transform-swf-updates — News on releases, bugs and patches for Transform SWF for Java

[transform-swf-updates] Status report for week ending June 4th

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

Breaking News: There will be an alpha release next week - there are only 
one or two random bugs left and all the tests are passing but I don't 
want to rush out a release on a Friday just in case I missed something. 
Since all the tests are passing it seems likely that a beta version will 
be released shortly - certainly before the end of June and hopefully in 
the next week or so.

Progress this week:

API Changes:

Simplified Movie by moving the version, compressed, frameSize and 
frameRate attributes into a new class MovieHeader. Movie is now purely a 
container class. It is also easier to write custom decoders which for 
example only decode the meta data on a movie.

The decodeFromStream() and encodeToStream methods on Movie are now 
private. Compressed Flash files contain a mix of uncompressed and 
compressed data therefore the stream passed to the method was not 
necessarily the stream used to handle the encoding / decoding. For 
writing the encoding stream has to be closed to ensure all data is 
flushed to the file. To do this the stream actually used had to be 
returned so the calling code could close the correct one. This was ugly 
so the Movie class was simplified to make it easier to write custom 
encoders and decoders when required.

Added the method decodeFromUrl(URL) to Movie to make it easier to read 
files from a network. The method is simple with no reries if there is an 
error. If retry functionality is needed then a custom decoder i.e. your 
own version of Movie must be written - however that is really easy to do 
by simply copying the existing Movie class and modifying the read(URL) 
method.

This is probably the last of the changes to public methods so the  API  
can be considered extremely stable, if not actually frozen.

Code Changes:

Simplified Context so the variables are accessible directly.

Moved classes that handle decoding of the different types of object to 
packages to minimise the number of import directives.

Coalesced all widely used constants into the SWF class.

Moved ImageTag from com.flagstone.transform.coder to 
com.flagstone.transform.image.
Moved MovieTag from com.flagstone.transform.coder to 
com.flagstone.transform.
Moved DefineTag from com.flagstone.transform.coder to 
com.flagstone.transform.

Refactored (cleaned up) coder classes to make them more independent and 
removed unused code.

Testing:

Resolved the issue with decoding DefineFont3. This problem is a bug in 
Flash where the containsLayout flag is not set if the font definition is 
followed by a FontAlignment tag even though the font definition does 
contain layout information.

Closed Tickets:

#30       Upgrade coder classes to support streams directly.
#49       Add Event which merges (and replaces) MovieClipEvent and 
ButtonEvent.
#121     Update Maven assembly to generate jars for sources and docs.
#126     Group reads and write of bit fields into byte accesses.
#127     Remove length attribute when passing a stream to ImageDecoder.
#128     Remove length attribute when passing a stream to SoundDecoder.
#130     Upgrade JPGDecoder to filter extension blocks.

Closed Conditionally:

#131     SoundDecoders should only load data on demand.

This is closed for MP3 files where frames are only loaded on demand but 
this cannot be imp'lemented for WAV files since the the FMT block which 
defines the sound may be decoded from a file after the DATA block which 
contains the sound samples. This means that a WAV file must be 
completely loaded into memory before it can be processed.

Open Tickets:

#122     Register and deploy to a central Maven repository.
#125     Add support for decoding files with hidden actionscript.


Regards,

Stuart
--
Flagstone Software Ltd.

[transform-swf-updates] Status report for week ending May 28th (actually 29th)

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

First: No alpha release yet.

Progress this week:

API Changes:

Removed the identifier attribute from Movie. While getting movie to keep 
track of the identifiers used for object definitions was somewhat 
useful, it complicated the code and slowed the decoding of Movies. Now 
you will have to maintain the counter in your code rather than relying 
on Movie to do it.

Separated out the Flash Video (FLV) related code. It will be added to a 
new project.

Code Changes:

Encoding and Decoding Movies is now based on streams. The SWFDecoder now 
supports an internal buffer where the data is loaded from the underlying 
stream before being decoded. Similarly the SWFEncoder uses the buffer to 
write to the underlying stream in blocks.

Removed as many bit fields reads and writes, instead grouping them into 
calls to readByte / writeByte.

These two changed combined seem to have contributed to a MAJOR increase 
in speed - the code is at least twice as fast. The increase is probably 
a result of the code changes but perhaps more importantly the reduction 
in memory use - it is certainly obvious when the garbage collector runs 
between tests. Previously tests on the real-world Flash files has to run 
with the JVM setting -Xmx256m. Now that is no longer necessary :)

Testing:

There are still a few issues when decoding DefineFont3 objects. Most 
decode without any problems but there are a few cases that cause 
unbderflow errors. The root cause is not known - there might be some 
functionality that is poorly described or not listed in Adobe's file 
format specification.

Planned for next week:

Adding support for streams took waaaaay more work then expected so next 
week's plan is the same plan as last week: More debugging and a decision 
on whether to have an alpha release of the code.

