Add support for ATN tracing (#32)

Author: marcospassos

composer.json

@@ -12,7 +12,8 @@
   ],
   "require": {
     "php": "^8.0",
-    "ext-mbstring": "*"
+    "ext-mbstring": "^8.0",
+    "psr/log": "^2.0 || ^3.0"
   },
   "require-dev": {
     "ergebnis/composer-normalize": "^2.15",

composer.lock

@@ -636,6 +636,10 @@
         <severity>5</severity>
     </rule>
 
+    <rule ref="SlevomatCodingStandard.TypeHints.ParameterTypeHint.MissingNativeTypeHint">
+        <exclude-pattern>src/StdoutMessageLogger.php</exclude-pattern>
+    </rule>
+
     <rule ref="SlevomatCodingStandard.Classes.SuperfluousAbstractClassNaming.SuperfluousPrefix">
         <exclude-pattern>src/Atn/Transitions/AbstractPredicateTransition.php</exclude-pattern>
     </rule>

phpcs.xml.dist

@@ -157,7 +157,7 @@ public function toString(bool $showAlt): string
             $buf .= ',[' . $this->context . ']';
         }
 
-        if ($this->semanticContext->equals(SemanticContext::none())) {
+        if (!$this->semanticContext->equals(SemanticContext::none())) {
             $buf .= ',' . $this->semanticContext;
         }
 
@@ -177,9 +177,9 @@ public function __toString(): string
             $this->state,
             $this->alt,
             $this->context !== null ? ',[' . $this->context . ']' : '',
-            $this->semanticContext->equals(SemanticContext::none()) ?
-                ',' . $this->semanticContext :
-                '',
+            $this->semanticContext->equals(SemanticContext::none())
+                ? ''
+                : ',' . $this->semanticContext,
             $this->reachesIntoOuterContext > 0 ? ',up=' . $this->reachesIntoOuterContext : '',
         );
     }

src/Atn/ATNConfig.php

@@ -24,6 +24,7 @@
 use Antlr\Antlr4\Runtime\Error\Exceptions\RecognitionException;
 use Antlr\Antlr4\Runtime\IntervalSet;
 use Antlr\Antlr4\Runtime\IntStream;
+use Antlr\Antlr4\Runtime\LoggerProvider;
 use Antlr\Antlr4\Runtime\Parser;
 use Antlr\Antlr4\Runtime\ParserRuleContext;
 use Antlr\Antlr4\Runtime\PredictionContexts\PredictionContext;
@@ -35,6 +36,7 @@
 use Antlr\Antlr4\Runtime\Utils\BitSet;
 use Antlr\Antlr4\Runtime\Utils\DoubleKeyMap;
 use Antlr\Antlr4\Runtime\Utils\Set;
+use Psr\Log\LoggerInterface as Logger;
 
 /**
  * The embodiment of the adaptive LL(*), ALL(*), parsing strategy.
@@ -231,6 +233,8 @@
  */
 final class ParserATNSimulator extends ATNSimulator
 {
+    public static bool $traceAtnSimulation = false;
+
     protected Parser $parser;
 
     /** @var array<DFA> */
@@ -260,6 +264,8 @@ final class ParserATNSimulator extends ATNSimulator
 
     protected ?DFA $dfa = null;
 
+    private Logger $logger;
+
     /**
      * @param array<DFA> $decisionToDFA
      */
@@ -273,6 +279,7 @@ public function __construct(
 
         $this->parser = $parser;
         $this->decisionToDFA = $decisionToDFA;
+        $this->logger = LoggerProvider::getLogger();
     }
 
     public function reset(): void
@@ -296,6 +303,18 @@ public function clearDFA(): void
      */
     public function adaptivePredict(TokenStream $input, int $decision, ParserRuleContext $outerContext): int
     {
+        if (self::$traceAtnSimulation) {
+            $this->logger->debug(
+                'adaptivePredict decision {decision} exec LA(1)=={token} line {line}:{pos}',
+                [
+                    'decision' => $decision,
+                    'token' => $this->getTokenName($input->LA(1)),
+                    'line' => $input->LT(1)?->getLine(),
+                    'pos' => $input->LT(1)?->getCharPositionInLine(),
+                ],
+            );
+        }
+
         $this->input = $input;
         $this->startIndex = $input->getIndex();
         $this->outerContext = $outerContext;
@@ -404,6 +423,19 @@ public function execATN(
         int $startIndex,
         ParserRuleContext $outerContext,
     ): ?int {
+        if (self::$traceAtnSimulation) {
+            $this->logger->debug(
+                'execATN decision {decision}, DFA state {state}, LA(1)=={token} line {line}:{pos}',
+                [
+                    'decision' => $dfa->decision,
+                    'state' => $s0->__toString(),
+                    'token' => $this->getTokenName($input->LA(1)),
+                    'line' => $input->LT(1)?->getLine(),
+                    'pos' => $input->LT(1)?->getCharPositionInLine(),
+                ],
+            );
+        }
+
         $previousD = $s0;
 
