[cairo] Cairo & libpng/zlib parameters

[Jeroen van Rijn]

On Sun, Jan 17, 2010 at 11:37, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> No, the PNG support in Cairo is principally there for debugging purposes,
> it was never intended to be a complete interface. That task was better
> suited for a dedicated image decoding/encoding library. If you are happy
> with PNG as your graphic format, the simplest approach would be to take
> Cairo's code and make it extensible for your needs and overload the
> current write_to_png Python function.
>
> Hope this helps, and you have fun using Cairo.
> -ickle

After I add the optipng stage to the daemon pending the research I'm
doing, I'll focus the task on investigating the overloading of
write_to_png to use devIL or similar, provided that library allows
more bare metal access to libpng. I'd put that in as an optional
dependency, i.e. if detected it offers to forego the optipng stage in
favour of directly choosing a favorable set of parameters.

This might well be the sensible thing to do in case Cairo's png debug
path ends up being marked deprecated in the future.

I think I can work with this, thanks.

> P.S. I'm looking for rendering torture tests to add to the performance
> benchmarks in cairo-trcaces. The first one I took using oscitymap, shows
> that there is a marked improvement with the development branch, but I
> would like more examples to broaden our focus.

I'll certainly give it some thought, and with the CC to MapOSMatic's
dev list, maybe others will have some ideas too.

Internationalisation is well underway (it's how I got involved with
the project). If you'd like to run some cairo torture tests with
ocitysmap, cairo and pango to stress RTL rendering, that's a
possibility I can think of. Other than that, there's looking through
the maposmatic web service job list to identify long running tasks -
as they signify the more complex maps - and zeroing in on those maps
specifically for your test set?

Best regards,
Jeroen.

-- 
"That's like saying that a squirrel is 48% juicier than an orange -
maybe it's true, but anybody who puts the two in a blender to compare
them is kind of sick." -- Linus Torvalds