Store buffer content in a list of lines

Instead of a big std::string, buffer now store it's content in a
list of lines. In order to achieve O(log(n)) random access, lines
contains both their content and their offset since the start of
the file, making binary search usable.

BufferIterator now have a LineAndColumn coordinate into the buffer
instead of an offset so that access is still O(1).

gdb/kakoune.py

@@ -51,9 +51,9 @@ class BufferIterator:
 
     def to_string(self):
         if self.val['m_buffer'] != 0:
-            return "buffer<%s>@%d" % (self.val['m_buffer'].dereference()['m_name'], self.val['m_position'])
+            return "buffer<%s>@(%d, %d)" % (self.val['m_buffer'].dereference()['m_name'], self.val['m_coord']['line'], self.val['m_coord']['column'])
         else:
-            return "buffer<none>@%s" % (self.val['m_position'])
+            return "buffer<none>@(%d, %d)" % (self.val['m_coord']['line'], self.val['m_coord']['column'])
 
 def build_pretty_printer():
     pp = gdb.printing.RegexpCollectionPrettyPrinter("kakoune")

src/buffer.cc

@@ -47,37 +47,25 @@ Buffer::~Buffer()
 
 BufferIterator Buffer::iterator_at(const BufferCoord& line_and_column) const
 {
-    if (m_lines.empty())
+    return BufferIterator(*this, clamp(line_and_column));
-        return begin();
-
-    BufferCoord clamped = clamp(line_and_column);
-    return BufferIterator(*this, m_lines[clamped.line] + clamped.column);
 }
 
 BufferCoord Buffer::line_and_column_at(const BufferIterator& iterator) const
 {
-    BufferCoord result;
+    return iterator.m_coord;
-    if (not m_lines.empty())
-    {
-        result.line = line_at(iterator);
-        result.column = iterator.m_position - m_lines[result.line];
-    }
-    return result;
 }
 
 BufferPos Buffer::line_at(const BufferIterator& iterator) const
 {
-    auto it = std::upper_bound(m_lines.begin(), m_lines.end(),
+    return iterator.line();
-                               iterator.m_position);
-    return it - m_lines.begin() - 1;
 }
 
 BufferSize Buffer::line_length(BufferPos line) const
 {
-    assert(not m_lines.empty());
+    assert(line < line_count());
     BufferPos end = (line < m_lines.size() - 1) ?
-                    m_lines[line + 1] : length();
+                    m_lines[line + 1].start : length();
-    return end - m_lines[line];
+    return end - m_lines[line].start;
 }
 
 BufferCoord Buffer::clamp(const BufferCoord& line_and_column) const
@@ -87,36 +75,39 @@ BufferCoord Buffer::clamp(const BufferCoord& line_and_column) const
 
     BufferCoord result(line_and_column.line, line_and_column.column);
     result.line = Kakoune::clamp<int>(0, m_lines.size() - 1, result.line);
-    int max_col = std::max(0, line_length(result.line)-2);
+    int max_col = std::max(0, line_length(result.line) - 2);
     result.column = Kakoune::clamp<int>(0, max_col, result.column);
     return result;
 }
 
 BufferIterator Buffer::iterator_at_line_begin(const BufferIterator& iterator) const
 {
-    return BufferIterator(*this, m_lines[line_at(iterator)]);
+    return BufferIterator(*this, { iterator.line(), 0 });
 }
 
 BufferIterator Buffer::iterator_at_line_end(const BufferIterator& iterator) const
 {
-    BufferPos line = line_at(iterator) + 1;
+    BufferPos line = iterator.line();
-    return line < m_lines.size() ? BufferIterator(*this, m_lines[line]) : end();
+    return ++BufferIterator(*this, { line, std::max(line_length(line) - 1, 0) });
 }
 
-
 BufferIterator Buffer::begin() const
 {
-    return BufferIterator(*this, 0);
+    return BufferIterator(*this, { 0, 0 });
 }
 
 BufferIterator Buffer::end() const
 {
-    return BufferIterator(*this, length());
+    if (m_lines.empty())
+        return BufferIterator(*this, { 0, 0 });
+    return BufferIterator(*this, { (int)line_count()-1, (int)m_lines.back().content.length() });
 }
 
 BufferSize Buffer::length() const
 {
-    return m_content.size();
+    if (m_lines.empty())
+        return 0;
+    return m_lines.back().start + m_lines.back().content.length();
 }
 
