D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate added a comment.


  In D12016#323833 , @volkov wrote:
  
  > I get the following warning from valgrind because of this change:
  
  
  Fix in D15420 

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: kwrite-devel, kde-frameworks-devel, volkov, anthonyfieroni, mwolff, brauch, 
cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, demsking, sars, 
dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Alexander Volkov]

volkov added a comment.
Herald added subscribers: kde-frameworks-devel, kwrite-devel.


  I get the following warning from valgrind because of this change:
  
==3621== Conditional jump or move depends on uninitialised value(s)
==3621==at 0x4FF3605: 
KateViewAccessible::positionFromCursor(KateViewInternal*, KTextEditor::Cursor 
const&) const (kateviewaccessible.h:201)
==3621==by 0x4FE41CA: 
KateViewInternal::documentTextInserted(KTextEditor::Document*, 
KTextEditor::Range const&) (kateviewinternal.cpp:3830)
==3621==by 0x967762A: call (qobjectdefs_impl.h:376)
==3621==by 0x967762A: QMetaObject::activate(QObject*, int, int, void**) 
(qobject.cpp:3754)
==3621==by 0x50C25E6: 
KTextEditor::DocumentPrivate::textInserted(KTextEditor::Document*, 
KTextEditor::Range const&) (moc_katedocument_6APBFLF6VF33HC.cpp:985)
==3621==by 0x4F498DE: KTextEditor::DocumentPrivate::editInsertText(int, 
int, QString const&) (katedocument.cpp:1235)
==3621==by 0x4F4DF67: 
KTextEditor::DocumentPrivate::insertText(KTextEditor::Cursor const&, QString 
const&, bool) (katedocument.cpp:780)
==3621==by 0x4F51838: 
KTextEditor::DocumentPrivate::typeChars(KTextEditor::ViewPrivate*, QString 
const&) (katedocument.cpp:3024)
==3621==by 0x4FE9BF4: KateViewInternal::keyPressEvent(QKeyEvent*) 
(kateviewinternal.cpp:2522)
==3621==by 0x4FF1F12: KateViewInternal::eventFilter(QObject*, QEvent*) 
(kateviewinternal.cpp:2370)
==3621==by 0x96493EC: 
QCoreApplicationPrivate::sendThroughObjectEventFilters(QObject*, QEvent*) 
(qcoreapplication.cpp:1174)
==3621==by 0x8284834: QApplicationPrivate::notify_helper(QObject*, 
QEvent*) (in /usr/lib/x86_64-linux-gnu/libQt5Widgets.so.5.11.0)
==3621==by 0x828D30E: QApplication::notify(QObject*, QEvent*) (in 
/usr/lib/x86_64-linux-gnu/libQt5Widgets.so.5.11.0)
==3621==  Uninitialised value was created by a heap allocation
==3621==at 0x4C2C25F: operator new(unsigned long) (in 
/usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==3621==by 0x4FE3EF0: accessibleInterfaceFactory(QString const&, 
QObject*) (kateviewaccessible.h:288)
==3621==by 0x8A70D2C: QAccessible::queryAccessibleInterface(QObject*) 
(in /usr/lib/x86_64-linux-gnu/libQt5Gui.so.5.11.0)
==3621==by 0x8A7172E: QAccessibleEvent::accessibleInterface() const (in 
/usr/lib/x86_64-linux-gnu/libQt5Gui.so.5.11.0)
==3621==by 0x8A71BFC: 
QAccessible::updateAccessibility(QAccessibleEvent*) (in 
/usr/lib/x86_64-linux-gnu/libQt5Gui.so.5.11.0)
==3621==by 0x82C0A72: QWidgetPrivate::show_helper() (in 
/usr/lib/x86_64-linux-gnu/libQt5Widgets.so.5.11.0)
==3621==by 0x82C3CC4: QWidget::setVisible(bool) (in 
/usr/lib/x86_64-linux-gnu/libQt5Widgets.so.5.11.0)
==3621==by 0x82C0957: QWidgetPrivate::showChildren(bool) (in 
/usr/lib/x86_64-linux-gnu/libQt5Widgets.so.5.11.0)
==3621==by 0x82C09DE: QWidgetPrivate::show_helper() (in 
/usr/lib/x86_64-linux-gnu/libQt5Widgets.so.5.11.0)
==3621==by 0x82C0946: QWidgetPrivate::showChildren(bool) (in 
/usr/lib/x86_64-linux-gnu/libQt5Widgets.so.5.11.0)
==3621==by 0x82C09DE: QWidgetPrivate::show_helper() (in 
/usr/lib/x86_64-linux-gnu/libQt5Widgets.so.5.11.0)
==3621==by 0x82C3CC4: QWidget::setVisible(bool) (in 
/usr/lib/x86_64-linux-gnu/libQt5Widgets.so.5.11.0)