         $t = $input->LA(1);
@@ -655,6 +687,12 @@ protected function execATNWithFullContext(
         int $startIndex,
         ParserRuleContext $outerContext,
     ): int {
+        if (self::$traceAtnSimulation) {
+            $this->logger->debug('execATNWithFullContext {state}', [
+                'state' => $s0->__toString(),
+            ]);
+        }
+
         $fullCtx = true;
         $foundExactAmbig = false;
         $reach = null;
@@ -890,15 +928,18 @@ protected function computeReachSet(ATNConfigSet $closure, int $t, bool $fullCtx)
          * multiple alternatives are viable.*/
 
         if ($skippedStopStates !== null && (!$fullCtx || !PredictionMode::hasConfigInRuleStopState($reach))) {
-            if (\count($skippedStopStates) === 0) {
-                throw new \LogicException('Skipped stop states cannot be empty.');
-            }
-
             foreach ($skippedStopStates as $lValue) {
                 $reach->add($lValue, $this->mergeCache);
             }
         }
 
+        if (self::$traceAtnSimulation) {
+            $this->logger->debug('computeReachSet {closure} -> {reach}', [
+                'closure' => $closure->__toString(),
+                'reach' => $reach->__toString(),
+            ]);
+        }
+
         if ($reach->isEmpty()) {
             return null;
         }
@@ -964,6 +1005,13 @@ protected function computeStartState(ATNState $p, RuleContext $ctx, bool $fullCt
         $initialContext = PredictionContext::fromRuleContext($this->atn, $ctx);
         $configs = new ATNConfigSet($fullCtx);
 
+        if (self::$traceAtnSimulation) {
+            $this->logger->debug('computeStartState from ATN state {state} initialContext={initialContext}', [
+                'state' => $p->__toString(),
+                'initialContext' => $initialContext->__toString(),
+            ]);
+        }
+
         foreach ($p->getTransitions() as $i => $t) {
             $c = new ATNConfig(null, $t->target, $initialContext, null, $i + 1);
             $closureBusy = new Set();
@@ -1464,6 +1512,12 @@ protected function closureCheckingStopState(
         int $depth,
         bool $treatEofAsEpsilon,
     ): void {
+        if (self::$traceAtnSimulation) {
+            $this->logger->debug('closure({config})', [
+                'config' => $config->toString(true),
+            ]);
+        }
+
         if ($config->state instanceof RuleStopState) {
             // We hit rule end. If we have context info, use it run thru all possible stack tops in ctx
             $context = $config->context;
@@ -2150,6 +2204,12 @@ protected function addDFAState(DFA $dfa, DFAState $D): DFAState
         $existing = $dfa->states->get($D);
 
         if ($existing instanceof DFAState) {
+            if (self::$traceAtnSimulation) {
+                $this->logger->debug('addDFAState {state} exists', [
+                    'state' => $D->__toString(),
+                ]);
+            }
+
             return $existing;
         }
 
@@ -2160,6 +2220,12 @@ protected function addDFAState(DFA $dfa, DFAState $D): DFAState
             $D->configs->setReadonly(true);
         }
 
+        if (self::$traceAtnSimulation) {
+            $this->logger->debug('addDFAState new {state}', [
+                'state' => $D->__toString(),
+            ]);
+        }
+
         $dfa->states->add($D);
 
         return $D;

src/Atn/ParserATNSimulator.php

@@ -0,0 +1,26 @@
+<?php
+
+declare(strict_types=1);
+
+namespace Antlr\Antlr4\Runtime;
+
+use Psr\Log\LoggerInterface as PsrLogger;
+
+final class LoggerProvider
+{
+    private static ?PsrLogger $logger = null;
+
+    public static function setLogger(PsrLogger $logger): void
+    {
+        self::$logger = $logger;
+    }
+
+    public static function getLogger(): PsrLogger
+    {
+        if (self::$logger === null) {
+            self::$logger = new StdoutMessageLogger();
+        }
+
+        return self::$logger;
+    }
+}