PrecedencePredicateTransition) { |
2016
|
|
|
throw new \RuntimeException('Unexpected transition type.'); |
2017
|
|
|
} |
2018
|
|
|
|
2019
|
1 |
|
return $this->precedenceTransition($config, $t, $collectPredicates, $inContext, $fullCtx); |
2020
|
|
|
|
2021
|
|
|
case Transition::PREDICATE: |
2022
|
|
|
if (!$t instanceof PredicateTransition) { |
2023
|
|
|
throw new \RuntimeException('Unexpected transition type.'); |
2024
|
|
|
} |
2025
|
|
|
|
2026
|
|
|
return $this->predTransition($config, $t, $collectPredicates, $inContext, $fullCtx); |
2027
|
|
|
|
2028
|
|
|
case Transition::ACTION: |
2029
|
1 |
|
if (!$t instanceof ActionTransition) { |
2030
|
|
|
throw new \RuntimeException('Unexpected transition type.'); |
2031
|
|
|
} |
2032
|
|
|
|
2033
|
1 |
|
return $this->actionTransition($config, $t); |
2034
|
|
|
|
2035
|
|
|
case Transition::EPSILON: |
2036
|
1 |
|
return new ATNConfig($config, $t->target); |
2037
|
|
|
|
2038
|
|
|
case Transition::ATOM: |
2039
|
|
|
case Transition::RANGE: |
2040
|
|
|
case Transition::SET: |
2041
|
|
|
// EOF transitions act like epsilon transitions after the first EOF transition is traversed |
2042
|
|
|
|
2043
|
1 |
|
if ($treatEofAsEpsilon) { |
2044
|
|
|
if ($t->matches(Token::EOF, 0, 1)) { |
2045
|
|
|
return new ATNConfig($config, $t->target); |
2046
|
|
|
} |
2047
|
|
|
} |
2048
|
|
|
|
2049
|
1 |
|
return null; |
2050
|
|
|
|
2051
|
|
|
default: |
2052
|
|
|
return null; |
2053
|
|
|
} |
2054
|
|
|
} |
2055
|
|
|
|
2056
|
1 |
|
protected function actionTransition(ATNConfig $config, ActionTransition $t) : ATNConfig |
2057
|
|
|
{ |
2058
|
1 |
|
if (self::$debug) { |
2059
|
|
|
$this->log[] = \sprintf('ACTION edge %d:%d', $t->ruleIndex, $t->actionIndex); |
2060
|
|
|
} |
2061
|
|
|
|
2062
|
1 |
|
return new ATNConfig($config, $t->target); |
2063
|
|
|
} |
2064
|
|
|
|
2065
|
1 |
|
public function precedenceTransition( |
2066
|
|
|
ATNConfig $config, |
2067
|
|
|
PrecedencePredicateTransition $pt, |
2068
|
|
|
bool $collectPredicates, |
2069
|
|
|
bool $inContext, |
2070
|
|
|
bool $fullCtx |
2071
|
|
|
) : ?ATNConfig { |
2072
|
1 |
|
if (self::$debug) { |
2073
|
|
|
$this->log[] = \sprintf( |
2074
|
|
|
'PRED (collectPredicates=%s) %d>=_p, ctx dependent=true', |
2075
|
|
|
$collectPredicates, |
2076
|
|
|
$pt->precedence |
2077
|
|
|
); |
2078
|
|
|
|
2079
|
|
|
if ($this->parser !== null) { |
2080
|
|
|
$this->log[] = \sprintf( |
2081
|
|
|
'context surrounding pred is [%s]', |
2082
|
|
|
\implode(', ', $this->parser->getRuleInvocationStack()) |
2083
|
|
|
); |
2084
|
|
|
} |
2085
|
|
|
} |
2086
|
|
|
|
2087
|
1 |
|
$c = null; |
2088
|
|
|
|
2089
|
1 |
|
if ($collectPredicates && $inContext) { |
2090
|
1 |
|
if ($fullCtx) { |
2091
|
|
|
/* In full context mode, we can evaluate predicates on-the-fly |
2092
|
|
|
* during closure, which dramatically reduces the size of |
2093
|
|
|
