Added support for -- subtraction opetor.

2025-12-17 01:03:14 +01:00 · 2025-03-02 19:46:56 +01:00
parent ee83c42b71
commit 3664d50772
5 changed files with 180 additions and 198 deletions
--- a/javascript/extractor/src/com/semmle/js/ast/regexp/CharacterClassSubtraction.java
+++ b/javascript/extractor/src/com/semmle/js/ast/regexp/CharacterClassSubtraction.java
@@ -0,0 +1,22 @@
+package com.semmle.js.ast.regexp;
+
+import com.semmle.js.ast.SourceLocation;
+import java.util.List;
+
+public class CharacterClassSubtraction extends RegExpTerm {
+    private final List<RegExpTerm> subtraction;
+
+    public CharacterClassSubtraction(SourceLocation loc, List<RegExpTerm> subtraction) {
+        super(loc, "CharacterClassSubtraction");
+        this.subtraction = subtraction;
+    }
+
+    @Override
+    public void accept(Visitor v) {
+        v.visit(this);
+    }
+
+    public List<RegExpTerm> getSubtraction() {
+        return subtraction;
+    }
+}
--- a/javascript/extractor/src/com/semmle/js/ast/regexp/Visitor.java
+++ b/javascript/extractor/src/com/semmle/js/ast/regexp/Visitor.java
@@ -65,4 +65,6 @@ public interface Visitor {
  public void visit(CharacterClassQuotedString nd);

  public void visit(CharacterClassIntersection nd);
+
+  public void visit(CharacterClassSubtraction nd);
 }
--- a/javascript/extractor/src/com/semmle/js/extractor/RegExpExtractor.java
+++ b/javascript/extractor/src/com/semmle/js/extractor/RegExpExtractor.java
@@ -12,6 +12,7 @@ import com.semmle.js.ast.regexp.CharacterClass;
 import com.semmle.js.ast.regexp.CharacterClassEscape;
 import com.semmle.js.ast.regexp.CharacterClassQuotedString;
 import com.semmle.js.ast.regexp.CharacterClassRange;
+import com.semmle.js.ast.regexp.CharacterClassSubtraction;
 import com.semmle.js.ast.regexp.Constant;
 import com.semmle.js.ast.regexp.ControlEscape;
 import com.semmle.js.ast.regexp.ControlLetter;
@@ -96,6 +97,7 @@ public class RegExpExtractor {
    termkinds.put("UnicodePropertyEscape", 27);
    termkinds.put("CharacterClassQuotedString", 28);
    termkinds.put("CharacterClassIntersection", 29);
+    termkinds.put("CharacterClassSubtraction", 30);
  }

  private static final String[] errmsgs =
@@ -362,6 +364,14 @@ public class RegExpExtractor {
      for (RegExpTerm element : nd.getIntersections())
        visit(element, lbl, i++);
    }
+
+    @Override
+    public void visit(CharacterClassSubtraction nd) {
+      Label lbl = extractTerm(nd, parent, idx);
+      int i = 0;
+      for (RegExpTerm element : nd.getSubtraction())
+        visit(element, lbl, i++);
+    }
  }

  public void extract(String src, SourceMap sourceMap, Node parent, boolean isSpeculativeParsing, String flags) {
--- a/javascript/extractor/src/com/semmle/js/parser/RegExpParser.java
+++ b/javascript/extractor/src/com/semmle/js/parser/RegExpParser.java
@@ -8,6 +8,7 @@ import com.semmle.js.ast.regexp.CharacterClass;
 import com.semmle.js.ast.regexp.CharacterClassEscape;
 import com.semmle.js.ast.regexp.CharacterClassQuotedString;
 import com.semmle.js.ast.regexp.CharacterClassRange;
+import com.semmle.js.ast.regexp.CharacterClassSubtraction;
 import com.semmle.js.ast.regexp.Constant;
 import com.semmle.js.ast.regexp.ControlEscape;
 import com.semmle.js.ast.regexp.ControlLetter;
@@ -566,6 +567,7 @@ public class RegExpParser {
  private enum CharacterClassType {
    STANDARD,
    INTERSECTION,
+    SUBTRACTION,
  }

  // ECMA 2024 `v` flag allows nested character classes.
