On 5/22/07, Taco Hoekwater wrote:

Oliver Heins wrote:

...

I noticed that the latest revisions (441/442) are very slow. I just switched back to rev 339, which runs much faster (approximated 10 times).

Yes, that is (more or less) on purpose

Care to elaborate? Something new and interesting going on? Thanks. nikolai

On 5/22/07, Taco Hoekwater wrote:

Nikolai Weibull wrote:

...

On 5/22/07, Taco Hoekwater wrote:

...

Oliver Heins wrote:

...

...

I noticed that the latest revisions (441/442) are very slow. I just switched back to rev 339, which runs much faster (approximated 10 times).

...

Yes, that is (more or less) on purpose

Care to elaborate? Something new and interesting going on?

Hans and I are working on a new feature we call "attributes".

An attribute is an item that becomes attached to the actual nodes that are to be typeset. These can then be queried and/or acted upon in the node processing callbacks, and could for instance be used for color, glyph substitution, storing annotations in the pdf, or even for re-linebreaking a paragraph.

The abstract idea is pretty clear, but the literal implementation isn't -- it requires many small changes to the data structure and the routines that handle them. The current implementation (as started in #441) is pretty slow and wasteful on resources, to prevent bugs stemming from optimization from obscuring bugs caused by incorrect data structure assumptions.

OK. Sounds cool. Thanks for the information. nikolai