Pegbar alignment #1205

davidlamhauge · 2019-03-30T16:51:05Z

A feature relevant for animators that animate on paper and import their scanned drawings into Pencil2D. If they use the scanners document feeder, they can expect som inaccuracies, so their artwork is not aligned.
This feature alignes the center pegbar of the drawings.
Usage:

select the keyframe that is to be the point of reference. This is often a field guide.
With the select tool, select an area around the center peg hole, that encircles all center peg holes.
Select peg bar alignment in the menu.
The artwork is now horizontally and vertically aligned.

MrStevns · 2019-04-13T08:34:36Z

core_lib/src/graphics/bitmap/bitmapimage.cpp

+            }
+        }
+    }
+    return 10000;


Should return -1 here instead of 10000.

The function 'findLeft' searches for, and returns, the left-most x-value of the peg bar you have selected. This value could in some cases be -1, which is pretty close to the center of the canvas. That's why I chose 10000 as a return value.

Hmm... in that case I'd prefer if we could maybe settle on another return type. Given that the coordinate can be anything, even 10000, returning a custom value that can hold both the intended value but also a potential Errorcode would be better and easier to understand for people trying to read the code too.

So it could be something like...

... will be defined in pencilerror.h struct StatusInt { int value = 0; ErrorCode errorcode; }; and then return that value in your findLeft and findTop top function with an appropriate errorcode. Now you can simply return a StatusInt and catch the appropriate errorcode instead. This could come in handy in many other places, so I think that would be a nice addition to the project.

I have never made a struct before, but I can see that it often is defined as a class.
However, I can't get it to work unless I put it inside the Status class, after the ErrorCode Enum, like this:

ERROR_NEED_AT_LEAST_ONE_CAMERA_LAYER }; struct StatusInt { int value = 0; ErrorCode errorcode; };

Is that the way to do it? It must somehow know what ErrorCode is?

Yeah that's correct, the struct needs to know the origin of the values inside it. Putting the struct above Errorcode isn't allowed because ErrorCode doesn't exist yet.

MrStevns · 2019-04-13T08:34:43Z

core_lib/src/graphics/bitmap/bitmapimage.cpp

+            }
+        }
+    }
+    return 10000;


Should return -1 here instead of 10000.

As above...

MrStevns · 2019-04-13T08:42:51Z

app/src/mainwindow2.cpp

@@ -44,7 +44,11 @@ GNU General Public License for more details.
 #include "soundmanager.h"
 #include "viewmanager.h"

+#include "layer.h"


Core files should not exist in app. Only widgets and UI related should exist here.

MrStevns · 2019-04-13T08:42:59Z

app/src/mainwindow2.cpp

 #include "layercamera.h"
+#include "layerbitmap.h"


Core files should not exist in app. Only widgets and UI related should exist here.

MrStevns · 2019-04-13T08:43:04Z

app/src/mainwindow2.cpp

 #include "layercamera.h"
+#include "layerbitmap.h"
+#include "bitmapimage.h"


Core files should not exist in app. Only widgets and UI related should exist here.

Remember to remove the header files again that you don't use.

MrStevns

Although the feature itself looks good, the code in mainwindow2 will have to be moved elsewhere as you're mixing core_lib and app which is not allowed.

Anything that modifies content and the like without interacting through gui should be move to core_lib instead, you'll have to split up the code so some lies in Editor, then calls the respective classes to make the required changes.

        for (int i = 0; i < sl->count(); i++)
        {
            layerbitmap = static_cast<LayerBitmap*>(mEditor->layers()->findLayerByName(sl->at(i)));
            for (int k = layerbitmap->firstKeyFramePosition(); k <= layerbitmap->getMaxKeyFramePosition(); k++)
            {
                if (layerbitmap->keyExists(k))
                {
                    img = layerbitmap->getBitmapImageAtFrame(k);
                    int tmp_x = img->findLeft(rect, 121);
                    if (tmp_x == 10000)
                    {
                        QMessageBox::information(this, nullptr,
                                                 tr("Peg bar not found at %1, %2").arg(layerbitmap->name()).arg(k),
                                                 QMessageBox::Ok);
                        return;
                    }
                    int tmp_y = img->findTop(rect, 121);
                    img->moveTopLeft(QPoint(img->left() + (peg_x - tmp_x), img->top() + (peg_y - tmp_y)));
                }
            }
        }

I'd make this its own method, calling it adjustContentOfLayer and then another adjustContentOfLayers that calls adjustOfContentOfLayer but with the for loop integrated.