* the config sets. It also obviates the need to test predicates |
2094
|
|
|
* later during conflict resolution. |
2095
|
|
|
*/ |
2096
|
|
|
|
2097
|
|
|
$currentPosition = $this->input->getIndex(); |
2098
|
|
|
|
2099
|
|
|
$this->input->seek($this->startIndex); |
2100
|
|
|
|
2101
|
|
|
$predSucceeds = $this->outerContext !== null ? |
2102
|
|
|
$pt->getPredicate()->eval($this->parser, $this->outerContext) : |
2103
|
|
|
false; |
2104
|
|
|
|
2105
|
|
|
$this->input->seek($currentPosition); |
2106
|
|
|
|
2107
|
|
|
if ($predSucceeds) { |
2108
|
|
|
$c = new ATNConfig($config, $pt->target);// no pred context |
2109
|
|
|
} |
2110
|
|
|
} else { |
2111
|
1 |
|
$newSemCtx = SemanticContext::andContext($config->semanticContext, $pt->getPredicate()); |
2112
|
1 |
|
$c = new ATNConfig($config, $pt->target, null, $newSemCtx); |
2113
|
|
|
} |
2114
|
|
|
} else { |
2115
|
1 |
|
$c = new ATNConfig($config, $pt->target); |
2116
|
1 |
|
} if (self::$debug) { |
2117
|
|
|
$this->log[] = \sprintf('Config from pred transition=%s', (string) $c); |
2118
|
|
|
} |
2119
|
|
|
|
2120
|
1 |
|
return $c; |
2121
|
|
|
} |
2122
|
|
|
|
2123
|
|
|
protected function predTransition( |
2124
|
|
|
ATNConfig $config, |
2125
|
|
|
PredicateTransition $pt, |
2126
|
|
|
bool $collectPredicates, |
2127
|
|
|
bool $inContext, |
2128
|
|
|
bool $fullCtx |
2129
|
|
|
) : ?ATNConfig { |
2130
|
|
|
if (self::$debug) { |
2131
|
|
|
$this->log[] = \sprintf( |
2132
|
|
|
'PRED (collectPredicates=%s) %d:%d, ctx dependent=%s', |
2133
|
|
|
$collectPredicates, |
2134
|
|
|
$pt->ruleIndex, |
2135
|
|
|
$pt->predIndex, |
2136
|
|
|
$pt->isCtxDependent |
2137
|
|
|
); |
2138
|
|
|
|
2139
|
|
|
if ($this->parser !== null) { |
2140
|
|
|
$this->log[] = \sprintf( |
2141
|
|
|
'Context surrounding pred is [%s]', |
2142
|
|
|
\implode(', ', $this->parser->getRuleInvocationStack()) |
2143
|
|
|
); |
2144
|
|
|
} |
2145
|
|
|
} |
2146
|
|
|
|
2147
|
|
|
$c = null; |
2148
|
|
|
|
2149
|
|
|
if ($collectPredicates && (!$pt->isCtxDependent || $inContext)) { |
2150
|
|
|
if ($fullCtx) { |
2151
|
|
|
// In full context mode, we can evaluate predicates on-the-fly |
2152
|
|
|
// during closure, which dramatically reduces the size of |
2153
|
|
|
// the config sets. It also obviates the need to test predicates |
2154
|
|
|
// later during conflict resolution. |
2155
|
|
|
|
2156
|
|
|
$currentPosition = $this->input->getIndex(); |
2157
|
|
|
|
2158
|
|
|
$this->input->seek($this->startIndex); |
2159
|
|
|
|
2160
|
|
|
$predSucceeds = $this->outerContext !== null ? |
2161
|
|
|
$pt->getPredicate()->eval($this->parser, $this->outerContext) : |
2162
|
|
|
false; |
2163
|
|
|
|
2164
|
|
|
$this->input->seek($currentPosition); |
2165
|
|
|
|
2166
|
|
|
if ($predSucceeds) { |
2167
|
|
|
$c = new ATNConfig($config, $pt->target);// no pred context |
2168
|
|
|
} |
2169
|
|
|
} else { |
2170
|
|
|
