Gradient Brushes #542

jongleur1983 · 2018-04-19T18:24:44Z

Prerequisites

I have written a descriptive pull-request title
I have verified that there are no overlapping pull-requests open
I have verified that I am following matches the existing coding patterns and practice as demonstrated in the repository. These follow strict Stylecop rules 👮.
I have provided test coverage for my change (where applicable)

Description

Implementing Gradient Brushes:

Linear Gradient
Radial Gradient
Elliptical Gradient (supports size, height/length ratios and rotation of the ellipse)

Color Repetition Modes

DontFill: fill between start and end of the gradient, don't touch anything else.
None: fill between start and end of the gradient. Beyond those borders, stay with the same color
Repeat: repeat the gradient endlessly
Reflect: same as repeat, but each other repetition is inverted (=reflected).

What's missing

proper API documentation

What I'd like to add in future PRs:

gradienting modes (e.g. HSV-gradienting. According to @JimBobSquarePants he might implement the corresponding formulas/algorithms somewhere else, so it might just have to be utilized here.
performance improvements at least for the common cases
- Linear Gradients: vertical lin gradients: each row is unicolored, so we have to calculate the color only once per row. Horizontal gradients: not sure how best to do it, but here each row is the same bunch of pixels, so we can cache it between the Apply calls.
- Radial Gradients: from the center point towards the edges, each row is mirrored, so only half of the colors have to be calculated.
Use different BrushApplicators to speed up the special cases (e.g. Elliptic Brush with AxisRatio 1 is a RadialBrush)
Another gradient Brush: StarShapedGradient: Consider a Polygon and a point inside, where the direct connection between that point and any position along the outline does not intersect the outline before. The class of these polygons contains all convex polygons, simple stars and more. Given a Star-Shaped polygon and one point inside where the condition described above holds, it's possible to color each point in an unambiguous way referring to a gradient description. ColorRepetitionMode.Reflect and ColorRepetitionMode.Repeat would not be supported here as that get's ambiguous outside the polygon.

CLAassistant · 2018-04-19T18:24:50Z

All committers have signed the CLA.

…ush, not yet working - what's wrong?

antonfirsov · 2018-04-19T21:25:25Z

src/ImageSharp.Drawing/Processing/Drawing/Brushes/LinearGradientBrush.cs

+
+            private readonly Point end;
+
+            private readonly Tuple<float, TPixel>[] colorStops;


I suggest to define a small private struct instead of Tuple<float, TPixel>. It provides:

Better readability

Better memory locality (Tuple<>-s are classes on the heap)

good point. Thanks. Except that the struct can't be private as it's used as a parameter in the Brush constructor. Therefore I'll use a public nested struct instead.

tocsoft

the reason you are getting a black screen is because of a bug in the base Apply method.

the test if (this.Options.BlendPercentage < 1) is incorrect and instead should read if (this.Options.BlendPercentage <= 1) i.e. less than or equal to instead of less than

tocsoft · 2018-04-19T21:23:50Z

tests/ImageSharp.Tests/Drawing/FillLinearGradientBrushTests.cs

+                    Assert.Equal(Rgba32.Red, sourcePixels[0, 0]);
+                    Assert.Equal(Rgba32.Red, sourcePixels[9, 9]);
+                    Assert.Equal(Rgba32.Red, sourcePixels[199, 149]);
+                    Assert.Equal(Rgba32.Red, sourcePixels[500, 500]);


this need to be 499 not 500 as its one pixel wider than the buffer.

thanks, you're right. fixed.

tocsoft · 2018-04-19T21:29:11Z

I've renamed the PR, we will happily keep reviewing it as you make improvements, once your fully happy and want a final review/merge just update the title again and drop the [WIP] prefix

@tocsoft

thanks @tocsoft for investigation and finding the bug. He proposed to just replace < by <= in line 78, but that would only shift the problem to values > 1. Those values should not be used from a semantic point, but it's not forbidden in a float value. My fix here keeps the <, but adds an else path that uses the original value from scanline[i]. This adds an else to the code (which technically adds a jump after the then-part), but omits the multiplication. The simple solution @tocsoft proposed might be faster, but should be preferred if and only if BlendPercentage can be guaranteed to be <=1.

… one smaller

@antonfirsov

as proposed by @antonfirsov: improving readability and memory locality.

codecov · 2018-04-19T23:32:01Z

Codecov Report

Merging #542 into master will decrease coverage by 0.4%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #542      +/-   ##
==========================================
- Coverage   87.67%   87.26%   -0.41%     
==========================================
  Files         838      840       +2     
  Lines       34684    35107     +423     
  Branches     2533     2553      +20     
==========================================
+ Hits        30410    30637     +227     
- Misses       3524     3719     +195     
- Partials      750      751       +1