Overall Progress:

Transform now fully supports Flash 10.
Now passing tests to decode, encode and copy 300 real-world Flash 8 and 
Flash 9 files - except some DefineFont3 tags.

AWTDecoder is able to render text using Java AWT Fonts.
TTFDecoder is able to render text using TrueType Fonts.

BMPDecoder is passing all acceptance tests for Window BMP images.
JPGDecoder is passing all acceptance tests for JPEG images.
PNGDecoder is passing all acceptance tests for JPEG images.
BufferedImageDecoder is mostly passing acceptance tests for PNGimages.

WAVDecoder is passing tests for generating event and streaming sounds.
MP3Decoder is passing tests for generating event and streaming sounds.

Canvas is passing tests for drawing shapes.

Open Tickets:

#49 - Add Event which merges (and replaces) MovieClipEvent and ButtonEvent.
#125 - Add support for decoding files with hidden actionscript.
#121 - Update Maven assembly to generate jars for sources and docs.
#122 - Register and deploy to a central Maven repository.
#126 - Group reads and write of bit fields into byte accesses.

Closed - but Trac is not updated yet.

#30 - Upgrade coder classes to support streams directly. 
#43 - Sound decoders do not report incorrect sample rates.


General Tasks Still ToDo:

Assess test coverage.
Fix problems reported by PMD.
Complete suite of Unit Tests.
Remove all the magic numbers reported by CheckStyle
Add javadoc comments for all methods and fields.
Review the javadoc comments.

It has been a good week with great progress.

Stuart
--
Flagstone Software Ltd.

[transform-swf-updates] Status report for week ending May 21st

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

Progress this week:

API Changes:

1. Added the CharacterSet convenience class for creating sorted lists of 
character codes for generating font definitions.

2. Added a new method to the SoundDecoder interface to specify the 
number of frames when generating a streaming sound. This should be 
useful when adding soundtrack to Movies where the soundtrack is longer 
than the movie and so avoids having to read the entire sound before 
generating the necessary objects.

3. Moved the methods in Place2.Builder and Place3.Builder back to being 
static methods on the parent class since the Builder classes were not 
full Builders and using as static methods is easier.

4. Added convenience methods to Bounds to generate a new bounds 
increased by a padding amount.

Code Changes:

1. Added character encoding to Context to simplify SWFEncodable 
interface and simplify unit test classes.
2. Added TrueTypeGlyph to support decoding glyph information from 
TrueType fonts.

Testing:

1. Acceptance tests for the sound decoder classes for playing MP3 files 
as streaming sounds are passing.
2. Acceptance tests for generating fonts displaying text using 
AWTDecoder and TTFDecoder now pass.
3. Acceptance tests for displaying PNG images using  
BufferedImageDecoder  now pass but with a few issues with colour.
4. Added a test reports to quickly allow the results from the 
integration tests.
5. Reogranised tests and test resources to allow failsafe plugin to run 
integration tests and provide better integration with Maven.

Documentation:

1. Cleaned up most of the javadoc problems reported by CheckStyle - 
leaving only undocumented methods and fields to add.


Planned for next week:

More debugging and a decision on whether to have an alpha release of the 
code.


Overall Progress:

Transform now fully supports Flash 10.
Now passing tests to decode, encode and copy 300 real-world Flash 8 and 
Flash 9 files.

AWTDecoder is able to render text using Java AWT Fonts.
TTFDecoder is able to render text using TrueType Fonts.

BMPDecoder is passing all acceptance tests for Window BMP images.
JPGDecoder is passing all acceptance tests for JPEG images.
PNGDecoder is passing all acceptance tests for JPEG images.
BufferedImageDecoder is mostly passing acceptance tests for PNGimages.

WAVDecoder is passing tests for generating event and streaming sounds.
MP3Decoder is passing tests for generating event and streaming sounds.

Canvas is passing tests for drawing shapes.


Open Tickets:

#30 - Upgrade coder classes to support streams directly.
#43 - Sound decoders do not report incorrect sample rates.
#49 - Add Event which merges (and replaces) MovieClipEvent and ButtonEvent.
#125 - Add support for decoding files with hidden actionscript.
#121 - Update Maven assembly to generate jars for sources and docs.
#122 - Register and deploy to a central Maven repository.
#126 - Group reads and write of bit fields into byte accesses.

General Tasks Still ToDo:

Assess test coverage.
Fix problems reported by PMD.
Complete suite of Unit Tests.
Remove all the magic numbers reported by CheckStyle
Add javadoc comments for all methods and fields.
Review the javadoc comments.

[transform-swf-updates] Transform SWF For Java 2.3.3 Released

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

Transform SWF 2.3.3 for Java is now available for download. This release 
fixes two bugs when decoding AWT Fonts to generate font definitions 
using the FSTextConstructor.