 BufferSize Buffer::line_count() const
@@ -126,7 +117,18 @@ BufferSize Buffer::line_count() const
 
 String Buffer::string(const BufferIterator& begin, const BufferIterator& end) const
 {
-    return m_content.substr(begin.m_position, end - begin);
+    String res;
+    for (BufferPos line = begin.line(); line <= end.line(); ++line)
+    {
+       size_t start = 0;
+       if (line == begin.line())
+           start = begin.column();
+       size_t count = -1;
+       if (line == end.line())
+           count = end.column() - start;
+       res += m_lines[line].content.substr(start, count);
+    }
+    return res;
 }
 
 void Buffer::begin_undo_group()
@@ -182,79 +184,115 @@ bool Buffer::redo()
     ++m_history_cursor;
 }
 
-void Buffer::update_lines(const Modification& modification)
+void Buffer::check_invariant() const
 {
-    size_t length = modification.content.length();
+    BufferSize start = 0;
-    const BufferPos pos = modification.position.m_position;
+    for (auto& line : m_lines)
-    const BufferPos endpos = pos + length;
-
-    // find the first line beginning after modification position
-    auto line_it = std::upper_bound(m_lines.begin(), m_lines.end(), pos);
-
-    if (modification.type == Modification::Insert)
     {
-        // all following lines advanced by length
+        assert(line.start == start);
-        for (auto it = line_it; it != m_lines.end(); ++it)
+        start += line.content.length();
-            *it += length;
+    }
+}
+
+void Buffer::insert(const BufferIterator& pos, const String& content)
+{
+    BufferSize offset = pos.offset();
+
+    // all following lines advanced by length
+    for (size_t i = pos.line()+1; i < line_count(); ++i)
+        m_lines[i].start += content.length();
 
-        std::vector<BufferPos> new_lines;
     // if we inserted at the end of the buffer, we may have created a new
     // line without inserting a '\n'
-        if (endpos == m_content.size() and
+    if (pos == end() and (pos == begin() or *(pos-1) == '\n'))
-            (pos == 0 or m_content[pos-1] == '\n'))
-            new_lines.push_back(pos);
-
-        // every \n inserted that was not the last buffer character created a
-        // new line
-        for (BufferPos i = pos; i < endpos and i + 1 < m_content.size(); ++i)
     {
-            if (m_content[i] == '\n')
+        int start = 0;
-                new_lines.push_back(i + 1);
+        for (int i = 0; i < content.length(); ++i)
-        }
-        m_lines.insert(line_it, new_lines.begin(), new_lines.end());
-    }
-    else if (modification.type == Modification::Erase)
         {
-        // find the first line beginning after endpos
+            if (content[i] == '\n')
-        auto end = std::upper_bound(line_it, m_lines.end(), endpos);
-
-        // all the lines until the end moved back by length
-        for (auto it = end; it != m_lines.end(); ++it)
             {
-            *it -= length;
+                m_lines.push_back({ offset + start, content.substr(start, i + 1 - start) });
-            assert(m_content[(*it)-1] == '\n');
+                start = i + 1;
             }
-
+        }
-        // if we erased from the beginning of a line until the end of
+        if (start != content.length())
-        // the buffer, that line also needs to be erased
+            m_lines.push_back({ offset + start, content.substr(start) });
-        if (pos == m_content.size() and *(line_it-1) == pos)
-            --line_it;
-        m_lines.erase(line_it, end);
     }
     else
-        assert(false);
+    {
+        String prefix = m_lines[pos.line()].content.substr(0, pos.column());
+        String suffix = m_lines[pos.line()].content.substr(pos.column());
+
+        auto line_it = m_lines.begin() + pos.line();
+        line_it = m_lines.erase(line_it);
+
+        int start = 0;
+        for (int i = 0; i < content.length(); ++i)
+        {
+            if (content[i] == '\n')
+            {
+                String line_content = content.substr(start, i + 1 - start);
+                if (start == 0)
+                {
+                    line_content = prefix + line_content;
+                    line_it = m_lines.insert(line_it, { offset + start - (int)prefix.length(),
+                                                        std::move(line_content) });
+                }
+                else
+                    line_it = m_lines.insert(line_it, { offset + start,
+                                                        std::move(line_content) });
+
+                ++line_it;
+                start = i + 1;
+            }
+        }
+        if (start == 0)
+            m_lines.insert(line_it, { offset + start - (int)prefix.length(), prefix + content + suffix });
+        else
+            m_lines.insert(line_it, { offset + start, content.substr(start) + suffix });
+    }
+
+    check_invariant();
+}
+
+void Buffer::erase(const BufferIterator& pos, BufferSize length)
+{
+    BufferIterator end = pos + length;
+    String prefix = m_lines[pos.line()].content.substr(0, pos.column());
+    String suffix = m_lines[end.line()].content.substr(end.column());
+    Line new_line = { m_lines[pos.line()].start, prefix + suffix };
+
+    m_lines.erase(m_lines.begin() + pos.line(), m_lines.begin() + end.line() + 1);
+    m_lines.insert(m_lines.begin() + pos.line(), new_line);
+
+    for (size_t i = pos.line()+1; i < line_count(); ++i)
+        m_lines[i].start -= length;
+
+    check_invariant();
 }
 