$newSemCtx = SemanticContext::andContext($config->semanticContext, $pt->getPredicate()); |
2171
|
|
|
$c = new ATNConfig($config, $pt->target, null, $newSemCtx); |
2172
|
|
|
} |
2173
|
|
|
} else { |
2174
|
|
|
$c = new ATNConfig($config, $pt->target); |
2175
|
|
|
} |
2176
|
|
|
|
2177
|
|
|
if (self::$debug) { |
2178
|
|
|
$this->log[] = \sprintf('Config from pred transition=%s', (string) $c); |
2179
|
|
|
} |
2180
|
|
|
|
2181
|
|
|
return $c; |
2182
|
|
|
} |
2183
|
|
|
|
2184
|
1 |
|
protected function ruleTransition(ATNConfig $config, RuleTransition $t) : ATNConfig |
2185
|
|
|
{ |
2186
|
1 |
|
if (self::$debug) { |
2187
|
|
|
$this->log[] = \sprintf( |
2188
|
|
|
'CALL rule %s, ctx=%s', |
2189
|
|
|
$this->getRuleName($t->target->ruleIndex), |
2190
|
|
|
$config->context |
2191
|
|
|
); |
2192
|
|
|
} |
2193
|
|
|
|
2194
|
1 |
|
$returnState = $t->followState; |
2195
|
1 |
|
$newContext = SingletonPredictionContext::create($config->context, $returnState->stateNumber); |
2196
|
|
|
|
2197
|
1 |
|
return new ATNConfig($config, $t->target, $newContext); |
2198
|
|
|
} |
2199
|
|
|
|
2200
|
|
|
/** |
2201
|
|
|
* Gets a {@see BitSet} containing the alternatives in `configs` |
2202
|
|
|
* which are part of one or more conflicting alternative subsets. |
2203
|
|
|
* |
2204
|
|
|
* @param ATNConfigSet $configs The {@see ATNConfigSet} to analyze. |
2205
|
|
|
* |
2206
|
|
|
* @return BitSet The alternatives in `configs` which are part of one or |
2207
|
|
|
* more conflicting alternative subsets. If `configs` does |
2208
|
|
|
* not contain any conflicting subsets, this method returns |
2209
|
|
|
* an empty {@see BitSet}. |
2210
|
|
|
*/ |
2211
|
|
|
protected function getConflictingAlts(ATNConfigSet $configs) : BitSet |
2212
|
|
|
{ |
2213
|
|
|
$altsets = PredictionMode::getConflictingAltSubsets($configs); |
2214
|
|
|
|
2215
|
|
|
return PredictionMode::getAlts($altsets); |
2216
|
|
|
} |
2217
|
|
|
|
2218
|
|
|
/** |
2219
|
|
|
Sam pointed out a problem with the previous definition, v3, of |
2220
|
|
|
ambiguous states. If we have another state associated with conflicting |
2221
|
|
|
alternatives, we should keep going. For example, the following grammar |
2222
|
|
|
|
2223
|
|
|
s : (ID | ID ID?) ';' ; |
2224
|
|
|
|
2225
|
|
|
When the ATN simulation reaches the state before ';', it has a DFA |
2226
|
|
|
state that looks like: [12|1|[], 6|2|[], 12|2|[]]. Naturally |
2227
|
|
|
12|1|[] and 12|2|[] conflict, but we cannot stop processing this node |
2228
|
|
|
because alternative to has another way to continue, via [6|2|[]]. |
2229
|
|
|
The key is that we have a single state that has config's only associated |
2230
|
|
|
with a single alternative, 2, and crucially the state transitions |
2231
|
|
|
among the configurations are all non-epsilon transitions. That means |
2232
|
|
|
we don't consider any conflicts that include alternative 2. So, we |