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: kwrite-devel, kde-frameworks-devel, volkov, anthonyfieroni, mwolff, brauch, 
cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, demsking, sars, 
dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

This revision was automatically updated to reflect the committed changes.
Closed by commit R39:81996f55b1ab: [ktexteditor] much faster positionFromCursor 
(authored by jtamate).

CHANGED PRIOR TO COMMIT
  https://phabricator.kde.org/D12016?vs=31853=33615#toc

REPOSITORY
  R39 KTextEditor

CHANGES SINCE LAST UPDATE
  https://phabricator.kde.org/D12016?vs=31853=33615

REVISION DETAIL
  https://phabricator.kde.org/D12016

AFFECTED FILES
  src/view/kateviewaccessible.h
  src/view/kateviewinternal.cpp

To: jtamate, #kate, cullmann, #frameworks
Cc: anthonyfieroni, mwolff, brauch, cullmann, #frameworks, michaelh, 
kevinapavew, ngraham, bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Christoph Cullmann]

cullmann added a comment.


  If the stuff still works for you and is fast now, I am still for merging that.
  If somebody has time to do a even nicer integration, a new request can be 
done afterwards.

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: anthonyfieroni, mwolff, brauch, cullmann, #frameworks, michaelh, 
kevinapavew, ngraham, bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate added a comment.


  In D12016#257862 , @cullmann wrote:
  
  > Actually, given the functionality is only used there, if that code is in 
the view or in that helper class makes no real difference, or?
  
  
  I've tried to extend KateViewInternal in such a way that positionFromCursor 
in KateViewAccessible becomes unnecessary, but it became a mess.
  The problem is that all the "coordinates" used there are Cursors, that are in 
the form line+column. 
  But in KateViewAccessible it uses characters from the beginning, and caching 
only in one place (positionFromCursor) does not create a mess.

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: anthonyfieroni, mwolff, brauch, cullmann, #frameworks, michaelh, 
kevinapavew, ngraham, bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Christoph Cullmann]

cullmann added a comment.


  Actually, given the functionality is only used there, if that code is in the 
view or in that helper class makes no real difference, or?

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: anthonyfieroni, mwolff, brauch, cullmann, #frameworks, michaelh, 
kevinapavew, ngraham, bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Anthony Fieroni]

anthonyfieroni added a comment.


  In D12016#257325 , @jtamate wrote:
  
  > Do you mean to extend KateViewInternal in such a way that 
positionFromCursor in KateViewAccessible becomes unnecessary?
  
  
  Yes, mainly it should be a tiny helper.

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: anthonyfieroni, mwolff, brauch, cullmann, #frameworks, michaelh, 
kevinapavew, ngraham, bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate added a comment.


  In D12016#257324 , @anthonyfieroni 
wrote:
  
  > Maybe i miss something, but i think if something should be extended it's 
view not accessible.
  
  
  Do you mean to extend KateViewInternal in such a way that positionFromCursor 
in KateViewAccessible becomes unnecessary?