@@ -588,6 +590,10 @@ public class RegExpParser {
        this.match("&&");
        classType = CharacterClassType.INTERSECTION;
      }
+      else if (lookahead("--")) {
+        this.match("--");
+        classType = CharacterClassType.SUBTRACTION;
+      }
      else {
        elements.add(this.parseCharacterClassElement());
      }
@@ -597,6 +603,8 @@ public class RegExpParser {
    switch (classType) {
      case INTERSECTION:
        return this.finishTerm(new CharacterClass(loc, Collections.singletonList(new CharacterClassIntersection(loc, elements)), inverted));
+      case SUBTRACTION:
+        return this.finishTerm(new CharacterClass(loc, Collections.singletonList(new CharacterClassSubtraction(loc, elements)), inverted));
      case STANDARD:
      default:
        return this.finishTerm(new CharacterClass(loc, elements, inverted));
@@ -614,7 +622,7 @@ public class RegExpParser {
          return atom;
      }
    }
-    if (!this.lookahead("-]") && this.match("-") && !(atom instanceof CharacterClassEscape))
+    if (!this.lookahead("-]") && !this.lookahead("--") && this.match("-") && !(atom instanceof CharacterClassEscape))
      return this.finishTerm(new CharacterClassRange(loc, atom, this.parseCharacterClassAtom()));
    return atom;
  }
--- a/javascript/extractor/tests/es2024/output/trap/subtraction.js.trap
+++ b/javascript/extractor/tests/es2024/output/trap/subtraction.js.trap
@@ -83,246 +83,186 @@ regexpterm(#20025,23,#20024,0,"[\p{Script_Extensions=Greek}--\p{Letter}]")
 locations_default(#20026,#10000,1,2,1,42)
 hasLocation(#20025,#20026)
 #20027=*
-regexpterm(#20027,24,#20025,0,"\p{Script_Extensions=Greek}--")
-#20028=@"loc,{#10000},1,3,1,31"
-locations_default(#20028,#10000,1,3,1,31)
-hasLocation(#20027,#20028)
-#20029=*
-regexpterm(#20029,27,#20027,0,"\p{Script_Extensions=Greek}")
-#20030=@"loc,{#10000},1,3,1,29"
-locations_default(#20030,#10000,1,3,1,29)
-hasLocation(#20029,#20030)
-unicode_property_escapename(#20029,"Script_Extensions")
-unicode_property_escapevalue(#20029,"Greek")
-#20031=*
-regexpterm(#20031,14,#20027,1,"-")
-#20032=@"loc,{#10000},1,31,1,31"
-locations_default(#20032,#10000,1,31,1,31)
-hasLocation(#20031,#20032)
-regexp_const_value(#20031,"-")
+regexpterm(#20027,30,#20025,0,"[\p{Script_Extensions=Greek}--\p{Letter}]")
+hasLocation(#20027,#20026)
+#20028=*
+regexpterm(#20028,27,#20027,0,"\p{Script_Extensions=Greek}")
+#20029=@"loc,{#10000},1,3,1,29"
+locations_default(#20029,#10000,1,3,1,29)
+hasLocation(#20028,#20029)
+unicode_property_escapename(#20028,"Script_Extensions")
+unicode_property_escapevalue(#20028,"Greek")
+#20030=*
+regexpterm(#20030,27,#20027,1,"\p{Letter}")
+#20031=@"loc,{#10000},1,32,1,41"
+locations_default(#20031,#10000,1,32,1,41)
+hasLocation(#20030,#20031)
+unicode_property_escapename(#20030,"Letter")