Bug Fixes:

* Using characters with high value character codes causes errors.
* AWT Fonts on Mac OSX are offset.

Release notes: 
http://www.flagstonesoftware.com/downloads/transform-java-2.3.3.txt
Downloads: http://www.flagstonesoftware.com/downloads/index.html

Regards,

Stuart MacKay
Flagstone Software Ltd.

[transform-swf-updates] Status report for week ending May 14th

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

Another week with lots done:

API Changes:

    Actions are now immutable to speed up copying of movies. Nested 
Builder classes were created for ExceptionHandler, NewFunction[2] & Push.
    Moved the static methods on Place2 and Place3 into a nested Builder 
class.
    Renamed identifier() method on Movie to nextIdentifier() so its 
purpose is clearer.
    Deleted ArraySIzeException - only used a couple of times and having 
an extra class did not justify it.
    Deleted StringSIzeException - only used a couple of times and having 
an extra class did not justify it.

Code Changes

    Added envent codes to the enum MovieClipEvent to simplify the code 
in MovieClipEventHandler.

Code Cleanup
   
    Removed all the errors and most of the warnings reported by PMD. 
Only complexity warnings remain which will be fixed in due course.
    Removed the rect, circle and ellipse methods from canvas to reduce 
the class size. The code will be made available in the cookbook.
    Converted a large number of magic numbers into constants to make the 
code easier to read. A lot remain which will be fixed in due course.

Testing

    Acceptance tests for the image decoder classes for displaying BMP, 
JPG and PNG are passing (with 1 or two exceptions).
    Acceptance tests for the sound decoder classes for playing WAV and 
MP3 files as event sounds are passing.
    Acceptance tests for the sound decoder classes for playing WAV files 
as streaming sounds are passing (MP3 files are failing).


The plan for next week is to concentrate on getting the remaining 
acceptance tests passing, then review the implementation plan and take a 
decision on releasing the code as an alpha or or possibly beta.

Regards,

Stuart
Flagstone Software Ltd.

[transform-swf-updates] Now with Flash 10

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

Well that didn't take long. Admittedly there were only two new classes 
(DefineFont4 and DefineJPEGImage4) and a new sound format (Speex) to 
deal with. It will be interested to see if there is anything missing 
from the file format specification published by Adobe - there was quite 
a bit missing from the Version 9 specification.

Regards,

Stuart MacKay
Flagstone Ltd.

[transform-swf-updates] Status report for week ending May 7th

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

After a pause of a couple of weeks I made excellent progress over the 
past five days. Version 3.0 now passes encoding, decoding and copying 
tests for 299 out of 300 real-world Flash 8 and Flash 9 files. This is 
the same type of tests that I used for acceptance tests of the current 
2.3 version so at this point I am tempted to declare (a tactical)  
victory and release at least an alpha version - though this is probably 
a little premature since the utilities classes for adding imags etc are 
still untested.

The next stage will be to add support for Flash 10 since the changes 
published in the specification are relatively minor. From there the 
utilities classes will be tested and a beta declared. After that the 
polishing of documentation, unit tests and other "housekeeping" tasks 
will proceed.

Regards,

Stuart MacKay
Flagstone Software Ltd.

[transform-swf-updates] BUG: Version 2.3.2 - Glyphs are offset when using Mac OS X

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

When an AWT Font object is used to generate a font definition on Mac OS 
X the characters are offset vertically. The problem appears to be the 
way the TrueType 
<https://sourceforge.net/apps/trac/transform-swf/wiki/TrueType> fonts 
are decoded by Java. In Sun's JDK 1.4 no translation was required as 
font glyphs were decoded but that changed in JDK 1.6 so the translation 
was applied.

First assessment is that Apple's version of Java 1.6 operates the same 
as 1.4 so the translation in the current version is wrongly applied.

It is hoped to have a fix ready for a scheduled release at the end of April.

--
Stuart MacKay
Flagstone Software Ltd.

[transform-swf-updates] BUG: Version 2.3.2 - Using high value characters causes errors in FSTextConstructor

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

In Transform SWF 2.3.2, when displaying characters with a character code 
of 16384 or more then an exception is thrown:

java.lang.ArrayIndexOutOfBoundsException: -3999
at com.flagstone.transform.util.FSTextConstructor.advancesForGlyphIndice 
(Unknown Source)
at com.flagstone.transform.util.FSTextConstructor.defineText(Unknown
Source).

The problem is the use of short[] for the tables used to index the 
glyphs. When a character code of 16384 or higher is stored in a table 
and read back it is not bit-masked so the value gets sign-extended. The 
result is that a negative value is used to index into the table 
containing the advances for each character.

A patch has been uploaded for the ticket: 
https://sourceforge.net/apps/trac/transform-swf/ticket/123

A new release will be posted at the end of April once one other issue 
regarding the display of fonts on Mac OS X systems, see Ticket 124 has 
been resolved.