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: anthonyfieroni, mwolff, brauch, cullmann, #frameworks, michaelh, 
kevinapavew, ngraham, bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Anthony Fieroni]

anthonyfieroni added a comment.


  Maybe i miss something, but i think if something should be extended it's view 
not accessible.

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: anthonyfieroni, mwolff, brauch, cullmann, #frameworks, michaelh, 
kevinapavew, ngraham, bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Christoph Cullmann]

cullmann added a comment.


  Hmm, I don't understand that comment?
  Now you have one (or more) object per view, or?
  That looks sane to me.
  Or do I misunderstand how that works?

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: anthonyfieroni, mwolff, brauch, cullmann, #frameworks, michaelh, 
kevinapavew, ngraham, bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Anthony Fieroni]

anthonyfieroni added a comment.


  So design is broken,for me. Function should be static and cursor from view 
(KateViewInternal) should be compared against cursor parameter.

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: anthonyfieroni, mwolff, brauch, cullmann, #frameworks, michaelh, 
kevinapavew, ngraham, bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Christoph Cullmann]

cullmann accepted this revision.
cullmann added a comment.
This revision is now accepted and ready to land.


  I am ok with it as it is ;=)
  I would have been ok with the version that still inherits from QObject, too, 
but now that one has done all the hassle to avoid that ok ;)

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: mwolff, brauch, cullmann, #frameworks, michaelh, kevinapavew, ngraham, 
bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Milian Wolff]

mwolff added a comment.


  lgtm, @cullmann ?
  
  regarding the accessibility interface, I agree that it's out of the scope of 
this patch. I'd say let's keep it like that for now...

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: mwolff, brauch, cullmann, #frameworks, michaelh, kevinapavew, ngraham, 
bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate marked an inline comment as done.
jtamate added a comment.




INLINE COMMENTS

> mwolff wrote in kateviewaccessible.h:191
> I know the old code used int already, but shouldn't this better be a quint64 
> as it's a file offset?

All the methods of QAccesibeTextInrface 
 expect an int.
What should be done if the document has more data than INT_MAX?
deactivate the class and show a warning? In any case, I think this is out of 
scope of this patch.

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: mwolff, brauch, cullmann, #frameworks, michaelh, kevinapavew, ngraham, 
bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate updated this revision to Diff 31853.
jtamate marked 5 inline comments as done.
jtamate added a comment.


  Included more comments.
  Addressed style coding.
  
  In a callgrind log file with 1606189 lines, at the end of the file, pressing 
PageUp, using the code included at the end:
  
  old implementation elapsed =  122
  new implementation elapsed =  0
  old implementation elapsed =  119
  new implementation elapsed =  0
  
  After some insertions:
  old implementation elapsed =  137
  new implementation elapsed =  118
  old implementation elapsed =  123
  new implementation elapsed =  0
  
  - src/view/kateviewaccessible.h
  
  +++ src/view/kateviewaccessible.h
  @@ -29,6 +29,8 @@
   #include 
   #include 
   #include 
  +#include 
  +#include 
  
  /**
  
  - This class implements a QAccessible-interface for a KateViewInternal.
  
  @@ -194,6 +196,9 @@ public:
  
 */
int positionFromCursor(KateViewInternal *view, const KTextEditor::Cursor 
) const
{
  
  +QElapsedTimer t;
  +t.start();
  +
  
int pos = m_lastPosition;
const auto *doc = view->view()->document();
 
  
  @@ -229,6 +234,22 @@ public:
  
m_lastCursor = cursor;
m_lastPosition = pos;
 
