search

GoogleCodeArchive / piccolo2d

Automatically exported from code.google.com/p/piccolo2d
0 stars 0 forks source link

Text rendering in PText and JLabel via PSwing appear not to respect antialiasing rendering hints #117

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
Consider the attached screenshots which demonstrate JLabel, PText, and
JLabel via PSwing text rendering with different text antialiasing rendering
hints provided.  See
http://java.sun.com/javase/6/docs/api/java/awt/RenderingHints.html#KEY_TEXT_ANTI
ALIASING,
in particular those values available in jdk 1.6+.

Neither PText nor JLabel via PSwing renders text the same as JLabel.  PText
always antialiases, poorly at that, and JLabel via PSwing does not appear
to respect any of the LCD_* values.

An interesting observation is that when JLabel renders directly to a
BufferedImage instead of to the screen, it renders text similar to that of
JLabel via PSwing.  See attached tarball of a web page.

I still need to add JTextComponent, PStyledText, PSWTText, direct Java2D
text rendering and the new PHtml/PHtmlView node to the benchmark.

Original issue reported on code.google.com by heue...@gmail.com on 31 Jul 2009 at 3:26

Attachments:

GoogleCodeExporter commented 9 years ago 
I added JTextComponent and direct Java2D text rendering to the benchmark and 
things
got even stranger on WinXP.

With System Properties -> Advanced -> Performance -> Settings -> [x] Smooth 
edges of
screen fonts turned on, JLabel, JTextComponent, and JLabel via PSwing do not
antialias for any rendering hint values, whereas PText antialiases poorly for 
all
hint values, and direct Java2D renders all the hint values appropriately.

Turn off that setting and JLabel, JTextComponent, and Java2D all render 
correctly,
PText still antialiases poorly on all, and JLabel via PSwing renders OFF and 
GASP
hints correctly with no antialiasing and renders the rest antialiased poorly.

There's now too much to show on one screen shot, so I have attached the 
benchmark
code as a runnable jar.  JDK 1.6+ only.  Don't look at this stuff too long, for 
all
the fuzzy text will make your eyes hurt.

Original comment by heue...@gmail.com on 1 Aug 2009 at 6:15

Attachments:

GoogleCodeExporter commented 9 years ago 
Attaching new version of text-rendering-benchmark.jar with PHtmlView.

PHtmlView appears to have the same behaviour as PSwing described above.

Original comment by heue...@gmail.com on 6 Aug 2009 at 6:38

Attachments:

GoogleCodeExporter commented 9 years ago 
Results on mac os x

java version "1.6.0_13"
Java(TM) SE Runtime Environment (build 1.6.0_13-b03-211)
Java HotSpot(TM) 64-Bit Server VM (build 11.3-b02-83, mixed mode)

Original comment by she...@gmail.com on 12 Aug 2009 at 4:11

Attachments:

GoogleCodeExporter commented 9 years ago 
Source code is in svn at

http://dishevelled.svn.sourceforge.net/svnroot/dishevelled/trunk/text-rendering-
benchmark/

http://dishevelled.svn.sourceforge.net/viewvc/dishevelled/trunk/text-rendering-b
enchmark/

Original comment by heue...@gmail.com on 9 Apr 2010 at 9:39

GoogleCodeExporter commented 9 years ago 
This is my most frustrating Piccolo2D issue.  Can anyone else take a look at it?

Original comment by heue...@gmail.com on 24 Aug 2010 at 3:04

GoogleCodeExporter commented 9 years ago 
Attaching current version of text-rendering-benchmark runnable jar, built with 
Piccolo2D svn trunk.

Original comment by heue...@gmail.com on 2 Mar 2011 at 5:32

Attachments:

GoogleCodeExporter commented 9 years ago 
Further analysis:

Swing text rendering is dependent on operating system and installed Look and 
Feel.  See jlabel-*, pswing-*, jtextcomponent-*, and phtmlview-* in the text 
rendering benchmark.

Java2D text rendering using Graphics2D.drawSwing respects LCD_* text rendering 
hints of the current graphics context on all platforms.  See java2d-* in the 
text rendering benchmark.

PText text rendering using java.awt.font.TextLayout.draw() ignores any text 
rendering hints specified in the current graphics context.  See ptext-* in the 
text rendering benchmark.

TextLayout receives its text rendering hints from 
PPaintContext.RENDER_QUALITY_HIGH_FRC passed to a 
java.awt.font.LineBreakMeasurer ctr.

http://code.google.com/p/piccolo2d/source/browse/piccolo2d.java/trunk/core/src/m
ain/java/org/piccolo2d/nodes/PText.java#436

RENDER_QUALITY_HIGH_FRC is currently implemented as new FontRenderContext(null, 
true, true)

http://code.google.com/p/piccolo2d/source/browse/piccolo2d.java/trunk/core/src/m
ain/java/org/piccolo2d/util/PPaintContext.java#63

http://download.java.net/jdk7/docs/api/java/awt/font/FontRenderContext.html#Font
RenderContext(java.awt.geom.AffineTransform,%20boolean,%20boolean)

It is not possible to modify the text rendering hints supplied to TextLayout or 
LineBreakMeasurer after they have been constructed.

The javadoc for FontRenderContext reads

Typically, instances of FontRenderContext are obtained from a Graphics2D 
object. A FontRenderContext which is directly constructed will most likely not 
represent any actual graphics device, and may lead to unexpected or incorrect 
results.

Incorrect results such as the poorly anti-aliased text shown in the text 
rendering benchmark.

Looking forward, I see a few options

1) Attempt to instantiate a "better" FontRenderContext for 
PPaintContext.RENDER_QUALITY_HIGH_FRC.

2) Re-implement PText to create a new instance of TextLayout with the current 
text rendering hints from the graphics context every time PText.paintText is 
called.

3) Re-implement PText to use the instance of TextLayout created with 
PPaintContext.RENDER_QUALITY_HIGH_FRC only for layout purposes and render the 
text in PText.paintText with Graphics2D.drawString.

4) Re-implement PText with the same pattern as PHtmlView, delegating text 
rendering to a JLabel.

There are pros and cons to each, thought I should post this analysis and await 
comment before proceeding.

Original comment by heue...@gmail.com on 7 Mar 2011 at 10:46

GoogleCodeExporter commented 9 years ago

Original comment by heue...@gmail.com on 26 Nov 2013 at 11:22