Thank you very much Phil, for this informative message.
I'm developing an application for the public at large, I don't have any
control over their environment. But yes, I'm testing on an HP Postscript
printer, and I am using 1.2. And the component in question is a Swing
component with various layers piled into a JLayeredPane. I haven't had the
time, but I will certainly follow your guidelines.
-yvon
>From: Phil Race <[EMAIL PROTECTED]>
>Reply-To: Phil Race <[EMAIL PROTECTED]>
>To: [EMAIL PROTECTED], [EMAIL PROTECTED]
>Subject: Re: [JAVA2D] Printing performance and spool size
>Date: Wed, 13 Dec 2000 17:36:51 -0800 (PST)
>
>Hi,
> V. large spool files is a problem we've been tackling by reducing
> the cases where the circumsances that cause them are necessary.
> They are encountered only (so far as I know) when using a
>Postscript or
> perhaps PCL printer. Raster printers don't have this problem.
>
> Here's about every gory detail I can think of on this subject,so
> you can make informed decisions.
> Without more specific information its difficult to give
> specific suggestions.
>
> 1) What O/S are you using? NT, or win2k, or win 9x, or Solaris ?
>
> I have noticed that the 9x drivers generates much larger files
> than NT for the same GDI calls. We've put some 9x specific
> optimisations in the next major release (1.4).
>
> 2) Are you using 1.3? If not, upgrade. Its much smaller for many
> cases.
>
> 3) What DPI are you using? Eg if the page is being generated as a
> raster, do you really need 1200dpi? That's going to create a file
> that's 16 times as big as the same output at 300dpi.
> Select the lower-res option in the printer properties.
> BTW printing as a raster is only used in 1.3 when compositing, or
> you draw using a paint which isn't a solid color. Perhaps you can
> avoid these things which trigger the problem. If you really need
> to use alpha colours (for example), then doing just that
> part of the rendering into an image and then drawing that image
> to the printer graphics will also avoid the large spool files.
>
> 4) Here's a performance tip: when printing look at the clip.
> Sometimes you can save by being smarter about not rendering
>outside
> the clip. This tip most usefully applies to the case when the
> page is printed as a raster in multiple bands.
>
> 5) Are you printing a Swing UI?
> Consider that printing Swing components is often better
> done by explicitly drawing it into a image and then printing that
> image. This is because Swing paints its components back to front
> and some rendering is simply being overwritten by newer rendering.
> This doesn't matter for a raster printer, but Postscript printers
> record all of that and replay it.
> Apps which exhibit similar behaviours: eg if you take a terrain
> map, draw it, draw other images, annotations etc on top, may also
> see the problem that they are replaying all of that in PS,
> rather than just the final result.
> This is one case where for some pathological examples printing the
> page as a raster may be better because there's so much rendering
> to be replayed in Postscript.
> The app's way to print as a raster is to draw into an image as
> described earlier.
>
> Here's some data points to compare your experiences:
> JDK : 1.3
> OS : NT 5.0 SP5
> Printer: HP 5000 (Postscript) at 600dpi
> Paper: 8.5 x 11 with 1in border (ie 6.5in x9in used)
> Content: 1 page with approx 20 images with the entire imageable
> area having a black background:
> Spool file size: 400Kb.
>
> If I "force" the same page to be printed as a raster, which
> is what happens if you are using paints other than solid colours,
> then the same job becomes 3.5Mb. Still more or less manageable.
>
> However it is true that this is not the worst case scenario which
>is for
> Postcript at 600 dpi an uncompressed page size raster of RGB
>colours
> encoded using ASCII which may be around 80Mb. Printing that does
>take
> several minutes, almost entirely for transmitting the spool file
>over
> the cable to the printer. The tips above may help find a way
> to improve where you encounter cases like that.
>
> Note that the same raster sent to a cheap inkjet is about a 2Mb
> spool file, although at a lower resolution (360dpi).
>
> Also as mentioned above 1.3 is smaller for many cases. We've
> tried to make the raster case get hit for fewer & fewer use
> cases. Fewer developers are hitting the remaining cases.
> If test cases can be provided to illustrate real problems
> people are still having with their apps we can use those to focus
> resources on improving those cases.
>
>-Phil.
>
_____________________________________________________________________________________
Get more from the Web. FREE MSN Explorer download : http://explorer.msn.com
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA2D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