  
  +qWarning() << "new implementation elapsed = " << t.restart();
  +// previous implementation
  +int _pos = 0;
  +for (int _line = 0; _line < cursor.line(); ++_line) {
  +// length of the line plus newline
  +_pos += view->view()->document()->line(_line).size() + 1;
  +}
  +_pos += cursor.column();
  +qWarning() << "old implementation elapsed = " << t.elapsed();
  +
  +return pos;
  +if (_pos != (pos + cursor.column())) {
  +qWarning() << "implementations differ, old=" << _pos << " new=" 
<<
  +pos + cursor.column();
  +}
  +
  
return pos + cursor.column();
}

REPOSITORY
  R39 KTextEditor

CHANGES SINCE LAST UPDATE
  https://phabricator.kde.org/D12016?vs=31760=31853

REVISION DETAIL
  https://phabricator.kde.org/D12016

AFFECTED FILES
  src/view/kateviewaccessible.h
  src/view/kateviewinternal.cpp

To: jtamate, #kate, cullmann, #frameworks
Cc: mwolff, brauch, cullmann, #frameworks, michaelh, kevinapavew, ngraham, 
bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Milian Wolff]

mwolff added a comment.


  some nits, otherwise lgtm assuming all tests pass
  
  bonus points for a proper benchmark that shows the before/after gain in hard 
numbers

INLINE COMMENTS

> kateviewaccessible.h:191
>  {
> -int pos = 0;
> -for (int line = 0; line < cursor.line(); ++line) {
> -// length of the line plus newline
> -pos += view->view()->document()->line(line).size() + 1;
> +int pos = m_lastPosition;
> +const KTextEditor::Document *doc = view->view()->document();

I know the old code used int already, but shouldn't this better be a quint64 as 
it's a file offset?

> kateviewaccessible.h:192
> +int pos = m_lastPosition;
> +const KTextEditor::Document *doc = view->view()->document();
> +

const auto *doc =

> kateviewaccessible.h:204
> +m_lastView = view;
>  }
> +else {

style: join with next line

> kateviewaccessible.h:215
> +pos += cursor.line() - m_lastCursor.line();
> +}
> +else {

dito

> kateviewaccessible.h:231
>  private:
> +
>  inline KateViewInternal *view() const

unrelated change?

> kateviewaccessible.h:270
> +mutable KTextEditor::Cursor m_lastCursor;
> +// m_lastPosition stores the positionFromCursor, with the cursor always 
> in column 0
> +mutable int m_lastPosition;

maybe also mention that this is a file offset?

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: mwolff, brauch, cullmann, #frameworks, michaelh, kevinapavew, ngraham, 
bruns, demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate updated this revision to Diff 31760.
jtamate added a comment.


  No QObject inheritance, only a QMetaObject::Connection member to disconnect 
the connection.
  As this is a non exported class, use the same method name and make it 
non-static.
  KateViewInternal::cursorMoved checks if QAccessible::isActive().

REPOSITORY
  R39 KTextEditor

CHANGES SINCE LAST UPDATE
  https://phabricator.kde.org/D12016?vs=31737=31760

REVISION DETAIL
  https://phabricator.kde.org/D12016

AFFECTED FILES
  src/view/kateviewaccessible.h
  src/view/kateviewinternal.cpp

To: jtamate, #kate, cullmann, #frameworks
Cc: brauch, cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, 
demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Sven Brauch]

brauch added a comment.


  Oh and, you do not need to inherit QObject to use connect; you can connect to 
a lambda calling the member function AFAIK or so. Just omit the third argument 
in connect(). What you lose by doing this is the automatic disconnect of the 
connection when the receiver object is deleted, so make sure that doesn't 
happen.

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: brauch, cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, 
demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Sven Brauch]

brauch added a comment.


  Re. binary compatibility: should be fine because this class is not exported 
(no KTEXTEDITOR_EXPORT macro).

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: brauch, cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, 
demsking, sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate updated this revision to Diff 31737.
jtamate edited the summary of this revision.
jtamate added a comment.


  Implemented now as object method instead of static.
  Changed the calls from KateViewInternal from static to use the reference the 
method QAccessible::queryAccessibleInterface(this) returns.

