Archive for Januar, 2008

Messages from happy users

Montag, Januar 28th, 2008

Zeit für ein bisschen Bauchpinselei.

Bug Reports

Sonntag, Januar 27th, 2008

Heute habe ich mal wieder einen guten Artikel zum Thema Bug-Reports gefunden. Der Artikel von ESR, How to ask questions the smart way, sollte ja hinlänglich bekannt und referenziert sein. Aber how to report bugs effectively liest sich gerade aus der Sicht eines Programmierers ganz gut, und ist vor allem um einiges kürzer. :) Insbesondere die Zusammenfassung ist kurz, prägnant und lesenswert für jeden Bug-Melder.

Es gibt auch eine deutsche Übersetzung.

Berlios kills projects

Freitag, Januar 18th, 2008

Two days ago, the Berlios Developer site lost one of my projects: Cython. I have no idea where it went, it just disappeared silently. The last mail went through the mailing list at 18:42, 2008/01/16, and since then, the list is dead and the project is no longer available from the developer site. I’ve been trying to contact any of the admins through either the problem tracker or the support form on the Berlios site, but didn’t get a response for two days so far. Personally, I find that a pretty inacceptable delay, but the back list in the problem tracker actually speaks for itself: the first open issue dates back to mid 2005! And Cython doesn’t seem to be the first project to have disappeared. I guess we’re just lucky we weren’t using their Subversion…

Looks like that was the last project I’ll ever host there…

XPath 2.0, XSLT 2.0 and Python

Montag, Januar 7th, 2008

I noticed a few people getting desperate about XPath/XSLT 2.0 support in Python, especially since the normal sources like XML-SIG or comp.lang.python magically failed to provide helpful answers in the past.

A good place to look is the web site of the W3C XQuery group, which has a general and quite comprehensive list of XQuery 1.0 implementations. By design, these also implement XPath 2.0. Now, some of them provide their XPath2 support separately, some additionally implement XSLT2, and some even seem to have Python bindings. So there should at least be something useful to start from.

Another take on this: while some implementations are in C (just as the CPython interpreter) or at least C++, many others are written in Java. GNU’s gcj is pretty good in compiling such tools and libraries into executable binaries and it also has support for generating the necessary C headers. What I’d love to see is one of the good XPath2/XQuery tools compiled as a Python extension and/or interfaced with lxml at the C level.

One project that did something like that was PyLucene. They now seem to have switched to a new tool called JCC, which (according to the docs) generates a complete CPython extension from a Java application that interfaces with the JVM through the JNI. While I like the simplicity of this (I haven’t used it yet), I would still prefer a GCJ compiled binary over a dependency on a full-fledged JVM. But I definitely like both approaches. Any volunteers? :)