2233
|
|
|
ignore the conflict between alts 1 and 2. We ignore a set of |
2234
|
|
|
conflicting alts when there is an intersection with an alternative |
2235
|
|
|
associated with a single alt state in the state→config-list map. |
2236
|
|
|
|
2237
|
|
|
It's also the case that we might have two conflicting configurations but |
2238
|
|
|
also a 3rd nonconflicting configuration for a different alternative: |
2239
|
|
|
[1|1|[], 1|2|[], 8|3|[]]. This can come about from grammar: |
2240
|
|
|
|
2241
|
|
|
a : A | A | A B ; |
2242
|
|
|
|
2243
|
|
|
After matching input A, we reach the stop state for rule A, state 1. |
2244
|
|
|
State 8 is the state right before B. Clearly alternatives 1 and 2 |
2245
|
|
|
conflict and no amount of further lookahead will separate the two. |
2246
|
|
|
However, alternative 3 will be able to continue and so we do not |
2247
|
|
|
stop working on this state. In the previous example, we're concerned |
2248
|
|
|
with states associated with the conflicting alternatives. Here alt |
2249
|
|
|
3 is not associated with the conflicting configs, but since we can continue |
2250
|
|
|
looking for input reasonably, I don't declare the state done. We |
2251
|
|
|
ignore a set of conflicting alts when we have an alternative |
2252
|
|
|
that we still need to pursue. |
2253
|
|
|
*/ |
2254
|
2 |
|
protected function getConflictingAltsOrUniqueAlt(ATNConfigSet $configs) : ?BitSet |
2255
|
|
|
{ |
2256
|
2 |
|
if ($configs->uniqueAlt !== ATN::INVALID_ALT_NUMBER) { |
2257
|
2 |
|
$conflictingAlts = new BitSet(); |
2258
|
|
|
|
2259
|
2 |
|
$conflictingAlts->add($configs->uniqueAlt); |
2260
|
|
|
|
2261
|
2 |
|
return $conflictingAlts; |
2262
|
|
|
} |
2263
|
|
|
|
2264
|
|
|
return $configs->getConflictingAlts(); |
2265
|
|
|
} |
2266
|
|
|
|
2267
|
|
|
public function getTokenName(int $t) : string |
2268
|
|
|
{ |
2269
|
|
|
if ($t === Token::EOF) { |
2270
|
|
|
return 'EOF'; |
2271
|
|
|
} |
2272
|
|
|
|
2273
|
|
|
$vocabulary = $this->parser !== null ? $this->parser->getVocabulary() : VocabularyImpl::emptyVocabulary(); |
2274
|
|
|
$displayName = $vocabulary->getDisplayName($t); |
2275
|
|
|
|
2276
|
|
|
if ($displayName === (string) $t) { |
2277
|
|
|
return $displayName; |
2278
|
|
|
} |
2279
|
|
|
|
2280
|
|
|
return \sprintf('%s<%d>', $displayName, $t); |
2281
|
|
|
} |
2282
|
|
|
|
2283
|
|
|
public function getLookaheadName(TokenStream $input) : string |
2284
|
|
|
{ |
2285
|
|
|
return $this->getTokenName($input->LA(1)); |
2286
|
|
|
} |
2287
|
|
|
|
2288
|
4 |
|
protected function noViableAlt( |
2289
|
|
|
TokenStream $input, |
2290
|
|
|
$outerContext, |
2291
|
|
|
?ATNConfigSet $configs, |
2292
|
|
|
int $startIndex |
2293
|
|
|
) : NoViableAltException { |
2294
|
4 |
|
return new NoViableAltException( |
2295
|
4 |
|
$this->parser, |
2296
|
|
|
$input, |
2297
|
4 |
|
$input->get($startIndex), |
2298
