[texworks] Wishlist for pdf previewer
Charlie Sharpsteen
chuck at sharpsteen.net
Sun Sep 25 23:25:15 CEST 2011
On Thu, Sep 22, 2011 at 9:19 AM, Jonathan Kew <jfkthame at googlemail.com>wrote:
> One thing I've been meaning to suggest: it might be worth experimenting
> with the Cairo backend to poppler (in place of the Splash renderer). Some
> comments I've seen (sorry, don't have references offhand) have suggested
> that it gives superior results; and in particular, it might serve as a
> stepping-stone towards printing support (as cairo can, for example, render
> to a Windows GDI printing surface or an OS X Quartz context).
>
> AFAIR, the combination of the Qt4 bindings and the Cairo backend is not
> "officially" supported by the main developers (cairo is associated with GTK
> rather than with Qt), but I believe it's possible to make them work together
> - I think I recall someone posting (to the poppler mailing list? cairo
> list?) about it in the past.
>
> Alternatively, it might also be worth seeing whether the Arthur backend
> (i.e. the Qt-specific poppler renderer) is usable, and how its performance
> and quality compares to Splash. When I started TW, Arthur was in a seriously
> broken state, but there has been work on it since then so it may be in
> better shape, and in principle it too could offer hope for better printing.
>
> JK
>
Arthur is definitely out---tried it and the results weren't that good. Some
major quality issues:
- Arthur looks like it may be using a different font choices as the text
looks like it is set in a different face compared to the Splash backend.
- The Arthur output device doesn't understand some graphical elements
used by TikZ---nodes are consistently bigger and bolder then they are
supposed to be. Gradients are off. The front page of the TikZ manual looks
like a mess. Most everything else looks passible aside from the text issues.
- Anti-aliasing is pretty poor.
Performance is a real killer---it looks like Arthur is up to an order of
magnitude slower than Splash when it comes to rendering output. Plus there
appears to be a fixed overhead of 200-400 ms, even for simple text only
pages, which is about as long as it takes Spalsh to render a very
compilcated page on my machine.
I also tried the 0.17.4 release candidate and there wasn't
any noticeable improvement.
-Charlie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://tug.org/pipermail/texworks/attachments/20110925/553e57a2/attachment.html>
More information about the texworks
mailing list