+#20032=*
+stmts(#20032,2,#20001,1,"/[[abc]--[cbd]]/v;")
+hasLocation(#20032,#20005)
+stmt_containers(#20032,#20001)
 #20033=*
-regexpterm(#20033,27,#20025,1,"\p{Letter}")
-#20034=@"loc,{#10000},1,32,1,41"
-locations_default(#20034,#10000,1,32,1,41)
-hasLocation(#20033,#20034)
-unicode_property_escapename(#20033,"Letter")
-#20035=*
-stmts(#20035,2,#20001,1,"/[[abc]--[cbd]]/v;")
-hasLocation(#20035,#20005)
-stmt_containers(#20035,#20001)
+exprs(#20033,5,#20032,0,"/[[abc]--[cbd]]/v")
+hasLocation(#20033,#20013)
+enclosing_stmt(#20033,#20032)
+expr_containers(#20033,#20001)
+literals("/[[abc]--[cbd]]/v","/[[abc]--[cbd]]/v",#20033)
+#20034=*
+regexpterm(#20034,23,#20033,0,"[[abc]--[cbd]]")
+#20035=@"loc,{#10000},2,2,2,15"
+locations_default(#20035,#10000,2,2,2,15)
+hasLocation(#20034,#20035)
 #20036=*
-exprs(#20036,5,#20035,0,"/[[abc]--[cbd]]/v")
-hasLocation(#20036,#20013)
-enclosing_stmt(#20036,#20035)
-expr_containers(#20036,#20001)
-literals("/[[abc]--[cbd]]/v","/[[abc]--[cbd]]/v",#20036)
+regexpterm(#20036,30,#20034,0,"[[abc]--[cbd]]")
+hasLocation(#20036,#20035)
 #20037=*
-regexpterm(#20037,1,#20036,0,"[[abc]--[cbd]]")
-#20038=@"loc,{#10000},2,2,2,15"
-locations_default(#20038,#10000,2,2,2,15)
+regexpterm(#20037,23,#20036,0,"[abc]")
+#20038=@"loc,{#10000},2,3,2,7"
+locations_default(#20038,#10000,2,3,2,7)
 hasLocation(#20037,#20038)
 #20039=*
-regexpterm(#20039,23,#20037,0,"[[abc]--[cbd]")
-#20040=@"loc,{#10000},2,2,2,14"
-locations_default(#20040,#10000,2,2,2,14)
+regexpterm(#20039,14,#20037,0,"a")
+#20040=@"loc,{#10000},2,4,2,4"
+locations_default(#20040,#10000,2,4,2,4)
 hasLocation(#20039,#20040)
+regexp_const_value(#20039,"a")
 #20041=*
-regexpterm(#20041,23,#20039,0,"[abc]")
-#20042=@"loc,{#10000},2,3,2,7"
-locations_default(#20042,#10000,2,3,2,7)
+regexpterm(#20041,14,#20037,1,"b")
+#20042=@"loc,{#10000},2,5,2,5"
+locations_default(#20042,#10000,2,5,2,5)
 hasLocation(#20041,#20042)
+regexp_const_value(#20041,"b")
 #20043=*
-regexpterm(#20043,14,#20041,0,"a")
-#20044=@"loc,{#10000},2,4,2,4"
-locations_default(#20044,#10000,2,4,2,4)
+regexpterm(#20043,14,#20037,2,"c")
+#20044=@"loc,{#10000},2,6,2,6"
+locations_default(#20044,#10000,2,6,2,6)
 hasLocation(#20043,#20044)
-regexp_const_value(#20043,"a")
+regexp_const_value(#20043,"c")
 #20045=*
-regexpterm(#20045,14,#20041,1,"b")
-#20046=@"loc,{#10000},2,5,2,5"
-locations_default(#20046,#10000,2,5,2,5)
+regexpterm(#20045,23,#20036,1,"[cbd]")
+#20046=@"loc,{#10000},2,10,2,14"
+locations_default(#20046,#10000,2,10,2,14)
 hasLocation(#20045,#20046)
-regexp_const_value(#20045,"b")
 #20047=*
-regexpterm(#20047,14,#20041,2,"c")
-#20048=@"loc,{#10000},2,6,2,6"
-locations_default(#20048,#10000,2,6,2,6)
+regexpterm(#20047,14,#20045,0,"c")
+#20048=@"loc,{#10000},2,11,2,11"
+locations_default(#20048,#10000,2,11,2,11)