The QMessage boxes will still have to be in the widget though so you'll have to make return values that can execute that still if the alignment fails etc...

Remember that for each time you implement something, it should be possible to write tests for it that doesn't rely on UI. so adjusting layer content should be 100% independent of widgets.

davidlamhauge · 2019-04-17T08:32:15Z

I've separated the functions, so the core functionality is in the core-lib. I've tested it as good as I can, and it works fine.
One day I would like to know how you "...write tests for it that doesn't rely on UI." @candyface

MrStevns

LGTM now aside from a few comments. Although I think an introduction to peg bar registration is needed for new users but that can probably wait till some other time. It's a nice to have thing.

MrStevns · 2019-04-20T08:26:02Z

app/app.pro

@@ -89,7 +90,8 @@ SOURCES += \
    src/spinslider.cpp \
    src/doubleprogressdialog.cpp \
    src/colorslider.cpp \
-    src/checkupdatesdialog.cpp
+    src/checkupdatesdialog.cpp \
+    src/pegbarregistration.cpp


Should use same name convention as the other widget classes, ie. pegbarregistrationdialog because it's a dialog widget.

app/src/mainwindow2.cpp

MrStevns · 2019-04-20T09:04:13Z

app/src/mainwindow2.cpp

 #include "layercamera.h"
+#include "layerbitmap.h"
+#include "bitmapimage.h"


Remember to remove the header files again that you don't use.

davidlamhauge · 2019-04-20T15:29:24Z

I have removed unused headers, and renamed the class to the convention.
It compiles and runs fine, but somehow the checks failed?

MrStevns · 2019-04-20T16:00:33Z

You forgot to rename the _ui file.

../../app/src/pegbaralignmentdialog.cpp:2:35: fatal error: ui_pegbarregistration.h: No such file or directory
 #include "ui_pegbarregistration.h"

davidlamhauge · 2019-04-22T14:23:50Z

I finally got it renamed - I hope...

MrStevns · 2019-07-01T09:06:06Z

I'll be merging this tomorrow if there are no further comments.

Jose-Moreno · 2019-07-01T16:59:27Z

@candyface I can't build David's branch to test right now, it seems to give me a few errors for missing methods on ScribbleArea (i.e deselectAll() and getSelection() )

I'll wait until it's merged to test it as there were some issues I presented to David before and I think they are specific to windows since last time he told me he wasn't getting them at all with the same test resources (e.g drawings were not being properly aligned on first pass, zooming affected alignment, operating on already aligned would mess up the alignment instead of preserving it)

scribblemaniac · 2019-07-01T19:35:18Z

Sounds like there might be some merging issues from what @Jose-Moreno said. Additionally, I would like to review this before it gets merged. It shouldn't take too long so please hold out on merging this right away.

davidlamhauge · 2019-07-01T20:06:06Z

I have made the changes so it can compile now.
Tested it on Ubuntu, and it works as expected.

davidlamhauge · 2019-07-11T08:40:45Z

On discord we've (@Jose-Moreno , @candyface and @scribblemaniac ) had a discussion whether it would be an enhancement if the dialog was modal. With the last merge the dialog is modal. So far just checked on Ubuntu.
As @scribblemaniac suggested it was a challenge programming wise, but with some tweaking I got it to work :-)

MrStevns · 2019-07-13T08:28:04Z

Two quick comments from your observations in discord:

You are able to open several alignment-dialogs. This should be prevented.

This can be prevented by adding a reference to the header and seeing if it exists already, quick fix.

2.2) Apparently there are some issues on the modal thing. In Ubuntu it worked fine, and I could interact with the UI, despite of the dialog. On my MacBook the dialog does not stay on top, but disappears under the MainWindow...?