|
4 |
|
$input->LT(1), |
2299
|
|
|
$configs, |
2300
|
|
|
$outerContext |
2301
|
|
|
); |
2302
|
|
|
} |
2303
|
|
|
|
2304
|
3 |
|
protected static function getUniqueAlt(ATNConfigSet $configs) : int |
2305
|
|
|
{ |
2306
|
3 |
|
$alt = ATN::INVALID_ALT_NUMBER; |
2307
|
|
|
|
2308
|
3 |
|
foreach ($configs->elements() as $c) { |
2309
|
3 |
|
if ($alt === ATN::INVALID_ALT_NUMBER) { |
2310
|
3 |
|
$alt = $c->alt; // found first alt |
2311
|
2 |
|
} elseif ($c->alt !== $alt) { |
2312
|
1 |
|
return ATN::INVALID_ALT_NUMBER; |
2313
|
|
|
} |
2314
|
|
|
} |
2315
|
|
|
|
2316
|
3 |
|
return $alt; |
2317
|
|
|
} |
2318
|
|
|
|
2319
|
|
|
/** |
2320
|
|
|
* Add an edge to the DFA, if possible. This method calls |
2321
|
|
|
* {@see ParserATNSimulator::addDFAState()} to ensure the `to` state is |
2322
|
|
|
* present in the DFA. If `from` is `null`, or if `t` is outside the |
2323
|
|
|
* range of edges that can be represented in the DFA tables, this method |
2324
|
|
|
* returns without adding the edge to the DFA. |
2325
|
|
|
* |
2326
|
|
|
* If `to` is `null`, this method returns `null`. Otherwise, this method |
2327
|
|
|
* returns the {@see DFAState} returned by calling |
2328
|
|
|
* {@see ParserATNSimulator::addDFAState()} for the `to` state. |
2329
|
|
|
* |
2330
|
|
|
* @param DFA $dfa The DFA |
2331
|
|
|
* @param DFAState|null $from The source state for the edge |
2332
|
|
|
* @param int $t The input symbol |
2333
|
|
|
* @param DFAState|null $to The target state for the edge |
2334
|
|
|
* |
2335
|
|
|
* @return DFAState If `to` is `null` this method returns `null`, |
2336
|
|
|
* otherwise this method returns the result of calling |
2337
|
|
|
* {@see ParserATNSimulator::addDFAState()} on `to`. |
2338
|
|
|
*/ |
2339
|
3 |
|
protected function addDFAEdge(DFA $dfa, ?DFAState $from, int $t, ?DFAState $to) : ?DFAState |
2340
|
|
|
{ |
2341
|
3 |
|
if (self::$debug) { |
2342
|
|
|
$this->log[] = \sprintf('EDGE %s -> %s upon %s', (string) $from, (string) $to, $this->getTokenName($t)); |
2343
|
|
|
} |
2344
|
|
|
|
2345
|
3 |
|
if ($to === null) { |
2346
|
|
|
return null; |
2347
|
|
|
} |
2348
|
|
|
|
2349
|
3 |
|
$to = $this->addDFAState($dfa, $to);// used existing if possible not incoming |
2350
|
|
|
|
2351
|
3 |
|
if ($from === null || $t < -1 || $t > $this->atn->maxTokenType) { |
2352
|
|
|
return $to; |
2353
|
|
|
} |
2354
|
|
|
|
2355
|
3 |
|
if ($from->edges === null) { |
2356
|
1 |
|
$from->edges = new \SplFixedArray($this->atn->maxTokenType + 1 + 1); |
2357
|
|
|
} |
2358
|
|
|
|
2359
|
3 |
|
$from->edges[$t + 1] = $to; |
2360
|
|
|
|
2361
|
3 |
|
if (self::$debug) { |
2362
|
|
|
$this->log[] = 'DFA =' . \PHP_EOL . $dfa->toString($this->parser->getVocabulary()); |
2363
|
|
|
} |
2364
|
|
|
|
2365
|
3 |
|
return $to; |
2366
|
|
|
} |
2367
|
|
|
|
2368
|
|
|
/** |
2369
|
|
|
* Add state `D` to the DFA if it is not already present, and return |
