Multiline (word wrap, fit width, etc) nodes - should work for dummies to ?

[Arioch The]

VTV was always string in complex owner-draw rich content scenarios

However for simple scenarios it frequently was rather complex beast.

Now migrating a legacy app to a new Delphi/VTV i found a windows, consisting of VDT and having half of default paint method verbatim copied into OnMeasureItem event just to keep track of all the possible elements like checkboxes, etc.

It was a simplistic tree: no columns (-1), no pictures - just default text and checkboxes.

It did not more worked for a newer VTV, so i tried to reimplement it based on SVN VirtualStringTree

...and also making the form resizable at last!

So i want to review few features of VTV design, that seem to me related to being able to easily implement multiline items.

As far as i can see, multiline support boils down to

1) answering "Which Height is required to fit the given text into given width"

2) rendering the text wit htrespect to both width and height

It looks like for both points state is less than ideal

* One problem is that main column (or NoColumn mode) may contain a number of elements beside the "content".

* Another one is that vertical scrollbar may (or may not) affect an available width.

* And the 3rd one is the conflict of responsibilities (maybe that is my error, but since i could not find the clarification in neither chm nor pdf nor comments in sources, then i also see it somewhat a lack.

* Then i feel like OnMeasureXXXX events are somewhat too picky. It should ask for minimum values, not for final values IMHO.

Now i'll try to put more details in this.

======== Scroll bars =========

1. Usual mode for Windows apps is to only show them when required. That means that at any given time the scrollbar may or may not be present.

Then how should developer know about it ?

One possible answer would be "he does not need to, scrollbars are subtracted from clientwidth/clientheight/clientrect already). Like it is claimed to be done on VCL TCustomForm level. However that does not seem to happen in VTV (judging by its sources, which compares RangeX/RangeY with Client*** without extra checks for scrollbars)

Then i make a proposal to add a porperties, so that it would be clear whether scrollbars are visible this very moment and if they are - how much do they subtract from effectively visible area of a control.

Proposal is:

type
TVirtualStringTree = class(VirtualTrees.TVirtualStringTree)
protected
    FScrollBarsBands: TSize;
    procedure DoShowScrollbar(Bar: Integer; Show: Boolean); override;
public
    property ScrollBarVisibleWidth: integer read FScrollBarsBands.cx;
    property ScrollBarVisibleHeight: integer read FScrollBarsBands.cx;
end;

procedure TVirtualStringTree.DoShowScrollbar(Bar: Integer; Show: Boolean);
var _H, _V: boolean;
begin
   inherited;

   _H := Bar in [SB_HORZ, SB_BOTH];
   _V := Bar in [SB_VERT, SB_BOTH];

  if Show then begin
      if _H then FScrollBarsBands.cx := GetSystemMetrics(SM_CXHSCROLL);
      if _V then FScrollBarsBands.cy := GetSystemMetrics(SM_CXVSCROLL);
  end else begin
     if _H then FScrollBarsBands.cx := 0;
     if _V then FScrollBarsBands.cy := 0;
  end;
end;

PS. yes, One can directly compare RangeX with ClientWidth, then double-check this against ScrollOptions but that is "leaking implementation" and so it is both tedious, counter-intuitive and problematic for future compatibility.

2. I also did not tested if items re-measuring can be auto-triggered by VTV resizing without maualyl plumbing this into OnResize event. Probably not.

=========== VTV-managef elements plus Events andrResponsibilities ==========

VDT has OnMeasureItem event

VST adds OnMeasureTextWidth/Height events

Correlation between those stays unclear to me.

VST Demo (d:\DelphiProjects\Libs\VTW\trunk\Demos\Advanced\MultilineDemo.pas) uses base OnMeasureItem event and ignore more specific new OnMeasureItem events.

Also it suggests to use VST.ComputeNodeHeight method which i have problems working with (actually it seems to imply that in NoColumn mode there is no tree but a flat list with no indents).