 void Buffer::apply_modification(const Modification& modification)
 {
+    const String& content = modification.content;
+    const BufferIterator& pos = modification.position;
+
     switch (modification.type)
     {
     case Modification::Insert:
-        m_content.insert(modification.position.m_position,
+        insert(modification.position, modification.content);
-                         modification.content);
         break;
     case Modification::Erase:
     {
         size_t size = modification.content.size();
         assert(string(modification.position, modification.position + size)
                == modification.content);
-        m_content.erase(modification.position.m_position, size);
+        erase(modification.position, size);
         break;
     }
     default:
         assert(false);
     }
-    update_lines(modification);
+
     for (auto listener : m_modification_listeners)
         listener->on_modification(modification);
 }

src/buffer.hh

@@ -12,6 +12,7 @@ namespace Kakoune
 {
 
 class Buffer;
+class Modification;
 class Window;
 
 typedef int      BufferPos;
@@ -39,8 +40,8 @@ public:
     typedef const value_type& reference;
     typedef std::bidirectional_iterator_tag iterator_category;
 
-    BufferIterator() : m_buffer(NULL), m_position(0) {}
+    BufferIterator() : m_buffer(NULL) {}
-    BufferIterator(const Buffer& buffer, BufferPos position);
+    BufferIterator(const Buffer& buffer, BufferCoord coord);
     BufferIterator& operator=(const BufferIterator& iterator);
 
     bool operator== (const BufferIterator& iterator) const;
@@ -66,11 +67,17 @@ public:
     bool is_end() const;
     bool is_valid() const;
 
+    void update(const Modification& modification);
+
     const Buffer& buffer() const;
 
 private:
+    BufferSize line() const { return m_coord.line; }
+    BufferSize column() const { return m_coord.column; }
+    BufferSize offset() const;
+
     const Buffer* m_buffer;
-    BufferPos     m_position;
+    BufferCoord   m_coord;
     friend class Buffer;
 };
 
@@ -145,8 +152,6 @@ public:
 
     const std::string& name() const { return m_name; }
 
-    const String& content() const { return m_content; }
-
     Window* get_or_create_window();
     void delete_window(Window* window);
 
@@ -166,17 +171,26 @@ public:
     // not on the last one)
     BufferIterator iterator_at_line_end(const BufferIterator& iterator) const;
 
+    const String& line_content(size_t l) const { return m_lines[l].content; }
+
 private:
     friend class BufferIterator;
 
-    std::vector<BufferPos> m_lines;
+    void check_invariant() const;
-    void update_lines(const Modification& modification);
+
+    struct Line
+    {
+        BufferPos start;
+        String    content;
+    };
+    std::vector<Line> m_lines;
+
+    void insert(const BufferIterator& pos, const String& content);
+    void erase(const BufferIterator& pos, BufferSize length);
 
     BufferPos line_at(const BufferIterator& iterator) const;
     BufferSize line_length(BufferPos line) const;
 