--
Stuart MacKay
Flagstone Software Ltd.

[transform-swf-updates] Status Report for week ending 9th April 2010

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

Last week was cut short by the Easter weekend so there was not much to 
report. This week, though I did not have much time has gone well. Out of 
a small set of Flash 8/9 files version 3 can now decode 16 out of 20. 
Interestingly the files which have problems in decoding do not appear to 
follow the Flash File format specification. At this point is not 
possible to tell whether the app that generated them was buggy or 
perhaps there is some fields that are not documented in the specification.

Next step is to spend a little more time investigating the files that 
are failing, then if the root cause cannot be found for now, add them to 
the to-do list and start expanding the suite of test files to see if the 
errors are consistent. From there start testing the encoding process.

Regards,

Stuart MacKay
Flagstone Software Ltd.

[transform-swf-updates] Progress report for week ending 26th March 2010

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

Started running the acceptance tests with Flash 8 and Flash 9 files. 
Since the code is running well with files up to Flash 7 this will 
hopefully accelerate the development process allowing at least a fully 
functioning beta release while the work on the unit tests and other 
project related tasks is completed.

Regards,

Stuart MacKay
Flagstone Software Ltd.

[transform-swf-updates] Status report for week ending 21 March 2010

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

Completed the work started last week on running the acceptance tests for 
encoding and decoding movies. The complete suite of real-world Flash 
files - up to Flash 7 - can now be successfully decoded and encoded then 
re-decoded. This means that the version 3.0 code can now be used to do 
useful work - though problably the best approach is to use it for 
porting existing applications.

Regards,

Stuart
--
Stuart MacKay, Flagstone Software Ltd.

[transform-swf-updates] Progress report for week ending 12th March 2010

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

After a long pause work started in earnest to get version 3.0 out of the 
door. Here's a summary of what was completed:

1. Refactoring of packages moved all the video related classes into 
com.flagstone.transform.video. Previously the classes were spread across 
several packages since video can be encoded in SWF as well as FLV files. 
Moving everything into com.flagstone.transform.video helps reduce the 
size of other packages, particularly com.flagstone.transform and makes 
it easier to find all video related code.

2. Added a new class, PropertyValue which is used to handle undocumented 
or incorrectly encoded properties that prevent an enum Property being 
used. This appears to be a problem in Flash 4 files only.

3. Added new milestones and tasks to the project portal at 
http://sourceforge.net/apps/trac/transform-swf/roadmap Not all of the 
completed tasks are marked but over the next few weeks you should be 
able to get a reasonable view of how the project is progressing.

4. Ran the acceptance tests for Movies. 351 out of 359 files from a set 
of open source files downloaded from FlashKit can be now decoded and 
encoded successfully.  There are some issues that cause the decoding to 
run out of control and there are problems reading/writing bit fields 
that extend across 32-bit word boundaries but generally this means that 
the current snapshot from the repository can be used to read/write files 
up to Flash 7. Files for Flash 8 and 9 will be added but not until more 
unit tests are written and passing.

Overall I am pretty pleased with progress and hope to build more 
momentum in the project over the next few days. The next status update 
will be at the end of next week.

Regards,

Stuart MacKay
Flagstone Software Ltd.

[transform-swf-updates] Development work restarted for Version 3

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

After a long pause due to work and family commitments development work 
on Version 3.0 of Transform SWF, which will include support for Flash 8 
& 9 has now been restarted. Regular progress reports (probably weekly) 
will be posted to this list with descriptions of any changes 
(refactoring) to the current codebase.

The first task is to continue updating the unit tests. The test patterns 
for all the different variations in the way each object is encoded and 
decoded is not stored in YAML files. This makes it easier to add new 
patterns without having add new tests. One sticking point previously was 
the converting the binary data into Base64 so it can be stored in the 
YAML file however Notepad++ and an improved version of the HexEditor 
plugin now makes the conversion process a lot easier and faster.

Regards,

Stuart
-- 
Flagstone Software Ltd.

[transform-swf-updates] Transform SWF 2.3.2 for Java now available

[News on r. b. a. p. f. T. S. f. J. <tra...@li...>]

Transform SWF 2.3.2 for Java is now available for download. This release 
fixes bugs when encoding movies with not UTF-8 character sets and 
improves the handling of sound files.

Bug Fixes:
  41. FSMovie does not set the character encoding in FSCoder.
  42. FSSoundConstructor accepts sounds with any sample rate.
  48. Extended arrays in FSShapeStyle are encoded as bit fields

The release notes, 
http://www.flagstonesoftware.com/downloads/transform-java-2.3.2.txt has 
all the details.

Regards,

Stuart
-- 
Stuart MacKay
Flagstone Software Ltd.