 hasLocation(#20047,#20048)
 regexp_const_value(#20047,"c")
 #20049=*
-regexpterm(#20049,24,#20039,1,"--[")
-#20050=@"loc,{#10000},2,8,2,10"
-locations_default(#20050,#10000,2,8,2,10)
+regexpterm(#20049,14,#20045,1,"b")
+#20050=@"loc,{#10000},2,12,2,12"
+locations_default(#20050,#10000,2,12,2,12)
 hasLocation(#20049,#20050)
+regexp_const_value(#20049,"b")
 #20051=*
-regexpterm(#20051,14,#20049,0,"-")
-#20052=@"loc,{#10000},2,8,2,8"
-locations_default(#20052,#10000,2,8,2,8)
+regexpterm(#20051,14,#20045,2,"d")
+#20052=@"loc,{#10000},2,13,2,13"
+locations_default(#20052,#10000,2,13,2,13)
 hasLocation(#20051,#20052)
-regexp_const_value(#20051,"-")
+regexp_const_value(#20051,"d")
 #20053=*
-regexpterm(#20053,14,#20049,1,"[")
-#20054=@"loc,{#10000},2,10,2,10"
-locations_default(#20054,#10000,2,10,2,10)
-hasLocation(#20053,#20054)
-regexp_const_value(#20053,"[")
+stmts(#20053,2,#20001,2,"/[[abc] ... de]]/v;")
+hasLocation(#20053,#20007)
+stmt_containers(#20053,#20001)
+#20054=*
+exprs(#20054,5,#20053,0,"/[[abc] ... bde]]/v")
+hasLocation(#20054,#20017)
+enclosing_stmt(#20054,#20053)
+expr_containers(#20054,#20001)
+literals("/[[abc]--[cbd]--[bde]]/v","/[[abc]--[cbd]--[bde]]/v",#20054)
 #20055=*
-regexpterm(#20055,14,#20039,2,"c")
-#20056=@"loc,{#10000},2,11,2,11"
-locations_default(#20056,#10000,2,11,2,11)
+regexpterm(#20055,23,#20054,0,"[[abc]--[cbd]--[bde]]")
+#20056=@"loc,{#10000},3,2,3,22"
+locations_default(#20056,#10000,3,2,3,22)
 hasLocation(#20055,#20056)
-regexp_const_value(#20055,"c")
 #20057=*
-regexpterm(#20057,14,#20039,3,"b")
-#20058=@"loc,{#10000},2,12,2,12"
-locations_default(#20058,#10000,2,12,2,12)
-hasLocation(#20057,#20058)
-regexp_const_value(#20057,"b")
-#20059=*
-regexpterm(#20059,14,#20039,4,"d")
-#20060=@"loc,{#10000},2,13,2,13"
-locations_default(#20060,#10000,2,13,2,13)
-hasLocation(#20059,#20060)
-regexp_const_value(#20059,"d")
-#20061=*
-regexpterm(#20061,14,#20037,1,"]")
-#20062=@"loc,{#10000},2,15,2,15"
-locations_default(#20062,#10000,2,15,2,15)
-hasLocation(#20061,#20062)
-regexp_const_value(#20061,"]")
-#20063=*
-regexp_parse_errors(#20063,#20037,"unexpected character")
-hasLocation(#20063,#20062)
+regexpterm(#20057,30,#20055,0,"[[abc]--[cbd]--[bde]]")
+hasLocation(#20057,#20056)
+#20058=*
+regexpterm(#20058,23,#20057,0,"[abc]")
+#20059=@"loc,{#10000},3,3,3,7"
+locations_default(#20059,#10000,3,3,3,7)
+hasLocation(#20058,#20059)
+#20060=*
+regexpterm(#20060,14,#20058,0,"a")
+#20061=@"loc,{#10000},3,4,3,4"
+locations_default(#20061,#10000,3,4,3,4)
+hasLocation(#20060,#20061)
+regexp_const_value(#20060,"a")
+#20062=*
+regexpterm(#20062,14,#20058,1,"b")
+#20063=@"loc,{#10000},3,5,3,5"
+locations_default(#20063,#10000,3,5,3,5)
+hasLocation(#20062,#20063)
+regexp_const_value(#20062,"b")
 #20064=*
-stmts(#20064,2,#20001,2,"/[[abc] ... de]]/v;")
