add empty subexpression grammar and tests

Author: mvorisek

resources/RegexGrammar.pp

@@ -135,7 +135,7 @@
     alternation()
 
 alternation:
-    concatenation() ( ::alternation:: concatenation() #alternation )*
+    concatenation()? ( <alternation> concatenation()? #alternation )*
 
 concatenation:
     (   internal_options() | assertion() | quantification() | condition() )
@@ -154,8 +154,8 @@
         <index>
       | ::assertion_reference_:: alternation() #assertioncondition
     )
-    ::_capturing:: concatenation()?
-    ( ::alternation:: concatenation()? )?
+    ::_capturing::
+    alternation()
     ::_capturing::
 
 assertion:
@@ -165,7 +165,8 @@
       | ::lookbehind_::          #lookbehind
       | ::negative_lookbehind_:: #negativelookbehind
     )
-    alternation() ::_capturing::
+    alternation()
+    ::_capturing::
 
 quantification:
     ( class() | simple() ) ( quantifier() #quantification )?
@@ -208,7 +209,8 @@
       | ::atomic_group_:: #atomicgroup
       | ::capturing_::
     )
-    alternation() ::_capturing::
+    alternation()
+    ::_capturing::
 
 non_capturing_internal_options:
     <non_capturing_internal_option>

tests/PHPStan/Analyser/nsrt/preg_match_shapes.php

@@ -901,6 +901,48 @@ function bugUnescapedDashAfterRange (string $string): void
 	}
 }
 
+function bugEmptySubexpression (string $string): void {
+	if (preg_match('//', $string, $matches)) {
+		assertType("array{string}", $matches); // could be array{''}
+	}
+
+	if (preg_match('/()/', $string, $matches)) {
+		assertType("array{string, string}", $matches); // could be array{'', ''}
+	}
+
+	if (preg_match('/|/', $string, $matches)) {
+		assertType("array{string}", $matches); // could be array{''}
+	}
+
+	if (preg_match('~|(a)~', $string, $matches)) {
+		assertType("array{0: string, 1?: 'a'}", $matches);
+	}
+
+	if (preg_match('~(a)|~', $string, $matches)) {
+		assertType("array{0: string, 1?: 'a'}", $matches);
+	}
+
+	if (preg_match('~(a)||(b)~', $string, $matches)) {
+		assertType("array{0: string, 1?: 'a'}|array{string, '', 'b'}", $matches);
+	}
+
+	if (preg_match('~(|(a))~', $string, $matches)) {
+		assertType("array{0: string, 1: ''|'a', 2?: 'a'}", $matches);
+	}
+
+	if (preg_match('~((a)|)~', $string, $matches)) {
+		assertType("array{0: string, 1: ''|'a', 2?: 'a'}", $matches);
+	}
+
+	if (preg_match('~((a)||(b))~', $string, $matches)) {
+		assertType("array{0: string, 1: ''|'a'|'b', 2?: ''|'a', 3?: 'b'}", $matches);
+	}
+
+	if (preg_match('~((a)|()|(b))~', $string, $matches)) {
+		assertType("array{0: string, 1: string, 2?: ''|'a', 3?: string, 4?: 'b'}", $matches);
+	}
+}
+
 function bug11744(string $string): void
 {
 	if (!preg_match('~^((/[a-z]+)?)~', $string, $matches)) {