REPOSITORY
  R39 KTextEditor

CHANGES SINCE LAST UPDATE
  https://phabricator.kde.org/D12016?vs=31733=31737

REVISION DETAIL
  https://phabricator.kde.org/D12016

AFFECTED FILES
  src/view/kateviewaccessible.h
  src/view/kateviewinternal.cpp

To: jtamate, #kate, cullmann, #frameworks
Cc: cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, demsking, 
sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Christoph Cullmann]

cullmann added a comment.


  You can use a mutable member for that.

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, demsking, 
sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate added a comment.


  In D12016#243079 , @cullmann wrote:
  
  > Hi, must the variable be static or would an object member be good enough?
  
  
  It can't be a member object, because some methods that call 
positionFromCursor are const. I tried :-(

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, demsking, 
sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Christoph Cullmann]

cullmann added a comment.


  Hi, must the variable be static or would an object member be good enough?

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann, #frameworks
Cc: cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, demsking, 
sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate updated this revision to Diff 31733.
jtamate edited the summary of this revision.
jtamate edited the test plan for this revision.
jtamate added a reviewer: Frameworks.
jtamate added a comment.


  Cached the position in static variables of KateViewAccessible.
  The cache is invalidated when the signal Document::textChanged is received.
  Unfortunately, KateViewAccessible must inherit also QObject or connect will 
not work, because KateViewAccessible didn't inherit QObject.
  **Is this change Binary compatible?**
  
  At first I thought the changes will come in setText, but in my tests it has 
never been called.
  
  To check the correctness of the result, I've run the fast and slow paths in 
kate, with several windows over the same document, changing it in sereveral 
ways, and the results always match.

REPOSITORY
  R39 KTextEditor

CHANGES SINCE LAST UPDATE
  https://phabricator.kde.org/D12016?vs=31579=31733

REVISION DETAIL
  https://phabricator.kde.org/D12016

AFFECTED FILES
  src/CMakeLists.txt
  src/view/kateviewaccessible.cpp
  src/view/kateviewaccessible.h

To: jtamate, #kate, cullmann, #frameworks
Cc: cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, demsking, 
sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Christoph Cullmann]

cullmann requested changes to this revision.
cullmann added a comment.
This revision now requires changes to proceed.


  Hi,
  I appreciate the intend to speed this up.
  But I think, to do this proper, it might make more sense to cache the current 
position in the KateViewAccessible and invalidate it on changes to the document.
  The way you do it now, it might not be correct if text changes occur 
in-between. (if I don't overlook something)

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

To: jtamate, #kate, cullmann
Cc: cullmann, #frameworks, michaelh, kevinapavew, ngraham, bruns, demsking, 
sars, dhaumann

D12016: [ktexteditor] much faster positionFromCursor

[Jaime Torres Amate]

jtamate created this revision.
jtamate added a reviewer: Kate.
Restricted Application added projects: Kate, Frameworks.
Restricted Application added a subscriber: Frameworks.
jtamate requested review of this revision.

REVISION SUMMARY
  Use static variables to store information about last call.
  As most of the time the cursor is close to the previous position, avoid to 
calculate every time the size of the lines from the beginning of the document, 
just calculate it from the previous cursor position.

TEST PLAN
  Open a callgrind log file with 2.840.605 lines
  Go to the end of the file, and then play with the PageUp, PageDown and cursor 
keys.
  Before: it was unable to do continuous PageUp refreshes.
  After: the PageUp at any part of the file as as fast as in the beginning of 
the document.

REPOSITORY
  R39 KTextEditor

REVISION DETAIL
  https://phabricator.kde.org/D12016

AFFECTED FILES
  src/view/kateviewaccessible.h

To: jtamate, #kate
Cc: #frameworks, michaelh, kevinapavew, ngraham, bruns, demsking, cullmann, 
sars, dhaumann