Merge remote-tracking branch 'origin/master' into regex-java-flag-x

lmasroca · lmasroca · commit d309908c4179 · 2026-06-28T19:44:30.000-03:00
# Conflicts:
#	core/src/main/kotlin/org/evomaster/core/parser/GeneRegexJavaVisitor.kt
diff --git a/core/src/main/kotlin/org/evomaster/core/parser/GeneRegexJavaVisitor.kt b/core/src/main/kotlin/org/evomaster/core/parser/GeneRegexJavaVisitor.kt
@@ -36,14 +36,22 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
      * Capture groups in order of appearance (1-based index -> list index 0).
      * Populated as the tree is walked. A backreference is only valid if it
      * appears after the group it references, which Java regex requires anyway.
+     * The value is nullable to represent a captured group that is unsatisfiable,
+     * for example when the group contains an empty character class like `([a&&b])`.
+     * In that case the map holds null instead of a DisjunctionListRxGene.
+     * @see buildDisjunctionList
      */
     private val captureGroups = mutableListOf<DisjunctionListRxGene?>()
 
     /**
      * Same as [captureGroups] but for named backreferences, which can be accessed
      * with their name or number.
+     * The value is nullable to represent a captured group that is unsatisfiable,
+     * for example when the group contains an empty character class like `([a&&b])`.
+     * In that case the map holds null instead of a DisjunctionListRxGene.
+     * @see buildDisjunctionList
      */
-    private val namedCaptureGroups = mutableMapOf<String, DisjunctionListRxGene>()
+    private val namedCaptureGroups = mutableMapOf<String, DisjunctionListRxGene?>()
 
     /**
      * Tracks the flags active in the current lexical scope.
@@ -52,13 +60,48 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
      */
     private var currentFlags = externalRegexFlags
 
+    /**
+     * Builds DisjunctionListRxGenes from a disjunction context, returns null if disjunction is unsatisfiable.
+     */
+    private fun buildDisjunctionList(ctx: RegexJavaParser.DisjunctionContext): DisjunctionListRxGene? {
+        val res = ctx.accept(this)
+        val validDisjunctions = res.genes.map { it as DisjunctionRxGene }
+
+        val satisfiableDisjunctions = validDisjunctions.filter{ !it.isUnsatisfiable() }
+
+        if(satisfiableDisjunctions.isEmpty()){
+            // As DisjunctionListRxGene extends CompositeFixedGene, its disjunctions list cannot be empty.
+            // In this case we return null to represent an unsatisfiable DisjunctionListRxGene.
+            return null
+        }
+
+        val disjList = DisjunctionListRxGene(satisfiableDisjunctions)
+
+        //TODO tmp hack until full handling of ^$. Assume full match when nested disjunctions
+        for (gene in disjList.disjunctions) {
+            gene.extraPrefix = false
+            gene.extraPostfix = false
+            gene.matchStart = true
+            gene.matchEnd = true
+        }
+        return disjList
+    }
+
     override fun visitPattern(ctx: RegexJavaParser.PatternContext): VisitResult {
 
         val res = ctx.disjunction().accept(this)
 
         val text = RegexUtils.getRegexExpByParserRuleContext(ctx)
 
-        val disjList = DisjunctionListRxGene(res.genes.map { it as DisjunctionRxGene })
+        val satisfiableDisjunctions = res.genes
+            .map { it as DisjunctionRxGene }
+            .filter{ !it.isUnsatisfiable() }
+
+        if (satisfiableDisjunctions.isEmpty()) {
+            throw IllegalStateException("Regex is unsatisfiable.")
+        }
+
+        val disjList = DisjunctionListRxGene(satisfiableDisjunctions)
 
         // we remove the <EOF> token from end of the string to store as sourceRegex
         val gene = RegexGene(
@@ -79,9 +122,19 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
         val matchStart = assertionMatches.first
         val matchEnd = assertionMatches.second
 
-        val disj = DisjunctionRxGene("disj", altRes.genes.map { it  }, matchStart, matchEnd)
+        val res = VisitResult()
+
+        // add disjunction if it has genes, OR if the alternative was purely assertions (^$) or flag scopes
+        // in that case altRes.genes is empty but the alternative is valid (matches "")
+        val hasOnlyAssertionsOrFlagScopes = ctx.alternative().term().isNotEmpty() &&
+                ctx.alternative().term().all { it.assertion() != null || it.FLAG_SCOPE_OPEN() != null }
 