Impacted Files	Coverage Δ
.../Processing/Drawing/Brushes/LinearGradientBrush.cs	`100% <100%> (ø)`
...harp.Tests/Drawing/FillLinearGradientBrushTests.cs	`100% <100%> (ø)`
...wing/Processing/Drawing/Brushes/BrushApplicator.cs	`86.95% <100%> (+48.86%)`	⬆️
...mats/Generated/Rgba32.PixelOperations.Generated.cs	`75% <0%> (-25%)`	⬇️
...ats/Generated/PixelOperations{TPixel}.Generated.cs	`80% <0%> (-20%)`	⬇️
src/ImageSharp/PixelFormats/Bgr24.cs	`78.26% <0%> (-19.04%)`	⬇️
src/ImageSharp/PixelFormats/Rgb24.cs	`75.55% <0%> (-18.89%)`	⬇️
src/ImageSharp/PixelFormats/Alpha8.cs	`82.5% <0%> (-17.5%)`	⬇️
src/ImageSharp/PixelFormats/NormalizedByte2.cs	`83.95% <0%> (-16.05%)`	⬇️
src/ImageSharp/PixelFormats/NormalizedShort2.cs	`82.71% <0%> (-15.82%)`	⬇️
... and 23 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 07a811c...d29b601. Read the comment docs.

test images don't have to be that big for axial gradients. It's sufficient to show they're constant across the axis and correct along the axis.

the theory lacks color checks yet that should be added, but it produces output images for visual control

somehow these don't look correct yet, containing hard edges sometimes.

jongleur1983 · 2018-05-13T12:58:32Z

@antonfirsov thanks for the changes. I still have 71 failing tests locally. On the first sight those are the same we thought to fail due the Vector2 issue, and: the same tests fail on SixLabors/master, too here.

So I'm fine with merging the branch.

antonfirsov · 2018-05-13T15:11:14Z

Everything looks good now @jongleur1983's local test running issues are unrelated to his PR.

I'm planning to merge this soon, let me know if there are blocker issues we've missed, so I shouldn't!

antonfirsov · 2018-05-13T15:17:41Z

...ageSharp.Drawing/Processing/Drawing/Brushes/GradientBrushes/AbstractGradientBrush{TPixel}.cs

+    /// Base class for Gradient brushes
+    /// </summary>
+    /// <typeparam name="TPixel">The pixel format</typeparam>
+    public abstract class AbstractGradientBrush<TPixel> : IBrush<TPixel>


Ok, one naming question mostly to @JimBobSquarePants :
Shouldn't we name this GradientBrushBase or simply GradientBrush to match our current naming patterns?

Yeah this should be GradientBrushBase.

done (and renamed the BrushApplicator as well)

antonfirsov · 2018-05-14T21:34:55Z

Finally merged!
@jongleur1983 thanks for pushing the rename!

… one smaller

@antonfirsov

as proposed by @antonfirsov: improving readability and memory locality.

test images don't have to be that big for axial gradients. It's sufficient to show they're constant across the axis and correct along the axis.

the theory lacks color checks yet that should be added, but it produces output images for visual control

somehow these don't look correct yet, containing hard edges sometimes.

The current implementation lacks accurrancy and prefers to be fast. For usual color gradients that shouldn't matter: They're quite smooth, for the black-white gradient on very small length's it's easy to spot these rounding errors as they affect whole pixels one can count manually.

…ude generic)

- move to GradientBrushes namespace - add abstract base class for Gradient Brushes, that implements the GetGradientSegment implementtion and the color picking.

@antonfirsov

…rawing tests, as @antonfirsov suggested.

@antonfirsov

as requested by @antonfirsov.

changing parameters so that visual evaluation of correctness is easier.

…Brush test files

Gradient Brushes

tocsoft self-requested a review April 19, 2018 18:25

tocsoft self-assigned this Apr 19, 2018

jongleur1983 force-pushed the GradientBrush branch 6 times, most recently from d284b76 to bc6e35c Compare April 19, 2018 19:40

SixLabors#86: started to work on Gradient Brushes: Linear gradient br…

c02f9d8

…ush, not yet working - what's wrong?

jongleur1983 force-pushed the GradientBrush branch from bc6e35c to c02f9d8 Compare April 19, 2018 19:46

jongleur1983 changed the title ~~#86: started to work on Gradient Brushes: Linear gradient brush,~~ #86: Linear gradient brush Apr 19, 2018

fix some typos in documentation.

4975eae

antonfirsov reviewed Apr 19, 2018

View reviewed changes

tocsoft reviewed Apr 19, 2018

View reviewed changes

tocsoft changed the title ~~#86: Linear gradient brush~~ [WIP] Linear gradient brush Apr 19, 2018

tocsoft added area:drawing enhancement labels Apr 19, 2018

jongleur1983 added 3 commits April 20, 2018 00:00

SixLabors#542: fix test: indices are 0-based, so bottom left pixel is…

6d87844

… one smaller

SixLabors#542: use struct for ColorStops

1e587bc

as proposed by @antonfirsov: improving readability and memory locality.

jongleur1983 added 7 commits April 20, 2018 19:18

fix code styling issues

6dd544d

SixLabors#542: reduce test output image sizes

ce72683

test images don't have to be that big for axial gradients. It's sufficient to show they're constant across the axis and correct along the axis.