-hasLocation(#20064,#20007)
-stmt_containers(#20064,#20001)
-#20065=*
-exprs(#20065,5,#20064,0,"/[[abc] ... bde]]/v")
-hasLocation(#20065,#20017)
-enclosing_stmt(#20065,#20064)
-expr_containers(#20065,#20001)
-literals("/[[abc]--[cbd]--[bde]]/v","/[[abc]--[cbd]--[bde]]/v",#20065)
+regexpterm(#20064,14,#20058,2,"c")
+#20065=@"loc,{#10000},3,6,3,6"
+locations_default(#20065,#10000,3,6,3,6)
+hasLocation(#20064,#20065)
+regexp_const_value(#20064,"c")
 #20066=*
-regexpterm(#20066,1,#20065,0,"[[abc]--[cbd]--[bde]]")
-#20067=@"loc,{#10000},3,2,3,22"
-locations_default(#20067,#10000,3,2,3,22)
+regexpterm(#20066,23,#20057,1,"[cbd]")
+#20067=@"loc,{#10000},3,10,3,14"
+locations_default(#20067,#10000,3,10,3,14)
 hasLocation(#20066,#20067)
 #20068=*
-regexpterm(#20068,23,#20066,0,"[[abc]--[cbd]")
-#20069=@"loc,{#10000},3,2,3,14"
-locations_default(#20069,#10000,3,2,3,14)
+regexpterm(#20068,14,#20066,0,"c")
+#20069=@"loc,{#10000},3,11,3,11"
+locations_default(#20069,#10000,3,11,3,11)
 hasLocation(#20068,#20069)
+regexp_const_value(#20068,"c")
 #20070=*
-regexpterm(#20070,23,#20068,0,"[abc]")
-#20071=@"loc,{#10000},3,3,3,7"
-locations_default(#20071,#10000,3,3,3,7)
+regexpterm(#20070,14,#20066,1,"b")
+#20071=@"loc,{#10000},3,12,3,12"
+locations_default(#20071,#10000,3,12,3,12)
 hasLocation(#20070,#20071)
+regexp_const_value(#20070,"b")
 #20072=*
-regexpterm(#20072,14,#20070,0,"a")
-#20073=@"loc,{#10000},3,4,3,4"
-locations_default(#20073,#10000,3,4,3,4)
+regexpterm(#20072,14,#20066,2,"d")
+#20073=@"loc,{#10000},3,13,3,13"
+locations_default(#20073,#10000,3,13,3,13)
 hasLocation(#20072,#20073)
-regexp_const_value(#20072,"a")
+regexp_const_value(#20072,"d")
 #20074=*
-regexpterm(#20074,14,#20070,1,"b")
-#20075=@"loc,{#10000},3,5,3,5"
-locations_default(#20075,#10000,3,5,3,5)
+regexpterm(#20074,23,#20057,2,"[bde]")
+#20075=@"loc,{#10000},3,17,3,21"
+locations_default(#20075,#10000,3,17,3,21)
 hasLocation(#20074,#20075)
-regexp_const_value(#20074,"b")
 #20076=*
-regexpterm(#20076,14,#20070,2,"c")
-#20077=@"loc,{#10000},3,6,3,6"
-locations_default(#20077,#10000,3,6,3,6)
+regexpterm(#20076,14,#20074,0,"b")
+#20077=@"loc,{#10000},3,18,3,18"
+locations_default(#20077,#10000,3,18,3,18)
 hasLocation(#20076,#20077)
-regexp_const_value(#20076,"c")
+regexp_const_value(#20076,"b")
 #20078=*
-regexpterm(#20078,24,#20068,1,"--[")
-#20079=@"loc,{#10000},3,8,3,10"
-locations_default(#20079,#10000,3,8,3,10)
+regexpterm(#20078,14,#20074,1,"d")
+#20079=@"loc,{#10000},3,19,3,19"
+locations_default(#20079,#10000,3,19,3,19)
 hasLocation(#20078,#20079)
+regexp_const_value(#20078,"d")
 #20080=*
-regexpterm(#20080,14,#20078,0,"-")
-#20081=@"loc,{#10000},3,8,3,8"
-locations_default(#20081,#10000,3,8,3,8)
+regexpterm(#20080,14,#20074,2,"e")
+#20081=@"loc,{#10000},3,20,3,20"