-        val res = VisitResult(disj)
+        if (altRes.genes.isNotEmpty() || hasOnlyAssertionsOrFlagScopes || ctx.alternative().term().isEmpty()) {
+            val disj = DisjunctionRxGene("disj", altRes.genes.map { it }, matchStart, matchEnd)
+
+            res.genes.add(disj)
+        }
+        // else: had non-assertion terms but all produced nothing (empty char class etc.), skip
 
         if(ctx.disjunction() != null){
             val disjRes = ctx.disjunction().accept(this)
@@ -147,7 +200,7 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
                 // term is not a back ref: we use the default behavior, term results may only have 0-1 genes
                 // if there is a gene, we add it to result
                 res.genes.add(gene)
-            } else {
+            } else if (resTerm.data is String) {
 
                 val assertion = resTerm.data as String
                 if(i==0 && assertion == "^"){
@@ -162,6 +215,9 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
                      */
                     throw IllegalStateException("Cannot support $assertion at position $i")
                 }
+            } else {
+                // unsatisfiable term, return with no genes
+                return VisitResult(data=Pair(false, false))
             }
         }
 
@@ -181,12 +237,23 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
 
         val resAtom = ctx.atom().accept(this)
         val atom = resAtom.genes.firstOrNull()
-            ?: return res
 
         if(ctx.quantifier() != null){
 
             val limits = ctx.quantifier().accept(this).data as Pair<Int,Int>
 
+            // if quantified atom is unsatisfiable we must then check the limits
+            if (atom == null ||
+                ((atom as? RxTerm)?.isUnsatisfiable() == true) && resAtom.genes.size == 1) {
+                return if (limits.first == 0) {
+                    // if 0 appearances is allowed then the regex is satisfiable only with empty string
+                    VisitResult(PatternCharacterBlockGene("0_QuantifierOnEmptyRegex", ""))
+                } else {
+                    // if not then unsatisfiable, return with no genes
+                    res
+                }
+            }
+
             // if atom is not a back ref then we use the default behavior, results may only have one gene
             var template: Gene = atom
 
@@ -211,10 +278,11 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
             if (ctx.atom()?.atomEscape()?.BackReference() != null){
                 // if atom is a BackReference we addAll genes from result as there may be more than one if digits are dropped
                 res.genes.addAll(resAtom.genes)
-            } else {
+            } else if (atom != null) {
                 // if atom is not a back ref we fall back to the default behavior, results only have one gene
                 res.genes.add(atom)
             }
+            // else atom is unsatisfiable, return no genes
         }
 
         return res
@@ -284,21 +352,16 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
 
             currentFlags = merged
 
-            val res = ctx.disjunction().accept(this)
+            val disjList = buildDisjunctionList(ctx.disjunction())
 
             currentFlags = previous
 
-            val disjList = DisjunctionListRxGene(res.genes.map { it as DisjunctionRxGene })
-
-            //TODO tmp hack until full handling of ^$. Assume full match when nested disjunctions
-            for (gene in disjList.disjunctions) {
-                gene.extraPrefix = false
-                gene.extraPostfix = false
-                gene.matchStart = true
-                gene.matchEnd = true
+            return if (disjList != null) {
+                VisitResult(disjList)
+            } else {
+                // unsatisfiable, return with no genes.
+                VisitResult()
             }
-
-            return VisitResult(disjList)
         }
 
         if(ctx.quote() != null){
@@ -332,17 +395,7 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
             val groupIndex = captureGroups.size
             captureGroups.add(null) // add placeholder for the gene
 
-            val res = ctx.disjunction().accept(this)
-
-            val disjList = DisjunctionListRxGene(res.genes.map { it as DisjunctionRxGene })
-
-            //TODO tmp hack until full handling of ^$. Assume full match when nested disjunctions
-            for(gene in disjList.disjunctions){
-                gene.extraPrefix = false
-                gene.extraPostfix = false
-                gene.matchStart = true
-                gene.matchEnd = true
-            }
+            val disjList = buildDisjunctionList(ctx.disjunction())
 
             val isCapturingGroup = !ctx.text.startsWith("(?:")
             val isNamedCaptureGroup = ctx.NAMED_CAPTURE_GROUP_OPEN() != null
@@ -358,7 +411,12 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
                 namedCaptureGroups[name] = disjList
             }
 
