Anyone have any thoughts here? Any time I think about it, I just end up thinking its best to leave it ... the javadoc itself (package descriptions) still appears to come out correctly.

just stating the obvious, but for the analysis.cn, the problem is that I think we want javadocs in

• analyzers javadocs jar
• smartcn javadocs jar
• javadocs-all target (which one?, this is causing the error)

it would be nice not to have the error, but also still have package.html in all 3. any ideas?
the whole point of the package.html in this case is to help users figure out what all the chinese analysis options are...

How about putting the package.html files one level lower into the smartcn package? the package.html of the top-level analyzers's doc could then be only in the common contrib.

Uwe, i think you are right, but the real problem in this case is that SmartChineseAnalyzer is not in the smartcn package.

it is under o.a.l.analysis instead of o.a.l.analysis.smartcn (with all the other smartcn tokenizer/tokenfilters where it should really belong imho)

if we put package.html and SmartChineseAnalyzer one level lower, things would make more sense in my opinion.

+1

I was wondering why the analyzer is not in the sub-package. I think we can change this even in release phase (as the whole package is experimental...)

I think we can change this even in release phase (as the whole package is experimental...)

+1

Just as an expansion on my feelings about any changes:

I posted the rules that are suggested on the wiki - but I think they are a bit harsh (eg only serious bug fixes). My thought is that the important part is :

Keep in mind that it is our main intention to keep the branch as stable as possible.

I think that anything is fair game as long as its clear it will not affect stability. If everybody thinks something is a good idea, and they don't think it has the reach to affect stability (or undercut the testing that has already occured), I don't see why we wouldn't do it. As long as its discussed first and given a bit of time to ensure consensus.

patch for the smartcn issue.

mkdir contrib/analyzers/smartcn/src/test/org/apache/lucene/analysis/cn/smart
svn add contrib/analyzers/smartcn/src/test/org/apache/lucene/analysis/cn/smart
svn move contrib/analyzers/smartcn/src/test/org/apache/lucene/analysis/cn/TestSmartChineseAnalyzer.java contrib/analyzers/smartcn/src/test/org/apache/lucene/analysis/cn/smart
svn move contrib/analyzers/smartcn/src/resources/org/apache/lucene/analysis/cn/stopwords.txt contrib/analyzers/smartcn/src/resources/org/apache/lucene/analysis/cn/smart
svn move contrib/analyzers/smartcn/src/java/org/apache/lucene/analysis/cn/SmartChineseAnalyzer.java contrib/analyzers/smartcn/src/java/org/apache/lucene/analysis/cn/smart
svn delete contrib/analyzers/smartcn/src/java/org/apache/lucene/analysis/cn/smart/package.html
svn move contrib/analyzers/smartcn/src/java/org/apache/lucene/analysis/cn/package.html contrib/analyzers/smartcn/src/java/org/apache/lucene/analysis/cn/smart
patch -p0 < patch.txt 

note: i did not make any change to analysis/cn/smart/hhmm/package.html, although it should be modified like the other files. the problem is it has a stray control-M, (as well as not having eol-style set), so it is impossible for me to provide a well-formed patch that will actually apply. we should fix this in the future!

Anyone have any thoughts here? Any time I think about it, I just end up thinking its best to leave it ... the javadoc itself (package descriptions) still appears to come out correctly.

Uh ... not really. what you get is non-deterministic behavior, where one of the package.html files for each package gets picked, and the other one isn't used. (this can look particularly confusing with something like queryPrser, where you'll find one description in the "core" docs, a differnet version in the "contrib" docs, and it's a crap shoot as to which one of those will show up in the "all" docs.

Robert: i opened LUCENE-1882 to track the issue of moving SmartChineseAnalyzer into the correct package, and committed your suggested changes under that issue.

i know for a fact that part of your patch didn't make it in – the characters kept getting corrupted in one of hte package.html files, and i couldn't figure out an obvious reason/solution so i just committed what there was at that point so we'd at least have all the files in the right places to move forward from there.

please take a look at the current state of the pacakge.html files and let us konw what still needs done to make them "good"

FYI: the queryParser contrib doesn't even have any classes living in the org.apache.lucene.queryParser package, so contrib/queryparser/src/java/org/apache/lucene/queryParser/package.html is never even used when building the contrib-queryParser specific javadocs ... it's just a possible candidate when builidng the "all" javadocs.

the best solution seems to be moving that content into contrib/queryParser/src/java/overview.html ... i'll work up a patch.

patch fixing the duplicate package.html files for queryParser by moving the contrib version into the contrib's overview.html (the package one was never used in the contrib specific docs)

this patch also makes some other misc improvements to the docs, and tweaks the build.xml so that the appropraite subpackages are listed in the correct section.

LUCENE-1862-queryParser.patch ...

Committed revision 810286

this is done ... if people want to make other improvements to the javadocs for either package, let's open separate issues.