Aliasing in AutomationPatternView and AutomationEditor fixed #3386

WalterSmuts · 2017-02-26T15:44:24Z

Exactly the same as this pull request: #3385 but now from a different branch on my side. (Sorry still learning) It's referencing issue #2688

tresf · 2017-02-26T15:47:46Z

Sorry still learning

No need to apologize. There's no easy way to switch off of a master branch, this was the right thing to do. The only portion that may be worth carrying over is any remaining discussion items such as this:

Before:

After:

"After many hours of trying different things I am quite happy with what happens when the following code changes are applied. It solves the anti-aliasing issue and by drawing polygons instead of rectangles and uses the anti aliasing hint. I also made sure there are no shading inconsistencies.

For some reason however it sometimes doesn't want to draw an automation that is not linked to anything. I believe it is because the values ranges are between 0 and 1 and not 0 and 100 but this is also the case for the master branch and for some reason it works. Can anyone shed some light on it?"

Umcaruje · 2017-02-26T16:47:08Z

For some reason however it sometimes doesn't want to draw an automation that is not linked to anything. I believe it is because the values ranges are between 0 and 1 and not 0 and 100 but this is also the case for the master branch and for some reason it works. Can anyone shed some light on it?

By changing all the QPoints to QPointFs you solve the problem, but a new glitch occurs:

There's again transparency in the painter paths from the huge amount of transforming and antialiasing that happens here.

Could we try to push all the values into a singular path and then draw that path?

WalterSmuts · 2017-02-26T19:30:13Z

Ok awesome! Seems to work now without and shading inconsistencies or leaving out somethings... Now all that needs to be done is the same for the editor itself. @Umcaruje Thanks for pointing out the QPointF thing. It was driving me crazy.

Umcaruje · 2017-02-26T19:42:53Z

src/gui/AutomationPatternView.cpp

-				p.fillRect( QRectF( x1, 0.0f, x2 - x1, value ), col );
-			}
+		}
+		path.lineTo(x_base + ((it + 1).key()) * ppt / MidiTime::ticksPerTact(),values[(it + 1).key() - 1 - it.key()]);


ppt / MidiTime::ticksPerTact() appears a lot of time in the code, you could make it a seperate variable so we don't calculate this all the time.

WalterSmuts · 2017-02-26T20:40:25Z

ppt / MidiTime::ticksPerTact() appears a lot of time in the code, you could make it a seperate variable so we don't calculate this all the time.

Done!

zonkmachine · 2017-02-26T21:01:09Z

src/gui/AutomationPatternView.cpp

@@ -278,6 +278,7 @@ void AutomationPatternView::paintEvent( QPaintEvent * )

 	const float y_scale = max - min;
 	const float h = ( height() - 2 * TCO_BORDER_WIDTH ) / y_scale;
+	const int MidiTpT = MidiTime::ticksPerTact();


I think what was suggested was to do away with a computation by doing it once, such as:
const int MidiTpT = ppt / MidiTime::ticksPerTact();

WalterSmuts · 2017-02-26T21:35:01Z

Done... I was quite silly there

Umcaruje · 2017-02-26T23:38:11Z

Cool, could you also adjust the spaces in parenthesis to be consistent with the rest of the file? I know this is not required anymore, but it really stands out.

Also would you do the same rendering process for the automation editor in this PR? So we could kill two birds with one stone.

Umcaruje · 2017-02-27T10:13:27Z

src/gui/AutomationPatternView.cpp

@@ -278,6 +278,7 @@ void AutomationPatternView::paintEvent( QPaintEvent * )

 	const float y_scale = max - min;
 	const float h = ( height() - 2 * TCO_BORDER_WIDTH ) / y_scale;
+	const float ppTact  = ppt / MidiTime::ticksPerTact();


This should actually be named ppTick, cause ppt already means pixels per tact, and you're dividing it with the number of ticks in a tact 🙂

What is a tact?

It's the german/european word for a Bar in music: https://en.wikipedia.org/wiki/Bar_(music)

Umcaruje · 2017-02-27T10:14:21Z

src/gui/AutomationPatternView.cpp

 			if( x1 > ( width() - TCO_BORDER_WIDTH ) ) break;
+			float value = values[ i - it.key() ];
+			path.lineTo( QPointF( x1,value ) );


There should be a space after the comma: x1, value

Umcaruje · 2017-02-27T10:14:58Z

src/gui/AutomationPatternView.cpp

-						MidiTime::ticksPerTact();
-			const float x2 = x_base + (i + 1) * ppt /
-						MidiTime::ticksPerTact();
+			const float x1 = x_base + i * ppTact;


as there is no x2 now, this variable could be called just x to avoid confusion.