-            return VisitResult(disjList)
+            return if (disjList != null) {
+                VisitResult(disjList)
+            } else {
+                // unsatisfiable, return with no genes.
+                VisitResult()
+            }
         }
 
         if(ctx.DOT() != null){
@@ -577,15 +635,12 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
                 maxDigits > allDigits.length -> allDigits.length
                 allDigits.take(maxDigits).toInt() <= captureGroups.size -> maxDigits
                 maxDigits > 1 -> maxDigits - 1
-                else -> throw IllegalStateException(
-                    "Backreference ${txt.take(2)} refers to group ${allDigits[0]} but only ${captureGroups.size} " +
-                            "capture group(s) have been defined so far"
-                )
+                else -> 1
             }
 
             val n = allDigits.take(backRefDigitCount).toInt()
 
-            val result = VisitResult(BackReferenceRxGene(n, captureGroups[n - 1]!!))
+            val result = VisitResult(BackReferenceRxGene(n, captureGroups.getOrNull(n - 1)))
 
             val remainingChars = allDigits.drop(backRefDigitCount)
 
@@ -601,8 +656,10 @@ class GeneRegexJavaVisitor(externalRegexFlags: RegexFlags = RegexFlags()) : Rege
         if (ctx.NamedBackReference() != null) {
             // strip "\k<" and ">"
             val name = txt.drop(3).dropLast(1)
+            if(name !in namedCaptureGroups){
+                throw IllegalStateException("Named backreference \\k<$name> refers to unknown group '$name'")
+            }
             val group = namedCaptureGroups[name]
-                ?: throw IllegalStateException("Named backreference \\k<$name> refers to unknown group '$name'")
             val groupIndex = captureGroups.indexOf(group) + 1  // 1-based, for the gene name
             return VisitResult(BackReferenceRxGene(groupIndex, group))
         }
diff --git a/core/src/main/kotlin/org/evomaster/core/search/gene/regex/BackReferenceRxGene.kt b/core/src/main/kotlin/org/evomaster/core/search/gene/regex/BackReferenceRxGene.kt
@@ -14,12 +14,18 @@ import org.evomaster.core.search.service.mutator.genemutation.SubsetGeneMutation
  * Represents a backreference \N in a regex (N being a number).
  * Its value is always identical to the current value of its [captureGroup].
  * It has no independent state and is therefore immutable.
+ * If capture group is null then the referenced group was unsatisfiable,
+ * in which case the same is true for the backreference to it.
  */
 class BackReferenceRxGene(
     val groupIndex: Int,
-    val captureGroup: DisjunctionListRxGene
+    val captureGroup: DisjunctionListRxGene?
 ) : RxAtom, SimpleGene("\\$groupIndex") {
 
+    override fun isUnsatisfiable(): Boolean {
+        return captureGroup == null || captureGroup.isUnsatisfiable()
+    }
+
     override fun checkForLocallyValidIgnoringChildren(): Boolean = true
 
     /**
@@ -59,7 +65,12 @@ class BackReferenceRxGene(
         mode: GeneUtils.EscapeMode?,
         targetFormat: OutputFormat?,
         extraCheck: Boolean
-    ): String = captureGroup.getValueAsPrintableString(targetFormat = null)
+    ): String {
+        if (captureGroup == null) {
+            throw IllegalStateException("Cannot get value from invalid backreference \\$groupIndex")
+        }
+        return captureGroup.getValueAsPrintableString(previousGenes, mode, targetFormat)
+    }
 
     override fun containsSameValueAs(other: Gene): Boolean {
         if (other !is BackReferenceRxGene) return false
diff --git a/core/src/main/kotlin/org/evomaster/core/search/gene/regex/CharacterClassEscapeRxGene.kt b/core/src/main/kotlin/org/evomaster/core/search/gene/regex/CharacterClassEscapeRxGene.kt
@@ -82,7 +82,7 @@ class CharacterClassEscapeRxGene(
             // create both normal and negated version for all
             .flatMap { (key, value) ->
                 listOf(
-                    key     to MultiCharacterRange(value),
+                    key     to MultiCharacterRange(false, value),
                     "^$key" to MultiCharacterRange(true,  value)
                 )
             }.toMap()
@@ -132,6 +132,12 @@ class CharacterClassEscapeRxGene(
         }
     }
 