Add setWindowFlag(Qt::WindowStaysOnTopHint); to your dialog window, this will make it stay on top if it's possible. Works on Mac OS.

MrStevns

Functionality wise I think it looks good, my request of changes are mainly to avoid bloating mainwindow2 with unrelated methods. Additionally there are some rename suggestions and removal of GUI feedback in Editor but aside from that I think it looks good.

This will be a very nice addition to the program for those that rely on imports, great job :)

MrStevns · 2019-07-15T08:33:31Z

app/src/mainwindow2.cpp

+    pegreg = nullptr;
+}
+
+void MainWindow2::updatePegRegLayers()


This function should be moved into the pegbar class. If you need to access Editor and related core functionality, add a void setCore(Editor*) in your header and keep a reference of Editor named "mEditor". Alternatively you can add it as parameter in the constructor.

MrStevns · 2019-07-15T08:34:24Z

app/src/mainwindow2.cpp

+    pegreg->setLayerList(bitmaplayers);
+}
+
+void MainWindow2::updatePegReg()


This too is only related to pegbar so it should moved to that class.

MrStevns · 2019-07-15T08:35:31Z

app/src/mainwindow2.cpp

@@ -1397,3 +1492,27 @@ void MainWindow2::changePlayState(bool isPlaying)
    }
    update();
 }
+
+void MainWindow2::alignPegs()


Same as above, move to pegbar class.

MrStevns · 2019-07-15T08:39:13Z

app/src/mainwindow2.h

@@ -65,6 +66,10 @@ class MainWindow2 : public QMainWindow
    void undoActSetEnabled();
    void updateSaveState();
    void clearRecentFilesList();
+    void pegBarReg();


The intent of the function could be clearer, we're launching a dialog, so the method could be called openPegBarRegDialog or pegBarDialogPressed or similar.

MrStevns · 2019-07-15T08:55:19Z

core_lib/src/interface/editor.cpp

+                retLeft = img->findLeft(rect, 121);
+                if (retLeft.errorcode == Status::FAIL)
+                {
+                    QMessageBox::information(nullptr, nullptr,


I understand what you want to do here but Editor and Core in general should not contain GUI feedback because in case a test is made, the program will get stuck if something like a qmessagebox is triggered. Although we don't have many tests for now, it should always be possible to run a method when implementing in core, without triggering GUI feedback.

I think it would be a good idea to have a general error or warning popup that can be emitted from Core and accepted in Mainwindow2. I'll make a PR on your branch.

Decouple GUI from editor

davidlamhauge · 2019-07-16T12:13:22Z

I have adressed all your requests @candyface.
I forgot to erase some marked out code in MainWindow, but I'll wait to commit this small change, until I know if there will be more requests. Thanks for helping out!

MrStevns

@davidlamhauge Previously I've mainly focused on reviewing the code, now I'll comment on the UI/UX.

There's a bug that if you close the dialog via the window frame (pressing the cross), then you can't open the dialog again. You'll be told it's already open.
Usability wise I don't like that the user doesn't know that the dialog disappears when they've pressed align, I'd rather see the dialog being kept open and the user can choose to close it whenever they feel happy with the result. You could rename the cancel to "Close", so the choices would be "close" ---- "align", then it's fully up to the user whether they want to close or keep refining.
The default state is currently that cancel will be highlighted, this should be "Align Peg Bars"

You can change it like by doing the following:

Cancel is misleading since there's nothing to cancel, clicking on cancel simply closes the dialog again. If this is the intention then it should be named "close" instead, otherwise it should reverse its action.
"pegbar alignment failed!" doesn't give me much indication of what is wrong (in my case I tried to align white pegbars), why did it fail or what can I do to prevent it? Although there are prerequisites, the error message should still try to explain the situation.
A better error message could be "Was not able to find pegholes, make sure you've selected all of them and try again"

Future work

There's room for further improvement. The aligner only works with pegholes that are black, which might not always be the case. Often the result is still offset by one pixel or so but overall it works quite well.

Code wise it looks good to me, well except that bug :)

davidlamhauge · 2019-07-23T17:04:27Z

Thanks for the comments @candyface .