2370
|
|
|
* the actual instance stored in the DFA. If a state equivalent to `D` |
2371
|
|
|
* is already in the DFA, the existing state is returned. Otherwise this |
2372
|
|
|
* method returns `D` after adding it to the DFA. |
2373
|
|
|
* |
2374
|
|
|
* If `D` is {@see ParserATNSimulator::error()}, this method returns |
2375
|
|
|
* {@see ParserATNSimulator::error()} and does not change the DFA. |
2376
|
|
|
* |
2377
|
|
|
* @param DFA $dfa The dfa |
2378
|
|
|
* @param DFAState $D The DFA state to add |
2379
|
|
|
* |
2380
|
|
|
* @return DFAState The state stored in the DFA. This will be either |
2381
|
|
|
* the existing state if `D` is already in the DFA, or `D` |
2382
|
|
|
* itself if the state was not already present. |
2383
|
|
|
* |
2384
|
|
|
* @throws \InvalidArgumentException |
2385
|
|
|
*/ |
2386
|
3 |
|
protected function addDFAState(DFA $dfa, DFAState $D) : DFAState |
2387
|
|
|
{ |
2388
|
3 |
|
if ($D === self::error()) { |
2389
|
1 |
|
return $D; |
2390
|
|
|
} |
2391
|
|
|
|
2392
|
3 |
|
$existing = $dfa->states->get($D); |
2393
|
|
|
|
2394
|
3 |
|
if ($existing !== null && $existing instanceof DFAState) { |
2395
|
1 |
|
return $existing; |
2396
|
|
|
} |
2397
|
|
|
|
2398
|
3 |
|
$D->stateNumber = $dfa->states->count(); |
2399
|
|
|
|
2400
|
3 |
|
if (!$D->configs->isReadOnly()) { |
2401
|
3 |
|
$D->configs->optimizeConfigs($this); |
2402
|
3 |
|
$D->configs->setReadonly(true); |
2403
|
|
|
} |
2404
|
|
|
|
2405
|
3 |
|
$dfa->states->add($D); |
2406
|
|
|
|
2407
|
3 |
|
if (self::$debug) { |
2408
|
|
|
$this->log[] = \sprintf('Adding new DFA state: %s', (string) $D); |
2409
|
|
|
} |
2410
|
|
|
|
2411
|
3 |
|
return $D; |
2412
|
|
|
} |
2413
|
|
|
|
2414
|
|
|
protected function reportAttemptingFullContext( |
2415
|
|
|
DFA $dfa, |
2416
|
|
|
?BitSet $conflictingAlts, |
2417
|
|
|
ATNConfigSet $configs, |
2418
|
|
|
int $startIndex, |
2419
|
|
|
int $stopIndex |
2420
|
|
|
) : void { |
2421
|
|
|
if (self::$debug || self::$retry_debug) { |
2422
|
|
|
$interval = new Interval($startIndex, $stopIndex); |
2423
|
|
|
$tokenStream = $this->parser->getTokenStream(); |
2424
|
|
|
|
2425
|
|
|
$this->log[] = \sprintf( |
2426
|
|
|
'reportAttemptingFullContext decision = %d:%s, input = %s', |
2427
|
|
|
$dfa->decision, |
2428
|
|
|
(string) $configs, |
2429
|
|
|
$tokenStream === null ? '' : $tokenStream->getTextByInterval($interval) |
2430
|
|
|
); |
2431
|
|
|
} |
2432
|
|
|
|
2433
|
|
|
if ($this->parser !== null) { |
2434
|
|
|
$this->parser->getErrorListenerDispatch()->reportAttemptingFullContext( |
2435
|
|
|
$this->parser, |
2436
|
|
|
$dfa, |
2437
|
|
|
$startIndex, |
2438
|
|
|
$stopIndex, |
2439
|
|
|
$conflictingAlts, |
2440
|
|
|
$configs |
2441
|
|
|
); |
2442
|
|
|
} |
2443
|
|
|
} |
2444
|
|
|
|
2445
|
|
|
protected function reportContextSensitivity( |
2446
|
|
|
DFA $dfa, |
2447
|
|
|