-    String m_content;
-
     std::string  m_name;
     const Type   m_type;
 

src/buffer_iterator.inl.hh

@@ -6,10 +6,10 @@
 namespace Kakoune
 {
 
-inline BufferIterator::BufferIterator(const Buffer& buffer, BufferPos position)
+inline BufferIterator::BufferIterator(const Buffer& buffer, BufferCoord coord)
-    : m_buffer(&buffer),
+    : m_buffer(&buffer), m_coord(coord)
-      m_position(std::max(0, std::min(position, (BufferPos)buffer.length())))
 {
+    assert(is_valid());
 }
 
 inline const Buffer& BufferIterator::buffer() const
@@ -20,90 +20,175 @@ inline const Buffer& BufferIterator::buffer() const
 
 inline bool BufferIterator::is_valid() const
 {
-    return m_buffer;
+    return m_buffer and
+           ((line() < m_buffer->line_count() and
+             column() < m_buffer->m_lines[line()].content.length()) or
+            ((line() == m_buffer->line_count() and column() == 0)) or
+             (line() == m_buffer->line_count() - 1 and
+              column() == m_buffer->m_lines.back().content.length()));
 }
 
 inline BufferIterator& BufferIterator::operator=(const BufferIterator& iterator)
 {
     m_buffer = iterator.m_buffer;
-    m_position = iterator.m_position;
+    m_coord = iterator.m_coord;
     return *this;
 }
 
 inline bool BufferIterator::operator==(const BufferIterator& iterator) const
 {
     assert(m_buffer == iterator.m_buffer);
-    return (m_position == iterator.m_position);
+    return (m_coord == iterator.m_coord);
 }
 
 inline bool BufferIterator::operator!=(const BufferIterator& iterator) const
 {
     assert(m_buffer == iterator.m_buffer);
-    return (m_position != iterator.m_position);
+    return (m_coord != iterator.m_coord);
 }
 
 inline bool BufferIterator::operator<(const BufferIterator& iterator) const
 {
     assert(m_buffer == iterator.m_buffer);
-    return (m_position < iterator.m_position);
+    return (m_coord < iterator.m_coord);
 }
 
 inline bool BufferIterator::operator<=(const BufferIterator& iterator) const
 {
     assert(m_buffer == iterator.m_buffer);
-    return (m_position <= iterator.m_position);
+    return (m_coord <= iterator.m_coord);
 }
 
 inline bool BufferIterator::operator>(const BufferIterator& iterator) const
 {
     assert(m_buffer == iterator.m_buffer);
-    return (m_position > iterator.m_position);
+    return (m_coord > iterator.m_coord);
 }
 
 inline bool BufferIterator::operator>=(const BufferIterator& iterator) const
 {
     assert(m_buffer == iterator.m_buffer);
-    return (m_position >= iterator.m_position);
+    return (m_coord >= iterator.m_coord);
+}
+
+inline BufferCoord measure_string(const String& string)
+{
+    BufferCoord res;
+    for (auto c : string)
+    {
+        if (c == '\n')
+        {
+            ++res.line;
+            res.column = 0;
+        }
+        else
+           ++res.column;
+    }
+    return res;
+}
+
+inline BufferCoord advance(const BufferCoord& base, const BufferCoord& offset)
+{
+    if (offset.line == 0)
+        return BufferCoord{base.line, base.column + offset.column};
+    else
+        return BufferCoord{base.line + offset.line, offset.column};
+}
+
+inline void BufferIterator::update(const Modification& modification)
+{
+    const BufferIterator& pos = modification.position;
+    if (*this < pos)
+        return;
+
+    BufferCoord measure = measure_string(modification.content);
+    if (modification.type == Modification::Erase)
+    {
+        BufferCoord end = advance(pos.m_coord, measure);
+        if (m_coord <= end)
+            m_coord = pos.m_coord;
+        else
+        {
+            m_coord.line -= measure.line;
+            if (end.line == m_coord.line)
+                m_coord.column -= measure.column;
+        }
+
+        if (is_end())
+            operator--();
+    }
+    else
+    {
+        assert(modification.type == Modification::Insert);
+        if (pos.line() == line())
+        {
+            BufferCoord end = advance(pos.m_coord, measure);
+            m_coord.column = end.column + column() - pos.column();
+        }
+        m_coord.line += measure.line;
+    }
+    assert(is_valid());
 }
 
 inline BufferChar BufferIterator::operator*() const
 {
     assert(m_buffer);
-    return m_buffer->m_content[m_position];
+    return m_buffer->m_lines[line()].content[column()];
+}
+
+inline BufferSize BufferIterator::offset() const
+{
+    assert(m_buffer);
+    return line() == 0 ? column() : m_buffer->m_lines[line()].start + column();
 }
 
