Make $regex.*search() functions not to match empty substrings in non empty strings

4 files changed, 127 insertions, 9 deletions

diff --git a/doc/manual.cli b/doc/manual.cli
index e16b0fa..8beb4ea 100644
--- a/doc/manual.cli
+++ b/doc/manual.cli
@@ -6128,8 +6128,9 @@ source <functions-target.cli>;
 
 The \c{$regex.*()} function family contains function that provide
 comprehensive regular expression matching and substitution facilities. The
-supported regular expression flavor is ECMAScript (more specifically,
-ECMA-262-based C++11 regular expressions).
+supported regular expression flavor is ECMAScript, more precisely,
+ECMA-262-based C++11 regular expressions. Note that the \c{match_not_null}
+flag is in effect unless the string being matched is empty.
 
 In the \c{$regex.*()} functions the substitution escape sequences in the
 format string (the \ci{fmt} argument) are extended with a subset of the Perl
diff --git a/doc/testscript.cli b/doc/testscript.cli
index 18dac41..7f7522d 100644
--- a/doc/testscript.cli
+++ b/doc/testscript.cli
@@ -2372,8 +2372,8 @@ continuations.
 \h#syntax-regex|Output Regex|
 
 Instead of literal text the expected result in output here-strings and
-here-documents can be specified as ECMAScript regular expressions (more
-specifically, ECMA-262-based C++11 regular expressions). To signal the use of
+here-documents can be specified as ECMAScript regular expressions, more
+precisely, ECMA-262-based C++11 regular expressions. To signal the use of
 regular expressions the redirect must end with the \c{~} modifier, for
 example:
 
@@ -3082,7 +3082,9 @@ illegal.
 
 Note that this builtin implementation deviates substantially from POSIX
 \c{sed} (as described next). Most significantly, the regular expression flavor
-is ECMAScript (more specifically, ECMA-262-based C++11 regular expressions).
+is ECMAScript, more precisely, ECMA-262-based C++11 regular expressions. Note
+that the \c{match_not_null} flag is in effect unless the line being matched
+is empty.
 
 \dl|
 
diff --git a/libbuild2/functions-regex.cxx b/libbuild2/functions-regex.cxx
index c46f6f5..de34d63 100644
--- a/libbuild2/functions-regex.cxx
+++ b/libbuild2/functions-regex.cxx
@@ -138,12 +138,24 @@ namespace build2
     //
     string s (to_string (move (v)));
 
+    // Match flags.
+    //
+    // Note that by default std::regex_search() matches the empty substrings
+    // in non-empty strings for all the major implementations. We suppress
+    // such a counter-intuitive behavior with the match_not_null flag (see the
+    // butl::regex_replace_search() function implementation for details).
+    //
+    regex_constants::match_flag_type mf (regex_constants::match_default);
+
+    if (!s.empty ())
+      mf |= regex_constants::match_not_null;
+
     if (!match && !subs)
-      return value (regex_search (s, rge)); // Return boolean value.
+      return value (regex_search (s, rge, mf)); // Return boolean value.
 
     match_results<string::const_iterator> m;
 
-    if (regex_search (s, m, rge))
+    if (regex_search (s, m, rge, mf))
     {
       assert (!m.empty ());
 
@@ -483,7 +495,19 @@ namespace build2
 
     for (auto& n: ns)
     {
-      if (regex_search (convert<string> (move (n)), rge))
+      string s (convert<string> (move (n)));
+
+      // Match flags.
+      //
+      // Suppress matching of empty substrings in non-empty strings (see above
+      // for details).
+      //
+      regex_constants::match_flag_type mf (regex_constants::match_default);
+
+      if (!s.empty ())
+        mf |= regex_constants::match_not_null;
+
+      if (regex_search (s, rge, mf))
         return true;
     }
 
@@ -516,7 +540,17 @@ namespace build2
       bool s (n.simple ());
       string v (convert<string> (s ? move (n) : name (n)));
 
-      if (regex_search (v, rge) == matching)
+      // Match flags.
+      //
+      // Suppress matching of empty substrings in non-empty strings (see above
+      // for details).
+      //
+      regex_constants::match_flag_type mf (regex_constants::match_default);
+
+      if (!v.empty ())
+        mf |= regex_constants::match_not_null;
+
+      if (regex_search (v, rge, mf) == matching)
         r.emplace_back (s ? name (move (v)) : move (n));
     }
 
diff --git a/tests/function/regex/testscript b/tests/function/regex/testscript
index 538bdab..7fbcc8e 100644
--- a/tests/function/regex/testscript
+++ b/tests/function/regex/testscript
@@ -366,6 +366,33 @@
       EOI
     }
   }
+
+  : empty-substring
+  :
+  : Note that regex_search() ignores the match_not_null flag for older
+  : versions of libstdc++ and libc++.
+  :
+  if (($cxx.id != 'gcc'   || $cxx.version.major >= 7) && \
+      ($cxx.id != 'clang' || $cxx.version.major >= 6))
+  {
+    : empty
+    :
+    $* <<EOI >'true'
+    print $regex.search('', '.*')
+    EOI
+
+    : match
+    :
+    $* <<EOI >'true'
+    print $regex.search('a', 'a*')
+    EOI
+
+    : no-match
+    :
+    $* <<EOI >'false'
+    print $regex.search('aa', 'b*')
+    EOI
+  }
 }
 
 : split
@@ -576,6 +603,33 @@
     print $regex.find_search(Foo.cxx, 'f', icase)
     EOI
   }
+
+  : empty-substring
+  :
+  : Note that regex_search() ignores the match_not_null flag for older
+  : versions of libstdc++ and libc++.
+  :
+  if (($cxx.id != 'gcc'   || $cxx.version.major >= 7) && \
+      ($cxx.id != 'clang' || $cxx.version.major >= 6))
+  {
+    : empty
+    :
+    $* <<EOI >'true'
+    print $regex.find_search('', '.*')
+    EOI
+
+    : match
+    :
+    $* <<EOI >'true'
+    print $regex.find_search('a', 'a*')
+    EOI
+
+    : no-match
+    :
+    $* <<EOI >'false'
+    print $regex.find_search('aa', 'b*')
+    EOI
+  }
 }
 
 : filter-search
@@ -607,6 +661,33 @@
     $* <<EOI >''
     print $regex.filter_search(-g, '-O')
     EOI
+
+    : empty-substring
+    :
+    : Note that regex_search() ignores the match_not_null flag for older
+    : versions of libstdc++ and libc++.
+    :
+    if (($cxx.id != 'gcc'   || $cxx.version.major >= 7) && \
+        ($cxx.id != 'clang' || $cxx.version.major >= 6))
+    {
+      : empty
+      :
+      $* <<EOI >'{}'
+      print $regex.filter_search('', '.*')
+      EOI
+
+      : match
+      :
+      $* <<EOI >'a'
+      print $regex.filter_search('a', 'a*')
+      EOI
+
+      : no-match
+      :
+      $* <<EOI >''
+      print $regex.filter_search('aa', 'b*')
+      EOI
+    }
   }
 
   : filter-out