+    override fun isUnsatisfiable(): Boolean = multiCharRange.isEmpty
+
+    override fun isMutable(): Boolean {
+        return !isUnsatisfiable()
+    }
+
     override fun checkForLocallyValidIgnoringChildren() : Boolean{
         // we pass the same embedded flags to the regex to accurately match the expected behavior
         return value.matches(Regex("${flags.getScopeString()}\\$type"))
@@ -193,6 +199,9 @@ class CharacterClassEscapeRxGene(
     }
 
     override fun getValueAsPrintableString(previousGenes: List<Gene>, mode: GeneUtils.EscapeMode?, targetFormat: OutputFormat?, extraCheck: Boolean): String {
+        if (isUnsatisfiable()) {
+            throw IllegalStateException("Cannot get value from empty CharacterClassEscape")
+        }
         return if (!flags.isCaseable(value[0])) {
             value[0].toString()
         }
diff --git a/core/src/main/kotlin/org/evomaster/core/search/gene/regex/CharacterRangeRxGene.kt b/core/src/main/kotlin/org/evomaster/core/search/gene/regex/CharacterRangeRxGene.kt
@@ -30,14 +30,19 @@ class CharacterRangeRxGene(
         private val log = LoggerFactory.getLogger(CharacterRangeRxGene::class.java)
     }
 
-    var value : Char = validRanges[0].start
+    // '\u0000' is a placeholder for the unsatisfiable case (empty MCR with no valid ranges).
+    // This gene will never be randomized or mutated when isUnsatisfiable() is true, as it would be immutable.
+    // getValueAsPrintableString throws when isUnsatisfiable, so that value should be unreachable.
+    var value : Char = if (isUnsatisfiable()) '\u0000' else validRanges[0].start
 
     /**
      * Whether to output the character in uppercase.
      * Only meaningful when flags.caseInsensitive is true.
      */
     var useUpperCase: Boolean = false
 
+    override fun isUnsatisfiable(): Boolean = validRanges.isEmpty
+
     override fun checkForLocallyValidIgnoringChildren() : Boolean{
         return validRanges.any {
             value in it ||
@@ -49,6 +54,9 @@ class CharacterRangeRxGene(
     }
 
     override fun isMutable(): Boolean {
+        if (isUnsatisfiable()) {
+            return false
+        }
         // check if there is more than one character or if the character is caseable
         return validRanges.charCount > 1 || flags.isCaseable(value)
     }
@@ -134,6 +142,9 @@ class CharacterRangeRxGene(
             TODO should \ be handled specially?
             In any case, would have same handling as AnyCharacterRxGene
          */
+        if (isUnsatisfiable()) {
+            throw IllegalStateException("Cannot get value from empty CharacterRange")
+        }
         return if (!flags.isCaseable(value)) {
             value.toString()
         }
diff --git a/core/src/main/kotlin/org/evomaster/core/search/gene/regex/DisjunctionRxGene.kt b/core/src/main/kotlin/org/evomaster/core/search/gene/regex/DisjunctionRxGene.kt
@@ -47,6 +47,8 @@ class DisjunctionRxGene(
         private val log : Logger = LoggerFactory.getLogger(DisjunctionRxGene::class.java)
     }
 
+    override fun isUnsatisfiable(): Boolean =
+        terms.isNotEmpty() && terms.any { (it as? RxTerm)?.isUnsatisfiable() == true }
 
     override fun checkForLocallyValidIgnoringChildren() : Boolean{
         return true
diff --git a/core/src/main/kotlin/org/evomaster/core/search/gene/regex/QuantifierRxGene.kt b/core/src/main/kotlin/org/evomaster/core/search/gene/regex/QuantifierRxGene.kt
@@ -49,8 +49,8 @@ class QuantifierRxGene(
         if (min < 0) {
             throw IllegalArgumentException("Invalid min value '$min': should be positive")
         }
-        if (max < 1) {
-            throw IllegalArgumentException("Invalid max value '$max': should be at least 1")
+        if (max < 0) {
+            throw IllegalArgumentException("Invalid max value '$max': should be positive")
         }
         if (min > max) {
             throw IllegalArgumentException("Invalid min-max values '$min-$max': min is greater than max")
diff --git a/core/src/main/kotlin/org/evomaster/core/search/gene/regex/RxTerm.kt b/core/src/main/kotlin/org/evomaster/core/search/gene/regex/RxTerm.kt
@@ -4,4 +4,11 @@ import org.evomaster.core.search.StructuralElement
 import org.evomaster.core.search.gene.Gene
 