 inline BufferSize BufferIterator::operator-(const BufferIterator& iterator) const
 {
     assert(m_buffer == iterator.m_buffer);
-    return m_position - iterator.m_position;
+    return offset() - iterator.offset();
 }
 
 inline BufferIterator BufferIterator::operator+(BufferSize size) const
 {
     assert(m_buffer);
-    return BufferIterator(*m_buffer, m_position + size);
+    if (size >= 0)
+    {
+        BufferSize o = std::min(m_buffer->length(), offset() + size);
+        for (int i = line() + 1; i < m_buffer->line_count(); ++i)
+        {
+            if (m_buffer->m_lines[i].start > o)
+                return BufferIterator(*m_buffer, { i-1, o - m_buffer->m_lines[i-1].start });
+        }
+        int last_line = m_buffer->line_count() - 1;
+        return BufferIterator(*m_buffer, { last_line, o - m_buffer->m_lines[last_line].start });
+    }
+    return operator-(-size);
 }
 
 inline BufferIterator BufferIterator::operator-(BufferSize size) const
 {
     assert(m_buffer);
-    return BufferIterator(*m_buffer, m_position - size);
+    if (size >= 0)
+    {
+        BufferSize o = std::max(0, offset() - size);
+        for (int i = line(); i >= 0; --i)
+        {
+            if (m_buffer->m_lines[i].start <= o)
+                return BufferIterator(*m_buffer, { i, o - m_buffer->m_lines[i].start });
+        }
+        assert(false);
+    }
+    return operator+(-size);
 }
 
 inline BufferIterator& BufferIterator::operator+=(BufferSize size)
 {
-    assert(m_buffer);
+    return *this = (*this + size);
-    m_position = std::max(0, std::min((BufferSize)m_position + size,
-                                      m_buffer->length()));
-    return *this;
 }
 
 inline BufferIterator& BufferIterator::operator-=(BufferSize size)
 {
-    assert(m_buffer);
+    return *this = (*this - size);
-    m_position = std::max(0, std::min((BufferSize)m_position - size,
-                                      m_buffer->length()));
-    return *this;
 }
 
 inline BufferIterator& BufferIterator::operator++()
@@ -119,13 +204,13 @@ inline BufferIterator& BufferIterator::operator--()
 inline bool BufferIterator::is_begin() const
 {
     assert(m_buffer);
-    return m_position == 0;
+    return m_coord.line == 0 and m_coord.column == 0;
 }
 
 inline bool BufferIterator::is_end() const
 {
     assert(m_buffer);
-    return m_position == m_buffer->length();
+    return offset() == m_buffer->length();
 }
 
 }

src/file.cc

@@ -88,7 +88,9 @@ void write_buffer_to_file(const Buffer& buffer, const std::string& filename)
     if (fd == -1)
         throw file_access_error(filename, strerror(errno));
 
-    const String& content = buffer.content();
+    for (size_t i = 0; i < buffer.line_count(); ++i)
+    {
+        const String& content = buffer.line_content(i);
         ssize_t count = content.length() * sizeof(BufferChar);
         const char* ptr = content.c_str();
 
@@ -101,6 +103,7 @@ void write_buffer_to_file(const Buffer& buffer, const std::string& filename)
             if (written == -1)
                 throw file_access_error(filename, strerror(errno));
         }
+    }
     close(fd);
 }
 

src/selection.cc

@@ -54,36 +54,10 @@ void Selection::merge_with(const Selection& selection)
     m_last = selection.m_last;
 }
 
-static void update_iterator(const Modification& modification,
-                            BufferIterator& iterator)
-{
-    if (iterator < modification.position)
-        return;
-
-    size_t length = modification.content.length();
-    if (modification.type == Modification::Erase)
-    {
-        // do not move length on the other side of the inequality,
-        // as modification.position + length may be after buffer end
-        if (iterator - length <= modification.position)
-            iterator = modification.position;
-        else
-            iterator -= length;
-
-        if (iterator.is_end())
-            --iterator;
-    }
-    else
-    {
-        assert(modification.type == Modification::Insert);
-        iterator += length;
-    }
-}
-
 void Selection::on_modification(const Modification& modification)
 {
-    update_iterator(modification, m_first);
+    m_first.update(modification);
-    update_iterator(modification, m_last);
+    m_last.update(modification);
 }
 
 void Selection::register_with_buffer()