+locations_default(#20081,#10000,3,20,3,20)
 hasLocation(#20080,#20081)
-regexp_const_value(#20080,"-")
+regexp_const_value(#20080,"e")
 #20082=*
-regexpterm(#20082,14,#20078,1,"[")
-#20083=@"loc,{#10000},3,10,3,10"
-locations_default(#20083,#10000,3,10,3,10)
+entry_cfg_node(#20082,#20001)
+#20083=@"loc,{#10000},1,1,1,0"
+locations_default(#20083,#10000,1,1,1,0)
 hasLocation(#20082,#20083)
-regexp_const_value(#20082,"[")
 #20084=*
-regexpterm(#20084,14,#20068,2,"c")
-#20085=@"loc,{#10000},3,11,3,11"
-locations_default(#20085,#10000,3,11,3,11)
-hasLocation(#20084,#20085)
-regexp_const_value(#20084,"c")
-#20086=*
-regexpterm(#20086,14,#20068,3,"b")
-#20087=@"loc,{#10000},3,12,3,12"
-locations_default(#20087,#10000,3,12,3,12)
-hasLocation(#20086,#20087)
-regexp_const_value(#20086,"b")
-#20088=*
-regexpterm(#20088,14,#20068,4,"d")
-#20089=@"loc,{#10000},3,13,3,13"
-locations_default(#20089,#10000,3,13,3,13)
-hasLocation(#20088,#20089)
-regexp_const_value(#20088,"d")
-#20090=*
-regexpterm(#20090,14,#20066,1,"--")
-#20091=@"loc,{#10000},3,15,3,16"
-locations_default(#20091,#10000,3,15,3,16)
-hasLocation(#20090,#20091)
-regexp_const_value(#20090,"--")
-#20092=*
-regexpterm(#20092,23,#20066,2,"[bde]")
-#20093=@"loc,{#10000},3,17,3,21"
-locations_default(#20093,#10000,3,17,3,21)
-hasLocation(#20092,#20093)
-#20094=*
-regexpterm(#20094,14,#20092,0,"b")
-#20095=@"loc,{#10000},3,18,3,18"
-locations_default(#20095,#10000,3,18,3,18)
-hasLocation(#20094,#20095)
-regexp_const_value(#20094,"b")
-#20096=*
-regexpterm(#20096,14,#20092,1,"d")
-#20097=@"loc,{#10000},3,19,3,19"
-locations_default(#20097,#10000,3,19,3,19)
-hasLocation(#20096,#20097)
-regexp_const_value(#20096,"d")
-#20098=*
-regexpterm(#20098,14,#20092,2,"e")
-#20099=@"loc,{#10000},3,20,3,20"
-locations_default(#20099,#10000,3,20,3,20)
-hasLocation(#20098,#20099)
-regexp_const_value(#20098,"e")
-#20100=*
-regexpterm(#20100,14,#20066,3,"]")
-#20101=@"loc,{#10000},3,22,3,22"
-locations_default(#20101,#10000,3,22,3,22)
-hasLocation(#20100,#20101)
-regexp_const_value(#20100,"]")
-#20102=*
-regexp_parse_errors(#20102,#20066,"unexpected character")
-hasLocation(#20102,#20101)
-#20103=*
-entry_cfg_node(#20103,#20001)
-#20104=@"loc,{#10000},1,1,1,0"
-locations_default(#20104,#10000,1,1,1,0)
-hasLocation(#20103,#20104)
-#20105=*
-exit_cfg_node(#20105,#20001)
-hasLocation(#20105,#20021)
-successor(#20064,#20065)
-successor(#20065,#20105)
-successor(#20035,#20036)
-successor(#20036,#20064)
+exit_cfg_node(#20084,#20001)
+hasLocation(#20084,#20021)
+successor(#20053,#20054)
+successor(#20054,#20084)
+successor(#20032,#20033)
+successor(#20033,#20053)
 successor(#20023,#20024)
-successor(#20024,#20035)
-successor(#20103,#20023)
+successor(#20024,#20032)
+successor(#20082,#20023)
 numlines(#10000,3,3,0)
 filetype(#10000,"javascript")