Well, okay, i can use TCanvas.TextHeight and all that, but i then need to know the actual Width, available to the node.

And i don't want to copy half of WM_PAINT handler into my app - that would be fragile and too complex.

I just want to know which gaps would VTV take for scrollbars or checkboxes or state images or tree lines or margins or cell margins or indentation or... or whatever, now and in future.

So, there seems to be that GetDisplayRect(Node, -1, true, false, true); method...

But it is that complex that it probably just could not be fast (when i only need TRect.Left and nothing else!).

Worse: attempt to use it in OnMeasureTextHight handler immediately leads to stack overflow. This IMHO is the clearest way that GetdisplatRect is a heavy non-ubiquitous function, alas.

This problem IMHO manifests itself in VTV sources - just grepping for "FCheckImages.Width" text shows a dozen of places where the same left offset calculations are copy-pasted time and again, instead of being unified in some service function (or into PVirtualNode^ field).

Or maybe there is some reason that the code should actually be copy-pasted instead of keeping it in a single place of responsibility?

Side note: BTW, why is GetNodeIndent a memory-jumping loop, instead of just one more feld in PVirtualNode ? Doesn't it impact speed and CPU cache ?

SideNode 2: it seems that OnMeasureTextHight and OnMeasuretextWidth are always called in pair, then why they are separate events ? doesn't it only complicate things? 

I think OnMeasureItem or OnMeasureTextHeight methods should include Width parameter or Left and Right boundaries.

But if it seems hard due to compatibiltiy issues, then VTV could implement function like LeftOffset(Node,Column).

PS. And ScrollBarEffectiveWidth as described above.

=== dimensions: minimal or final =====

Now about minimums.

NodeHeight := vst.ComputeNodeHeight(TargetCanvas, Node, -1);

Okay ? Not, if you're running display with 120 DPI fonts instead of old Windows habit of 96DPi.

The checkboxes would just be clipped!

NodeHeight := vst.ComputeNodeHeight(TargetCanvas, Node, -1);
NodeHeight := Max( vst.CheckImages.Height + 2, NodeHeight);

That works better... but it is against leaking implementation details. And yet not enough (what if [+] marks or other pictures would be even taller than those checkboxes?) 

So again it looks like either a NodeMinimalHeight function is needed or a special extra parameter into event handler, specifying whether the height should be reduced or kept, if custom height is less than a default one.

http://uploadmb.com/dw.php?id=1378385247&/vst_multiline_should_be_simple.zip
This project is already somewhat too complex, but...

Run it in Delphi XE2 on Win7 / 120DPI fonts and play wit hthe form's width.

Vertical scrollbar tends to appear way too late.

Text strings are "dancing", their vertical alignment is continualyl switchign middle to top and back

Some strings are word wrappedand some are not.

...somewhat a weird behaviour for the task seems to be conceptually clear and concise.

[Arioch The]

> So, there seems to be that GetDisplayRect(Node, -1, true, false, true); method...

But it is that complex that it probably just could not be fast (when i only need TRect.Left and nothing else!).

Worse: attempt to use it in OnMeasureTextHight handler immediately leads to stack overflow. This IMHO is the clearest way that GetdisplatRect is a heavy non-ubiquitous function, alas.

One more stack blower, based on internet advice

procedure TForm8.vstBeforeCellPaint(Sender: TBaseVirtualTree;
TargetCanvas: TCanvas; Node: PVirtualNode; Column: TColumnIndex;
CellPaintMode: TVTCellPaintMode; CellRect: TRect; var ContentRect: TRect);
begin
    Exclude(Node.States, vsHeightMeasured);
    vst.MeasureItemHeight(TargetCanvas, Node);
end; 

nice ?

Not if 

procedure TForm8.vstMeasureItem(Sender: TBaseVirtualTree;
TargetCanvas: TCanvas; Node: PVirtualNode; var NodeHeight: Integer);
var r: TRect;
begin
r := vst.GetDisplayRect(Node, -1, true, false, true);

...