int $prediction, |
2448
|
|
|
ATNConfigSet $configs, |
2449
|
|
|
int $startIndex, |
2450
|
|
|
int $stopIndex |
2451
|
|
|
) : void { |
2452
|
|
|
if (self::$debug || self::$retry_debug) { |
2453
|
|
|
$interval = new Interval($startIndex, $stopIndex); |
2454
|
|
|
$tokenStream = $this->parser->getTokenStream(); |
2455
|
|
|
|
2456
|
|
|
$this->log[] = \sprintf( |
2457
|
|
|
'reportContextSensitivity decision = %d:%s, input = %s', |
2458
|
|
|
$dfa->decision, |
2459
|
|
|
(string) $configs, |
2460
|
|
|
$tokenStream === null ? '' : $tokenStream->getTextByInterval($interval) |
2461
|
|
|
); |
2462
|
|
|
} |
2463
|
|
|
|
2464
|
|
|
if ($this->parser !== null) { |
2465
|
|
|
$this->parser->getErrorListenerDispatch()->reportContextSensitivity( |
2466
|
|
|
$this->parser, |
2467
|
|
|
$dfa, |
2468
|
|
|
$startIndex, |
2469
|
|
|
$stopIndex, |
2470
|
|
|
$prediction, |
2471
|
|
|
$configs |
2472
|
|
|
); |
2473
|
|
|
} |
2474
|
|
|
} |
2475
|
|
|
|
2476
|
|
|
/** |
2477
|
|
|
* If context sensitive parsing, we know it's ambiguity not conflict. |
2478
|
|
|
*/ |
2479
|
|
|
protected function reportAmbiguity( |
2480
|
|
|
DFA $dfa, |
2481
|
|
|
DFAState $D, |
|
|
|
|
2482
|
|
|
int $startIndex, |
2483
|
|
|
int $stopIndex, |
2484
|
|
|
bool $exact, |
2485
|
|
|
?BitSet $ambigAlts, |
2486
|
|
|
ATNConfigSet $configs |
2487
|
|
|
) : void { |
2488
|
|
|
if (self::$debug || self::$retry_debug) { |
2489
|
|
|
$interval = new Interval($startIndex, $stopIndex); |
2490
|
|
|
$tokenStream = $this->parser->getTokenStream(); |
2491
|
|
|
|
2492
|
|
|
$this->log[] = \sprintf( |
2493
|
|
|
'reportAmbiguity %s:%s, input = %s', |
2494
|
|
|
(string) $ambigAlts, |
2495
|
|
|
(string) $configs, |
2496
|
|
|
$tokenStream === null ? '' : $tokenStream->getTextByInterval($interval) |
2497
|
|
|
); |
2498
|
|
|
} |
2499
|
|
|
|
2500
|
|
|
if ($this->parser !== null) { |
2501
|
|
|
$this->parser->getErrorListenerDispatch()->reportAmbiguity( |
2502
|
|
|
$this->parser, |
2503
|
|
|
$dfa, |
2504
|
|
|
$startIndex, |
2505
|
|
|
$stopIndex, |
2506
|
|
|
$exact, |
2507
|
|
|
$ambigAlts, |
2508
|
|
|
$configs |
2509
|
|
|
); |
2510
|
|
|
} |
2511
|
|
|
} |
2512
|
|
|
|
2513
|
|
|
public function setPredictionMode(int $mode) : void |
2514
|
|
|
{ |
2515
|
|
|
$this->mode = $mode; |
2516
|
|
|
} |
2517
|
|
|
|
2518
|
|
|
public function getPredictionMode() : int |
2519
|
|
|
{ |
2520
|
|
|
return $this->mode; |
2521
|
|
|
} |
2522
|
|
|
|
2523
|
|
|
public function getParser() : Parser |
2524
|
|
|
{ |
2525
|
|
|
return $this->parser; |
2526
|
|
|
} |
2527
|
|
|
} |
2528
|
|
|
|
This check looks for function or method calls that always return null and whose return value is used.
The method
getObject()
can return nothing but null, so it makes no sense to use the return value.The reason is most likely that a function or method is imcomplete or has been reduced for debug purposes.