never mind what I just said.  Your insistence on IndexedColorModel -- it's
finally sinking in ... and getting better results!

Thanks!

On Thu, Apr 30, 2009 at 7:36 PM, Daniel Wilson <
[email protected]> wrote:

> >>This gets confusing. Maybe we need to take a concrete PDF as example and
> discuss it.
>
> Quite!
>
> I'm looking at PDFBox\trunk\test\input\rendering\ArchiveRGB.ai.
>
> The PNG that is posted with it lacks the stencil mask.  The stencil mask,
> as I've implemented it, on this one puts a white screen over most of it and
> some gibberish dots near the top.
>
> Obviously, the stencil mask handling is wrong in this case (though it's
> much closer in most others).  This is Separation with an alternate colors
> space that is ICCBased.
>
> The raw PDF source that seems to have to do with this is:
> 31 0 obj
> [/Separation/J80-NAVY 24 0 R<</C0[0.0 0.0 0.0 0.0]/C1[0.992157 0.835297
> 0.44313 0.46666]/FunctionType 2/N 1.0/Domain[0 1]/Range[0.0 1.0 0.0 1.0 0.0
> 1.0 0.0 1.0]>>]
> endobj
>
> Though this one also has something to do with that ... this actually has
> the ImageMask flag set:
> 26 0 obj
> <</Intent/RelativeColorimetric/Subtype/Image/Length
> 96956/Filter/FlateDecode/Name/X/ImageMask true/BitsPerComponent 1/Width
> 1188/DecodeParms<</Columns 1188/Predictor 15/BitsPerComponent 1/Colors
> 1>>/Height 1188/Type/XObject/Decode[0.0 1.0]>>stream
>
> and again:
> 28 0 obj
> <</Intent/RelativeColorimetric/Subtype/Image/Length
> 57146/Filter/FlateDecode/Name/X/ImageMask true/BitsPerComponent 1/Width
> 864/DecodeParms<</Columns 864/Predictor 15/BitsPerComponent 1/Colors
> 1>>/Height 864/Type/XObject/Decode[0.0 1.0]>>stream
>
> Why do I think it's IICBased?  Well, that's the branch of logic it's taking
> in PDFBox.  I can't tell at this point whether that's right or wrong.
>
> Thanks again for any guidance here!
>
> Daniel
>
>
>
> On Thu, Apr 30, 2009 at 2:20 AM, Jeremias Maerki 
> <[email protected]>wrote:
>
>> On 30.04.2009 01:46:31 Daniel Wilson wrote:
>> > OK, I'm changing over to 4 spaces.  You're right, SCite has that option.
>> >
>> > I'm trying to run ant checkstyle, but am getting a
>> > UnsupportedClassVersionError.  I think this is b/c I'm running jdk 1.4
>> --
>> > which I'm doing to ensure I do not accidentally introduce something
>> > supported only in newer versions of Java.
>> >
>> > Do you have suggestions on this?
>>
>> I've configured Checkstyle in Eclipse which is running Java 6 but PDFBox
>> is configured to compile with Java 1.4, so I don't have that problem.
>>
>> I think you can also just run Checkstyle from Ant with Java >1.4 but
>> compile with Java 1.4. I've set up some scripts that let me switch JDKs
>> easily by changing the JAVA_HOME env variable.
>>
>> > >> I read this as only relevant if you are using DataBuffer.TYPE_BYTE
>> (for
>> > example) and per component less than 8 bits are used.
>> >
>> > Which is exactly what we're dealing with.
>>
>> Not in the case I've cited: PDICCBased. It uses ComponentColorModel, not
>> IndexColorModel. We have to keep "components" apart from "samples". When
>> an XObject uses an /ICCBased color space it usually (!) has 8 bits per
>> components and n components. But an XObject could also use a /Indexed
>> color space with an /ICCBased color space as base in which case PDFBox
>> would have to build an IndexColorModel-like instance that is based on an
>> arbitrary ICC_ColorSpace (rather than sRGB).
>>
>> This gets confusing. Maybe we need to take a concrete PDF as example and
>> discuss it.
>>
>> > Is there a better way to initialize the array?  Providing it with the
>> number
>> > of elements it needs?
>> >
>> > Thanks.
>> >
>> > Daniel Wilson
>> <snip/>
>>
>>
>> Jeremias Maerki
>>
>>
>

Reply via email to