LUCENE-9578: TermRangeQuery empty string lower bound edge case (#1976)

Currently a TermRangeQuery with the empty String ("") as lower bound and
includeLower=false leads internally constructs an Automaton that doesn't match
anything. This is unexpected expecially for open upper bounds where any string
should be considered to be "higher" than the empty string.

This PR changes "Automata#makeBinaryInterval" so that for an empty string lower
bound and an open upper bound, any String should match the query regardless or
the includeLower flag.

Author: Christoph Büscher

lucene/core/src/java/org/apache/lucene/util/automaton/Automata.java

@@ -85,7 +85,22 @@ public static Automaton makeAnyBinary() {
     a.finishState();
     return a;
   }
-  
+
+  /**
+   * Returns a new (deterministic) automaton that accepts all binary terms except
+   * the empty string.
+   */
+  public static Automaton makeNonEmptyBinary() {
+    Automaton a = new Automaton();
+    int s1 = a.createState();
+    int s2 = a.createState();
+    a.setAccept(s2, true);
+    a.addTransition(s1, s2, 0, 255);
+    a.addTransition(s2, s2, 0, 255);
+    a.finishState();
+    return a;
+  }
+
   /**
    * Returns a new (deterministic) automaton that accepts any single codepoint.
    */
@@ -254,8 +269,12 @@ public static Automaton makeBinaryInterval(BytesRef min, boolean minInclusive, B
       cmp = min.compareTo(max);
     } else {
       cmp = -1;
-      if (min.length == 0 && minInclusive) {
-        return makeAnyBinary();
+      if (min.length == 0) {
+        if (minInclusive) {
+          return makeAnyBinary();
+        } else {
+          return makeNonEmptyBinary();
+        }
       }
     }
 
@@ -266,7 +285,7 @@ public static Automaton makeBinaryInterval(BytesRef min, boolean minInclusive, B
         return makeBinary(min);
       }
     } else if (cmp > 0) {
-      // max > min
+      // max < min
       return makeEmpty();
     }
 

lucene/core/src/test/org/apache/lucene/util/automaton/TestAutomaton.java

@@ -1331,6 +1331,23 @@ public void testMakeBinaryIntervalBasic() throws Exception {
     assertFalse(Operations.run(a, intsRef("baq")));
     assertTrue(Operations.run(a, intsRef("bara")));
   }
+  
+  public void testMakeBinaryIntervalLowerBoundEmptyString() throws Exception {
+    Automaton a = Automata.makeBinaryInterval(new BytesRef(""), true, new BytesRef("bar"), true);
+    assertTrue(Operations.run(a, intsRef("")));
+    assertTrue(Operations.run(a, intsRef("a")));
+    assertTrue(Operations.run(a, intsRef("bar")));
+    assertFalse(Operations.run(a, intsRef("bara")));
+    assertFalse(Operations.run(a, intsRef("baz")));
+    
+    
+    a = Automata.makeBinaryInterval(new BytesRef(""), false, new BytesRef("bar"), true);
+    assertFalse(Operations.run(a, intsRef("")));
+    assertTrue(Operations.run(a, intsRef("a")));
+    assertTrue(Operations.run(a, intsRef("bar")));
+    assertFalse(Operations.run(a, intsRef("bara")));
+    assertFalse(Operations.run(a, intsRef("baz")));
+  }
 
   public void testMakeBinaryIntervalEqual() throws Exception {
     Automaton a = Automata.makeBinaryInterval(new BytesRef("bar"), true, new BytesRef("bar"), true);
@@ -1352,6 +1369,12 @@ public void testMakeBinaryIntervalCommonPrefix() throws Exception {
     assertFalse(Operations.run(a, intsRef("barfoop")));
   }
 
+  public void testMakeBinaryExceptEmpty() throws Exception {
+    Automaton a = Automata.makeNonEmptyBinary();
+    assertFalse(Operations.run(a, intsRef("")));
+    assertTrue(Operations.run(a, intsRef(TestUtil.randomRealisticUnicodeString(random(), 1, 10))));
+  }
+
   public void testMakeBinaryIntervalOpenMax() throws Exception {
     Automaton a = Automata.makeBinaryInterval(new BytesRef("bar"), true, null, true);
     assertFalse(Operations.run(a, intsRef("bam")));
@@ -1366,6 +1389,19 @@ public void testMakeBinaryIntervalOpenMax() throws Exception {
     assertTrue(Operations.run(a, intsRef("zzz")));
   }
 
+  public void testMakeBinaryIntervalOpenMaxZeroLengthMin() throws Exception {
+    // when including min, automaton should accept "a"
+    Automaton a = Automata.makeBinaryInterval(new BytesRef(""), true, null, true);
+    assertTrue(Operations.run(a, intsRef("")));
+    assertTrue(Operations.run(a, intsRef("a")));
+    assertTrue(Operations.run(a, intsRef("aaaaaa")));
+    // excluding min should still accept "a"
+    a = Automata.makeBinaryInterval(new BytesRef(""), false, null, true);
+    assertFalse(Operations.run(a, intsRef("")));
+    assertTrue(Operations.run(a, intsRef("a")));
+    assertTrue(Operations.run(a, intsRef("aaaaaa")));
+  }
+
   public void testMakeBinaryIntervalOpenMin() throws Exception {
     Automaton a = Automata.makeBinaryInterval(null, true, new BytesRef("foo"), true);
     assertFalse(Operations.run(a, intsRef("foz")));