~~The only way I can eliminate the close-button bug, is to disable the close button in the upper corner. If you know another way, please let me know.~~
The feature looks for peg holes that has a gray value of 121 or less, so peg holes could be red, blue or other colors, but they should have a tone that is medium gray or darker. We could add it as a fourth prerequisite and/or the level could be slightly higher (150-200), but we should not look for white pegholes. Nobody else does, and I've never heard of it before.

MrStevns

Looks good to me now! Well done @davidlamhauge and thanks for being patient :)

I'll merge it later if there are no further comments.

davidlamhauge added 5 commits March 17, 2019 10:08

peg bar registration

976d908

return value error corrected

62942df

Ui and final selctions works

b476b0e

Simplified Ui

14a8056

Ui enhancement

c814078

MrStevns reviewed Apr 13, 2019

View reviewed changes

MrStevns requested changes Apr 13, 2019

View reviewed changes

davidlamhauge added 3 commits April 16, 2019 12:12

Separated app and core_lib functions

639d2b6

Added check for alphavalues

373aa18

Merge conflict with master resolved

5f3305a

scribblemaniac added Enhancement import-export labels Apr 19, 2019

MrStevns approved these changes Apr 20, 2019

View reviewed changes

davidlamhauge added 2 commits April 20, 2019 16:12

Removed unnecessary headers. Renamed dialog files

f1ce0ab

Typo fixed

823167d

davidlamhauge added 2 commits April 22, 2019 15:18

Renamed ui class

8a718fa

Naming bug fixed

c9403bc

Merge branch 'master' into pegbar_alignment

b50f169

MrStevns requested review from scribblemaniac and chchwy July 1, 2019 09:01

Made changes deriving from refactoring scribblearea

f21b581

Feature got modal dialog interactively updated

1524402

davidlamhauge added 3 commits July 13, 2019 12:20

Modality, only one dialog, layer update fixed

bd45323

Moved disconnets to closePegreg() function

81e970f

Removed one layer type control. Tab error corrected

029256e

MrStevns requested changes Jul 15, 2019

View reviewed changes

MrStevns and others added 3 commits July 15, 2019 11:34

Decouple GUI from Editor

909752f

Merge pull request #5 from CandyFace/pegbar_aligment_PR

f57617d

Decouple GUI from editor

Moved functions to related file. Review requests met.

3c2c793

MrStevns requested changes Jul 23, 2019

View reviewed changes

Close dialog bug fixed. Button and message text changed.

694d9e0

MrStevns approved these changes Jul 24, 2019

View reviewed changes

MrStevns merged commit 623b585 into pencil2d:master Jul 25, 2019

Jose-Moreno added this to the 0.6.5 milestone Jan 15, 2020

davidlamhauge deleted the pegbar_alignment branch August 6, 2023 13:00

Pegbar alignment #1205

Pegbar alignment #1205

Conversation

davidlamhauge commented Mar 30, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

MrStevns Apr 14, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

MrStevns Apr 14, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

MrStevns left a comment

Choose a reason for hiding this comment

davidlamhauge commented Apr 17, 2019

MrStevns left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

davidlamhauge commented Apr 20, 2019

MrStevns commented Apr 20, 2019

davidlamhauge commented Apr 22, 2019

MrStevns commented Jul 1, 2019

Jose-Moreno commented Jul 1, 2019

scribblemaniac commented Jul 1, 2019

davidlamhauge commented Jul 1, 2019

davidlamhauge commented Jul 11, 2019

MrStevns commented Jul 13, 2019 • edited Loading

MrStevns left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

davidlamhauge commented Jul 16, 2019

MrStevns left a comment • edited Loading

Choose a reason for hiding this comment

Future work

davidlamhauge commented Jul 23, 2019 • edited Loading

MrStevns left a comment • edited Loading

Choose a reason for hiding this comment

MrStevns Apr 14, 2019 •

edited

Loading

MrStevns Apr 14, 2019 •

edited

Loading

MrStevns left a comment •

edited

Loading

MrStevns commented Jul 13, 2019 •

edited

Loading

MrStevns left a comment •

edited

Loading

davidlamhauge commented Jul 23, 2019 •

edited

Loading

MrStevns left a comment •

edited

Loading