Umcaruje · 2017-02-27T10:15:30Z

src/gui/AutomationPatternView.cpp

-				p.fillRect( QRectF( x1, 0.0f, x2 - x1, value ), col );
-			}
+		}
+		path.lineTo( x_base + ( ( it + 1 ).key() ) * ppTact,values[ ( it + 1 ).key() - 1 - it.key() ] );


same here, space after comma

Umcaruje · 2017-02-27T10:17:53Z

src/gui/AutomationPatternView.cpp

-			}
+		}
+		path.lineTo( x_base + ( ( it + 1 ).key() ) * ppTact,values[ ( it + 1 ).key() - 1 - it.key() ] );
+		path.lineTo( x_base + ( ( it + 1 ).key() ) * ppTact,0.0f );


same as previous comment

Umcaruje · 2017-02-27T10:44:55Z

src/gui/AutomationPatternView.cpp

-		for( int i = it.key(); i < (it + 1).key(); i++ )
+
+		QPainterPath path;
+		QPointF origin = QPointF(x_base + it.key() * ppTick,0.0f);


Missing spaces in parenthesis (sorry I somehow missed this one)

*( sorry I somehow missed this one )

WalterSmuts · 2017-02-27T15:37:48Z

Ok so I managed to add anti-aliasing in the Automation Editor with some minor points that need addressing:

The Shading is inconsistent (Although it was inconsistent in some themes before)
I'm not using the is_selected Boolean (Can't seem to find it's use)
I don't catch the points if they aren't in the "Visible area" as was done before

This is the result thus far:

After:

Keep in mind this is at extreme zoom in

tresf · 2017-02-27T17:12:56Z

@TheTravelingSpaceman @Umcaruje that shading won't ever (EVER) line up unless we either set a fixed gradient (yuck).

Can we just abolish it? I see this is being done against the old theme, has the new theme gotten rid of the gradient or is it still there?

WalterSmuts · 2017-02-27T17:24:27Z

Can we just abolish it?

Are you talking about the gradient?

The new theme works just fine since it doesn't use a gradient.

tresf · 2017-02-27T17:58:26Z

The new theme works just fine

👍

Umcaruje · 2017-02-27T20:04:05Z

The new theme works just fine since it doesn't use a gradient.

Actually it still does use one, but I agree this can be removed really. @RebeccaDeField thoughts?

Umcaruje · 2017-02-27T20:08:36Z

src/gui/editors/AutomationEditor.cpp

-					drawLevelTick( p, it.key() + i, values[i],
-									is_selected );
+				float nextValue;
+				if ( m_pattern->valuesAfter( ( it + 1 ).key() ) != NULL )


You should have brackets for the if/else statement, and there should be no space between if and the parenthesis: if(...

I thought it looked neater without the brackets but I'll add them to make it consistent?

You're still missing {} brackets here

Umcaruje · 2017-02-27T20:24:19Z

src/gui/editors/AutomationEditor.cpp

+				path.lineTo( QPointF( xCoordOfTick( ( it + 1 ).key() ), yCoordOfLevel( nextValue ) ) );
+				path.lineTo( QPointF( xCoordOfTick( ( it + 1 ).key() ), yCoordOfLevel( 0 ) ) );
+				path.lineTo( QPointF( xCoordOfTick( it.key() ), yCoordOfLevel( 0 ) ) );
+				p.fillPath( path, graphColor() );


The is_selected variable was used when a particular automation point was selected, but this is disabled and broken atm, it would requre some CSS changes to be properly implemented, and we can save that for when that feature gets fixed.

So can I delete the code referencing the is_selected variable and related code?

It's good when it's commented out as now.

RebeccaDeField · 2017-02-27T20:52:56Z

Actually it still does use one, but I agree this can be removed really. @RebeccaDeField thoughts?

The gradient is just for a touch of depth and it can definitely be flattened. I do think it's important to keep some transparency so that the grid can show through for usability. I'm thinking we could give the whole automation graph the same transparency as the top stop.

WalterSmuts · 2017-02-27T22:35:42Z

I'm thinking we could give the whole automation pattern the same transparency as the top stop.

What's the top stop?

And is it a problem that the Travis-CI failed it's tests?

tresf · 2017-02-28T02:56:41Z

What's the top stop?

I believe @RebeccaDeField is recommending to use the top-most automation point as the gradient end point for the entire section so that they draw without segmentation.

RebeccaDeField · 2017-02-28T03:38:58Z

@TheTravelingSpaceman It is currently using a gradient with two stops (top and bottom). If we wanted to take away the gradient, we would only have one shade of transparency. I'm suggesting that we use the transparency of stop:0 which seems to be 180.