 
-interface RxTerm
+interface RxTerm {
+    /**
+     * Returns true if this gene can never produce a valid value,
+     * for example an empty character class intersection like [a&&b].
+     * Used at construction time to filter unsatisfiable branches from disjunctions.
+     */
+    fun isUnsatisfiable(): Boolean = false
+}
diff --git a/core/src/main/kotlin/org/evomaster/core/utils/MultiCharacterRange.kt b/core/src/main/kotlin/org/evomaster/core/utils/MultiCharacterRange.kt
diff --git a/core/src/test/kotlin/org/evomaster/core/parser/GeneRegexJavaVisitorTest.kt b/core/src/test/kotlin/org/evomaster/core/parser/GeneRegexJavaVisitorTest.kt

Original file line number	Diff line number	Diff line change
`@@ -82,7 +82,7 @@ class CharacterClassEscapeRxGene(`
`82`	`82`	`// create both normal and negated version for all`
`83`	`83`	`.flatMap { (key, value) ->`
`84`	`84`	`listOf(`
`85`		`- key to MultiCharacterRange(value),`
	`85`	`+ key to MultiCharacterRange(false, value),`
`86`	`86`	`"^$key" to MultiCharacterRange(true, value)`
`87`	`87`	`)`
`88`	`88`	`}.toMap()`
`@@ -132,6 +132,12 @@ class CharacterClassEscapeRxGene(`
`132`	`132`	`}`
`133`	`133`	`}`
`134`	`134`
	`135`	`+ override fun isUnsatisfiable(): Boolean = multiCharRange.isEmpty`
	`136`	`+`
	`137`	`+ override fun isMutable(): Boolean {`
	`138`	`+ return !isUnsatisfiable()`
	`139`	`+ }`
	`140`	`+`
`135`	`141`	`override fun checkForLocallyValidIgnoringChildren() : Boolean{`
`136`	`142`	`// we pass the same embedded flags to the regex to accurately match the expected behavior`
`137`	`143`	`return value.matches(Regex("${flags.getScopeString()}\\$type"))`
`@@ -193,6 +199,9 @@ class CharacterClassEscapeRxGene(`
`193`	`199`	`}`
`194`	`200`
`195`	`201`	`override fun getValueAsPrintableString(previousGenes: List<Gene>, mode: GeneUtils.EscapeMode?, targetFormat: OutputFormat?, extraCheck: Boolean): String {`
	`202`	`+ if (isUnsatisfiable()) {`
	`203`	`+ throw IllegalStateException("Cannot get value from empty CharacterClassEscape")`
	`204`	`+ }`
`196`	`205`	`return if (!flags.isCaseable(value[0])) {`
`197`	`206`	`value[0].toString()`
`198`	`207`	`}`
Original file line number	Diff line number	Diff line change
`@@ -47,6 +47,8 @@ class DisjunctionRxGene(`
`47`	`47`	`private val log : Logger = LoggerFactory.getLogger(DisjunctionRxGene::class.java)`
`48`	`48`	`}`
`49`	`49`
	`50`	`+ override fun isUnsatisfiable(): Boolean =`
	`51`	`+ terms.isNotEmpty() && terms.any { (it as? RxTerm)?.isUnsatisfiable() == true }`
`50`	`52`
`51`	`53`	`override fun checkForLocallyValidIgnoringChildren() : Boolean{`
`52`	`54`	`return true`
Original file line number	Diff line number	Diff line change
`@@ -49,8 +49,8 @@ class QuantifierRxGene(`
`49`	`49`	`if (min < 0) {`
`50`	`50`	`throw IllegalArgumentException("Invalid min value '$min': should be positive")`
`51`	`51`	`}`
`52`		`- if (max < 1) {`
`53`		`- throw IllegalArgumentException("Invalid max value '$max': should be at least 1")`
	`52`	`+ if (max < 0) {`
	`53`	`+ throw IllegalArgumentException("Invalid max value '$max': should be positive")`
`54`	`54`	`}`
`55`	`55`	`if (min > max) {`
`56`	`56`	`throw IllegalArgumentException("Invalid min-max values '$min-$max': min is greater than max")`