SixLabors#542: add test for vertical gradient

90293cc

SixLabors#542: fix implementation of non-axial gradients and add tests

94dbb74

the theory lacks color checks yet that should be added, but it produces output images for visual control

SixLabors#542: add tests for multi-color gradients

7a3ba8a

somehow these don't look correct yet, containing hard edges sometimes.

cleanup whitespace

28457c8

SixLabors#542: add test creating black-white patterns by gradients

35d0f56

Merge remote-tracking branch 'origin/master' into GradientBrush

195a453

antonfirsov reviewed May 13, 2018

View reviewed changes

SixLabors#542: apply naming scheme for abstract classes

5d3daaa

antonfirsov merged commit dbe2b1b into SixLabors:master May 14, 2018

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: fix test: indices are 0-based, so bottom left pixel is…

1912311

… one smaller

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: use struct for ColorStops

ddf66d8

as proposed by @antonfirsov: improving readability and memory locality.

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: add test for vertical gradient

d40e49d

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: fix implementation of non-axial gradients and add tests

1efbd05

the theory lacks color checks yet that should be added, but it produces output images for visual control

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: add tests for multi-color gradients

70fd87f

somehow these don't look correct yet, containing hard edges sometimes.

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: add test creating black-white patterns by gradients

deb6cde

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: define debuggerDisplay of ColorStop,

497b052

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: fix rounding issues in tests

bd3faf7

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: rename file to match other files in the solution (incl…

6dda3f1

…ude generic)

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: apply change requests made by @tocsoft in code review.

29a37d0

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: refactor tests to follow the recommended pattern for d…

18cc05f

…rawing tests, as @antonfirsov suggested.

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: reduce test image sizes to save submudule size

c2561f1

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: remove pixel checks and base class

397352f

as requested by @antonfirsov.

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: improve tests for elliptic gradients

a5e2713

changing parameters so that visual evaluation of correctness is easier.

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: remove redundant Asserts, cleanup code

6a5c4fa

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: shorten test names, improve test image filenames

458aa53

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: code cleanup

16608b4

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: cleanup test file names, fix naming for RadialGradient…

9308b3a

…Brush test files

antonfirsov pushed a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

SixLabors#542: apply naming scheme for abstract classes

0b8f34e

antonfirsov added a commit to antonfirsov/ImageSharp that referenced this pull request Nov 11, 2019

Merge pull request SixLabors#542 from jongleur1983/GradientBrush

dab4a36

Gradient Brushes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Gradient Brushes #542

Gradient Brushes #542

jongleur1983 commented Apr 19, 2018 •

edited

Loading

CLAassistant commented Apr 19, 2018 •

edited

Loading

antonfirsov Apr 19, 2018 •

edited

Loading

jongleur1983 Apr 19, 2018

tocsoft left a comment

tocsoft Apr 19, 2018

jongleur1983 Apr 19, 2018

tocsoft commented Apr 19, 2018

codecov bot commented Apr 19, 2018 •

edited

Loading

jongleur1983 commented May 13, 2018

antonfirsov commented May 13, 2018 •

edited

Loading

antonfirsov May 13, 2018

JimBobSquarePants May 14, 2018

jongleur1983 May 14, 2018

antonfirsov commented May 14, 2018


		private readonly Point end;

		private readonly Tuple<float, TPixel>[] colorStops;

Gradient Brushes #542

Gradient Brushes #542

Conversation

jongleur1983 commented Apr 19, 2018 • edited Loading

Prerequisites

Description

What's missing

What I'd like to add in future PRs:

CLAassistant commented Apr 19, 2018 • edited Loading

antonfirsov Apr 19, 2018 • edited Loading

Choose a reason for hiding this comment

jongleur1983 Apr 19, 2018

Choose a reason for hiding this comment

tocsoft left a comment

Choose a reason for hiding this comment

tocsoft Apr 19, 2018

Choose a reason for hiding this comment

jongleur1983 Apr 19, 2018

Choose a reason for hiding this comment

tocsoft commented Apr 19, 2018

codecov bot commented Apr 19, 2018 • edited Loading

Codecov Report

jongleur1983 commented May 13, 2018

antonfirsov commented May 13, 2018 • edited Loading

antonfirsov May 13, 2018

Choose a reason for hiding this comment

JimBobSquarePants May 14, 2018

Choose a reason for hiding this comment

jongleur1983 May 14, 2018

Choose a reason for hiding this comment

antonfirsov commented May 14, 2018

jongleur1983 commented Apr 19, 2018 •

edited

Loading

CLAassistant commented Apr 19, 2018 •

edited

Loading

antonfirsov Apr 19, 2018 •

edited

Loading

codecov bot commented Apr 19, 2018 •

edited

Loading

antonfirsov commented May 13, 2018 •

edited

Loading