Sorry if that was confusing 😅

WalterSmuts · 2017-02-28T09:08:02Z

@RebeccaDeField So do we change that in the CSS or the c++ code? I can't seem to find the right place to limit it in the c++ code :/

Umcaruje · 2017-02-28T09:23:52Z

@TheTravelingSpaceman I think it would be wiser to actually comment out the code rather than remove it, cause we want that feature back sometime.

WalterSmuts · 2017-02-28T14:30:46Z

So? What's the next step?

WalterSmuts · 2017-03-02T08:11:30Z

@Umcaruje @RebeccaDeField @tresf

RebeccaDeField · 2017-03-04T06:45:35Z

@TheTravelingSpaceman Sorry for the late reply, this somehow got buried in my inbox. You will need to remove the gradient in the style.css file if that is what you're asking. 😊

WalterSmuts · 2017-03-05T13:56:46Z

@Umcaruje Should be fine to merge now

WalterSmuts · 2017-03-12T11:05:35Z

@Umcaruje I've looked at the two problems you've indicated and can't find a proper way to fix it (That I understand and am confident that it won't break something elsewhere).

I did "fix" the bottom pixel error in the final commit but if feels a bit hackish. Maybe the person who wrote the code originally should take a look (@jmuzz)

The problems however has nothing to do with the things I've added and can therefore be fixed separately in another PR. I would like to fix it but unfortunately I'm starting school tomorrow.

zonkmachine · 2017-03-12T12:46:52Z

Did another test. It looks really smooth, but you're off by a pixel when drawing in the automation editor:
...
As you see, there's a pixel gap on the bottom.

But this gap is there already in 1.0.3

I did "fix" the bottom pixel error in the final commit but if feels a bit hackish.

Works, but maybe not if fixing the top margin error too.

Also there seems to be some overflowing on the top end of the automation editor, this was probably not introduced with this PR, but it's just something that I've noticed now.

https://github.com/LMMS/lmms/blob/master/include/AutomationEditor.h#L176
If we don't take the bottom gap into account, Changing TOP_MARGIN value to ~20 seem to take care of this. I think rather the theming folks needs to take a look at this as it's a separate issue.

Umcaruje · 2017-03-12T14:26:27Z

But this gap is there already in 1.0.3

Huh, I never noticed that. Guess I never looked hard enough. Anyways @TheTravelingSpaceman the last commit doesn't fix it completely. I suggest you revert it, we merge this and open a seperate issue for the top margin and the pixel gap.

This reverts commit ff80186.

WalterSmuts · 2017-03-12T17:41:09Z

Reverted as requested. Should be fine to merge now @Umcaruje

zonkmachine · 2017-03-12T17:49:45Z

Reverted as requested. Should be fine to merge now @Umcaruje

I'm pretty sure he meant reverting 940c766

WalterSmuts · 2017-03-12T17:55:28Z

@zonkmachine I thought that was what I was doing. Was under the impression that the command reverts to the state that the commit was at that point. So just making sure now... I need to execute the following code to have it right:
git revert 4e127a7
git revert 940c766

The first line reverts the incorrect "revert" that I did. And the second is what I should have done initially

This reverts commit 4e127a7.

This reverts commit 940c766.

zonkmachine · 2017-03-12T18:04:18Z

git revert 4e127a7
git revert 940c766

Yes, this works.

For the future. I usually do rebase -i and remove the lines of the commits I want removed as it's neater ( first back up! ) .

tresf · 2017-03-12T19:29:42Z

revert reverts a single commit. reset --hard rewinds back in time to a particular commit.

Umcaruje · 2017-03-13T00:13:52Z

👍

RebeccaDeField · 2017-03-13T02:17:11Z

Nice work @TheTravelingSpaceman! 😁

* Ailiasing in AutomationPatternView; ERROR: Doesn't draw unreferanced tracks * Draws one polygon instead of 'poly'-polygons * Changed QPoints to QPointF * Added int MidiTpT constant * Added ppTact for reduced computation * Added spaces in parentheses to be consistent * Variable name change and spacing * S P A C E S * Anti-Aliasing of Automation Editor * Commented out all referances to unused is_selected * Changed css to non-gradient graphs * Added Brackets * Removed no-pixel line at bottom of graph * Revert "Added Brackets" This reverts commit ff80186. * Revert "Revert "Added Brackets"" This reverts commit 4e127a7. * Revert "Removed no-pixel line at bottom of graph" This reverts commit 940c766.

