WordDB: Use interned strings pointing directly into the buffer line data

src/buffer.hh

@@ -126,7 +126,7 @@ public:
     BufferIterator end() const;
     LineCount      line_count() const;
 
-    const StringView& operator[](LineCount line) const
+    const InternedString& operator[](LineCount line) const
     { return m_lines[line]; }
 
     // returns an iterator at given coordinates. clamp line_and_column

src/interned_string.hh

@@ -79,6 +79,21 @@ public:
     using StringView::operator==;
     using StringView::operator!=;
 
+    InternedString acquire_substr(ByteCount from, ByteCount length = INT_MAX) const
+    {
+        if (m_slot == -1)
+            return InternedString{};
+        StringRegistry::instance().acquire(m_slot);
+        return InternedString{StringView::substr(from, length), m_slot};
+    }
+    InternedString acquire_substr(CharCount from, CharCount length = INT_MAX) const
+    {
+        if (m_slot == -1)
+            return InternedString{};
+        StringRegistry::instance().acquire(m_slot);
+        return InternedString{StringView::substr(from, length), m_slot};
+    }
+
 private:
     friend class StringRegistry;
 

src/selectors.cc

@@ -424,7 +424,8 @@ Selection select_indent(const Buffer& buffer, const Selection& selection, Object
     LineCount begin_line = line - 1;
     if (flags & ObjectFlags::ToBegin)
     {
-        while (begin_line >= 0 and (buffer[begin_line] == "\n" or get_indent(buffer[begin_line], tabstop) >= indent))
+        while (begin_line >= 0 and (buffer[begin_line] == StringView{"\n"} or
+                                    get_indent(buffer[begin_line], tabstop) >= indent))
             --begin_line;
     }
     ++begin_line;
@@ -432,7 +433,8 @@ Selection select_indent(const Buffer& buffer, const Selection& selection, Object
     if (flags & ObjectFlags::ToEnd)
     {
         const LineCount end = buffer.line_count();
-        while (end_line < end and (buffer[end_line] == "\n" or get_indent(buffer[end_line], tabstop) >= indent))
+        while (end_line < end and (buffer[end_line] == StringView{"\n"} or
+                                   get_indent(buffer[end_line], tabstop) >= indent))
             ++end_line;
     }
     --end_line;

src/unit_tests.cc

@@ -35,20 +35,20 @@ void test_buffer()
     // check insert at end behaviour: auto add end of line if necessary
     pos = buffer.end()-1;
     buffer.insert(pos, "tchou");
-    kak_assert(buffer.string(pos.coord(), buffer.end_coord()) == "tchou\n");
+    kak_assert(buffer.string(pos.coord(), buffer.end_coord()) == StringView{"tchou\n"});
 
     pos = buffer.end()-1;
     buffer.insert(buffer.end(), "kanaky\n");
-    kak_assert(buffer.string((pos+1).coord(), buffer.end_coord()) == "kanaky\n");
+    kak_assert(buffer.string((pos+1).coord(), buffer.end_coord()) == StringView{"kanaky\n"});
 
     buffer.commit_undo_group();
     buffer.erase(pos+1, buffer.end());
     buffer.insert(buffer.end(), "mutch\n");
     buffer.commit_undo_group();
     buffer.undo();
-    kak_assert(buffer.string(buffer.advance(buffer.end_coord(), -7), buffer.end_coord()) == "kanaky\n");
+    kak_assert(buffer.string(buffer.advance(buffer.end_coord(), -7), buffer.end_coord()) == StringView{"kanaky\n"});
     buffer.redo();
-    kak_assert(buffer.string(buffer.advance(buffer.end_coord(), -6), buffer.end_coord()) == "mutch\n");
+    kak_assert(buffer.string(buffer.advance(buffer.end_coord(), -6), buffer.end_coord()) == StringView{"mutch\n"});
 }
 
 void test_undo_group_optimizer()

src/word_db.cc

@@ -7,7 +7,7 @@
 namespace Kakoune
 {
 
-static std::vector<InternedString> get_words(StringView content)
+static std::vector<InternedString> get_words(const InternedString& content)
 {
     std::vector<InternedString> res;
     using Iterator = utf8::iterator<const char*, utf8::InvalidPolicy::Pass>;
@@ -24,7 +24,9 @@ static std::vector<InternedString> get_words(StringView content)
         }
         else if (in_word and not word)
         {
-            res.push_back(StringView{word_start, it.base()});
+            const ByteCount start = word_start - content.begin();
+            const ByteCount length = it.base() - word_start;
+            res.push_back(content.acquire_substr(start, length));
             in_word = false;
         }
     }