Version 4.0.5: variable "note" can't be tested

bpreusche

Hi,
the following IF-test doesn't work correctly with the new version 4.0.5:

Settings: only the variable "note" is empty (no entry)!! The test should go to ELSE...

<if variable="author note" match="all">
<text value="should not be this"/>
</if>
<else>
<text value="should be this!"/>
</else>

The test without other vaiables is correct:

<if variable="note">
<text value="With these settings, not this!"/>
</if>
<else>
<text value="It's this one!"/>
</else>

Ciao, Berni

fbennett

Both tests work correctly in the current release of the processor. Is your test style valid CSL?

adamsmith

works correctly for me. We need to see this in context.

bpreusche

Here is a gist of my style. https://gist.github.com/anonymous/5407554

Perhaps there is an error in the info-part (?), because you are right: when i installed the style, i got the hint like: the file is not a validate CSL 1.0.1 style and it could work uncorrectly (the error is in germsn-lang.) but with zotero 4.0.4 it worked correctly.

adamsmith

I can confirm that with the simplified style. (It doesn't validate because you can't have a space in the id - it validates with http://www.zotero.org/styles/berni4 ). This has nothing to do with note but seems to be a problem with match="all" in that setting. e.g. "author abstract" also evaluates to true even when the abstract is empty.

bpreusche

besides the match-error above, the validation-error occurs again without space in the ID:

<?xml version="1.0" encoding="UTF-8"?>
<style xmlns="http://purl.org/net/xbiblio/csl" version="1.0" class="note">
<info>
<title>berni4</title>
<id>http://www.zotero.org/styles/berni4</id>;
<author>
<name>Bernhard Preusche</name>
<email>BPreusche@gmx.de</email>
<uri>http://wherever.com/</uri>;
</author>
<category citation-format="note"/>
<category field="theology"/>
<updated>2008-10-29T21:01:24+00:00</updated>
<summary>My great new style format.</summary>
<rights>Keine Rechte!</rights>
</info>

adamsmith

https://gist.github.com/adam3smith/5407923
validates, both in the validator and when installing.
You need a citation element.

The problem is the matching bug, the validation works fine.

fbennett

I've tracked down the bug. It may not be my doing.

That test creates two functions, bundled as a list. The list of two items is passed into the "all" function that evaluates them -- which receives a list with only one item in it. There isn't any code between those two steps, which has me wondering whether this might be a bug in the Firefox JavaScript engine.

A test built with the style runs fine in the Rhino JavaScript engine that I use to test the processor. That also makes me suspect the Firefox JS engine as the culprit.

fbennett

If someone has Firefox 17 installed, I'd be curious what the style does in that environment. My tests were done with Firefox 20.

fbennett

Okay, I'm pretty well convinced that this is due to a bug in the Firefox JS interpreter. I've found a workaround, though, that will produce correct output in the case, and will release the new code a little later in the day.

fbennett

Hold that thought! My apologies for the string of premature notes above. I believe that I do own this bug.

When I get this sorted out, this thread will be closed. :-)

fbennett

Finally got my head around this.

Technical run-down: In new code, when the style is built I'm doing preemptive execution, to encapsulate the test. The problem is that the attributes are evaluated in @bpreuche's test in a different order in Rhino (the testing engine) and in Tracemonkey (the Firefox JS engine). That's the bug in the processor: once it's fixed, it will come right.

fbennett

Fixed. You can test the revisions by installing the processor patch plugin.

bpreusche

Thanks a lot, now it works.
cu