Umcaruje requested changes Feb 26, 2017

View reviewed changes

zonkmachine reviewed Feb 26, 2017

View reviewed changes

Umcaruje reviewed Feb 27, 2017

View reviewed changes

WalterSmuts force-pushed the Anti-Aliasing branch from 1538466 to af9ed17 Compare February 27, 2017 11:33

Umcaruje approved these changes Feb 27, 2017

View reviewed changes

WalterSmuts force-pushed the Anti-Aliasing branch from 0031766 to 10de750 Compare February 28, 2017 10:27

WalterSmuts changed the title ~~Aliasing in AutomationPatternView; ERROR: Doesn't draw unreferencing tracks~~ Aliasing in AutomationPatternView and AutomationEditor fixed Feb 28, 2017

zonkmachine approved these changes Mar 6, 2017

View reviewed changes

WalterSmuts added 8 commits March 10, 2017 15:39

Added int MidiTpT constant

56cc028

Added ppTact for reduced computation

fb1e841

Added spaces in parentheses to be consistent

299380a

Variable name change and spacing

07596d6

S P A C E S

83edae4

Anti-Aliasing of Automation Editor

3897aa4

Commented out all referances to unused is_selected

9f689a3

Changed css to non-gradient graphs

7d0355e

WalterSmuts force-pushed the Anti-Aliasing branch from a57b823 to 7d0355e Compare March 10, 2017 13:51

WalterSmuts added 2 commits March 11, 2017 17:04

Added Brackets

ff80186

Removed no-pixel line at bottom of graph

940c766

Revert "Added Brackets"

4e127a7

This reverts commit ff80186.

WalterSmuts added 2 commits March 12, 2017 20:03

Revert "Revert "Added Brackets""

0da54d3

This reverts commit 4e127a7.

Revert "Removed no-pixel line at bottom of graph"

54098f5

This reverts commit 940c766.

Umcaruje approved these changes Mar 13, 2017

View reviewed changes

Umcaruje merged commit 899e386 into LMMS:master Mar 13, 2017

zonkmachine mentioned this pull request Mar 13, 2017

Anti-Aliasing in automation editior #2688

Closed

WalterSmuts deleted the Anti-Aliasing branch March 13, 2017 10:26

Umcaruje mentioned this pull request Mar 14, 2017

Update classic theme #3424

Merged

Aliasing in AutomationPatternView and AutomationEditor fixed #3386

Aliasing in AutomationPatternView and AutomationEditor fixed #3386

Conversation

WalterSmuts commented Feb 26, 2017

tresf commented Feb 26, 2017 • edited Loading

Umcaruje commented Feb 26, 2017

WalterSmuts commented Feb 26, 2017

Choose a reason for hiding this comment

WalterSmuts commented Feb 26, 2017

Choose a reason for hiding this comment

WalterSmuts commented Feb 26, 2017

Umcaruje commented Feb 26, 2017 • 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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

WalterSmuts commented Feb 27, 2017

tresf commented Feb 27, 2017

WalterSmuts commented Feb 27, 2017 • edited Loading

tresf commented Feb 27, 2017

Umcaruje commented Feb 27, 2017

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

RebeccaDeField commented Feb 27, 2017 • edited Loading

WalterSmuts commented Feb 27, 2017

tresf commented Feb 28, 2017

RebeccaDeField commented Feb 28, 2017 • edited Loading

WalterSmuts commented Feb 28, 2017

Umcaruje commented Feb 28, 2017

WalterSmuts commented Feb 28, 2017

WalterSmuts commented Mar 2, 2017

RebeccaDeField commented Mar 4, 2017

WalterSmuts commented Mar 5, 2017 • edited Loading

WalterSmuts commented Mar 12, 2017

zonkmachine commented Mar 12, 2017

Umcaruje commented Mar 12, 2017

WalterSmuts commented Mar 12, 2017 • edited Loading

zonkmachine commented Mar 12, 2017

WalterSmuts commented Mar 12, 2017 • edited Loading

zonkmachine commented Mar 12, 2017

tresf commented Mar 12, 2017

Umcaruje commented Mar 13, 2017

RebeccaDeField commented Mar 13, 2017

tresf commented Feb 26, 2017 •

edited

Loading

Umcaruje commented Feb 26, 2017 •

edited

Loading

WalterSmuts commented Feb 27, 2017 •

edited

Loading

RebeccaDeField commented Feb 27, 2017 •

edited

Loading

RebeccaDeField commented Feb 28, 2017 •

edited

Loading

WalterSmuts commented Mar 5, 2017 •

edited

Loading

WalterSmuts commented Mar 12, 2017 •

edited

Loading

WalterSmuts commented Mar 12, 2017 •

edited

Loading