Compute StartDesc with an offset to effective start

This means `.{2,4}foo` will now consider 4 or less before f as
a start candidate instead of every characters

1 files changed, 24 insertions, 3 deletions

diff --git a/src/regex_impl.cc b/src/regex_impl.cc
index b1630a1f..266f920e 100644
--- a/src/regex_impl.cc
+++ b/src/regex_impl.cc
@@ -886,8 +886,8 @@ private:
     }
 
     // Mutate start_desc with informations on which Codepoint could start a match.
-    // Returns true if the node possibly does not consume the char, in which case
-    // the next node would still be relevant for the parent node start chars computation.
+    // Returns true if the subsequent nodes are still relevant for computing the
+    // start desc
     template<RegexMode direction>
     bool compute_start_desc(ParsedRegex::NodeIndex index,
                              CompiledRegex::StartDesc& start_desc) const
@@ -916,10 +916,20 @@ private:
                     add_multi_byte_utf8();
                 return node.quantifier.allows_none();
             case ParsedRegex::AnyChar:
+                if (start_desc.offset + node.quantifier.max <= CompiledRegex::StartDesc::OffsetLimits::max())
+                {
+                    start_desc.offset += node.quantifier.max;
+                    return true;
+                }
                 for (auto& b : start_desc.map)
                     b = true;
                return node.quantifier.allows_none();
             case ParsedRegex::AnyCharExceptNewLine:
+                if (start_desc.offset + node.quantifier.max <= CompiledRegex::StartDesc::OffsetLimits::max())
+                {
+                    start_desc.offset += node.quantifier.max;
+                    return true;
+                }
                 for (Codepoint cp = 0; cp < single_byte_limit; ++cp)
                 {
                     if (cp != '\n')
@@ -1138,7 +1148,7 @@ String dump_regex(const CompiledRegex& program)
                     res += (char)c;
             }
         }
-        res += "]\n";
+        res += format("]+{}\n", static_cast<int>(desc.offset));
     };
     if (program.forward_start_desc)
         dump_start_desc(*program.forward_start_desc, "forward");
@@ -1557,6 +1567,17 @@ auto test_regex = UnitTest{[]{
     }
 
     {
+        TestVM<RegexMode::Forward | RegexMode::Search> vm{"(.{3,4}|f)oo"};
+        kak_assert(vm.forward_start_desc and vm.forward_start_desc->offset == 4);
+        for (int c = 0; c < CompiledRegex::StartDesc::count; ++c)
+            kak_assert(vm.forward_start_desc->map[c] == (c == 'f' or c == 'o'));
+
+        kak_assert(vm.exec("xxxoo", RegexExecFlags::None));
+        kak_assert(vm.exec("xfoo", RegexExecFlags::None));
+        kak_assert(not vm.exec("😄xoo", RegexExecFlags::None));
+    }
+
+    {
         auto eq = [](const CompiledRegex::NamedCapture& lhs,
                      const CompiledRegex::NamedCapture& rhs) {
             return lhs.name == rhs.name and