From 0befab300849be7ac0f77bc4228f8de50a108191 Mon Sep 17 00:00:00 2001
From: Karen Arutyunov <karen@codesynthesis.com>
Date: Fri, 25 Oct 2024 18:35:43 +0200
Subject: Make regex_replace_search() not to match empty substrings in
 non-empty strings

---
 libbutl/regex.hxx      |  2 +-
 libbutl/regex.txx      | 19 +++++++++++++++++++
 tests/regex/testscript | 23 +++++++++++++++++++++++
 3 files changed, 43 insertions(+), 1 deletion(-)

diff --git a/libbutl/regex.hxx b/libbutl/regex.hxx
index 9b31075..69009c3 100644
--- a/libbutl/regex.hxx
+++ b/libbutl/regex.hxx
@@ -52,7 +52,7 @@ namespace butl
                         const std::basic_string<C>& fmt,
                         F&& append,
                         std::regex_constants::match_flag_type =
-                        std::regex_constants::match_default);
+                          std::regex_constants::match_default);
 
   // As above but concatenate non-matched substrings and matched substring
   // replacements into a string returning it as well as whether the search
diff --git a/libbutl/regex.txx b/libbutl/regex.txx
index 214d949..ec9f7af 100644
--- a/libbutl/regex.txx
+++ b/libbutl/regex.txx
@@ -217,6 +217,25 @@ namespace butl
     bool first_only ((flags & regex_constants::format_first_only) != 0);
     bool no_copy ((flags & regex_constants::format_no_copy) != 0);
 
+    // Note that by default the std::regex_search(), std::regex_replace(), and
+    // std::regex_iterator() functions match the empty substrings in non-empty
+    // strings for all the major implementations. For example:
+    //
+    // - regex_search("bb", "a*") call returns true.
+    //
+    // - regex_replace("bb", "a*", "x") call returns "xbxbx".
+    //
+    // - regex_replace("a", ".*", "x") call returns "xx".
+    //
+    // - Iterating using the regex_iterator("a", ".*") object ends up with the
+    //   two matches: "a" and "".
+    //
+    // Since such a behavior feels counter-intuitive, we suppress it using the
+    // match_not_null flag, except for the empty string.
+    //
+    if (!s.empty ())
+      flags |= regex_constants::match_not_null;
+
     // Beginning of the last unmatched substring.
     //
     str_it ub (s.begin ());
diff --git a/tests/regex/testscript b/tests/regex/testscript
index 93ad4b6..137469d 100644
--- a/tests/regex/testscript
+++ b/tests/regex/testscript
@@ -63,6 +63,21 @@
     :
     $* xay '/a/\lVZ/' >xvZy
   }
+
+  : empty-substring
+  :
+  : Note that the regex search-based replacement with the match_not_null flag
+  : is broken for older versions of libstdc++ and libc++ (may ignore
+  : match_not_null for the former and may hang for some string/pattern for the
+  : latter).
+  :
+  if (($cxx.id != 'gcc'   || $cxx.version.major >= 7) && \
+      ($cxx.id != 'clang' || $cxx.version.major >= 6))
+  {
+    $* '' '/.*/x/' >'x'      : empty
+    $* a  '/a*/x/' >'x'      : match
+    $* aa '/b*/x/'      == 1 : no-match
+  }
 }
 
 : replace-match
@@ -72,6 +87,14 @@
 
   $* abc  '/a(b)c/x\1y/'  >xby : match
   $* abcd '/a(b)c/x\1yd/' == 1 : no-match
+
+  : empty-substring
+  :
+  {
+    $* '' '/.*/x/'          >'x' : empty
+    $* a  '/a*/x/'          >'x' : match
+    $* ab '/a(c*)(b)/\1\2/' >'b' : match-mid
+  }
 }
 
 : invalid-regex-fmt
-- 
cgit v1.1