[Joachim Marder]

Hi.

Would you mind opening issues for that at https://code.google.com/p/virtual-treeview/issues/

Please try to break your suggestions into as many issue as possible.

Thanks and best regards,

Joachim

2013/9/5 Arioch The <ario...@gmail.com>

--
You received this message because you are subscribed to the Google Groups "Virtual Treeview" group.
To unsubscribe from this group and stop receiving emails from it, send an email to virtual-treevi...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Arioch The]

there is only two ready suggestions, other notes require talkign and thinking about

четверг, 5 сентября 2013 г., 19:59:22 UTC+4 пользователь joachim.marder написал:

[Arioch The]

Also, if you're really "last active developer" then mayeb switch from SVN to some DVCS like Git ?

SVN makes branching and history reading much more slow and painful process, as it is always done by fetching from server.

For example i can hardly try a branch or two locally trying to implement something and act how it would go.

SVN is only good here for anonymous checkouts and one batch of local changes.

And even getting SVN access would not help as we here have very limited internet traffic volumes.

[Joachim Marder]

Hmm, as far as I can see Google Code supports Git as well. However I personally do not have experience with GIT and I am not eager to switch to an new VCS as I have a lot of tools and plugins for SVN. Any comments on this from other team members?

2013/9/6 Arioch The <ario...@gmail.com>

--

[Arioch The]

.GetDisplayRect seems non-fit to determine effective width of the item in NoColumn mode, due to

if Column > NoColumn then

...

else

  OffsetRect(Result, -FEffectiveOffsetX, FOffsetY);

четверг, 5 сентября 2013 г., 17:00:43 UTC+4 пользователь Arioch The написал:

[Arioch The]

Would you mind opening issues for that at https://code.google.com/p/virtual-treeview/issues/

Please try to break your suggestions into as many issue as possible.

i opened few, but i don't feel bug-tracker format would be good to discuss those things, they look complex and globall;y-changing to me,

https://code.google.com/p/virtual-treeview/issues/detail?id=367 

https://code.google.com/p/virtual-treeview/issues/detail?id=368 

https://code.google.com/p/virtual-treeview/issues/detail?id=369 

[Arioch The]

> Hmm, as far as I can see Google Code supports Git as well. However I personally do not

> have experience with GIT and I am not eager to switch to an new VCS as I have a lot of tools

> and plugins for SVN. Any comments on this from other team members?

As a matter of interest:

1) switch was already requested: https://code.google.com/p/virtual-treeview/issues/detail?id=344

2) the transition is already happening de facto, but generating an abandoned outdated repositories

https://github.com/salvadorbs/Asuite/tree/master/3p/Virtual%20Treeview

https://bitbucket.org/ltogniolli/xananews/src/4843ea8a69d46b44f2985ca3a288fa8e354c0540/components/thirdparty/VirtualTreeView/Source?at=default

https://bitbucket.org/esasse/xananews/src/8a974552ac893b1a83f3577f575f928a90f2e16c/components/thirdparty/VirtualTreeView/Source?at=default

SVN is reasonably fit for a development as "little close company" plus many passive consumers.

But if you're the last of a team, then it mostly is a barrier for anyone else to play and try changing things, SVN is for consuming.

Perhaps you glance through http://www.ericsink.com/vcbe/

[Arioch The]

BTW, you seem to be registered twice at 

https://www.ohloh.net/people?q=joachim%20marder

perhaps you an change URL of repository http://www.ohloh.net/p/VirtualTreeview/enlistments

to  http://virtual-treeview.googlecode.com/svn/ 

[Joachim Marder]

I didn't mean to judge GIT. The time it costs me to update my whole toolchain of batches, makefiles, CI-Server, IDE-plugins etc. would simply be magnitudes greater than the time GIT could save me compared to stick with the good old SVN.

2013/9/6 Arioch The <ario...@gmail.com>

--

[Joachim Marder]

I actually *never* registered on that site.

2013/9/6 Arioch The <ario...@gmail.com>