Provide consistent ordering for JavaClass.getBeanProperties

I agree, ordering of collections should be consistent across JDKs, and across runs of QDox. IMHO source elements should be listed in the order they were declared in the original source.

I thought this was largely the case already, actually; in most cases, we maintain ordered lists of elements, append to the end while parsing the source, and make them available as arrays in the API.

Obviously, there must be cases where this is not true (or you wouldn't have raised this issue). Konstanstin, can you provided specific examples of calls that return Collections/Maps with inconsistent ordering? Thanks.

for example JavaClass.getBeanPropertyMap()
gathers them into HashMap, and then converts
values() to array - using threeMap instead helps

I' attaching small patch for just one change.

just small patch for JavaClass

Thanks Konstantin; I've re-titled the issue. I'll apply your s/HashMap/TreeMap/ patch as a quick fix.

But the behaviour may change again, as longer-term, I'd prefer the ordering as declared in the source to be preserved.

well, this is not only JavaClass, but everywhere else...

Sorry; there are plenty of Collections that do have consistent ordering, so "everywhere else" doesn't help me much. Can you be more specific about where ordering is inconsistent?

• Bean properties, we've already covered.
• Looks like named-parameters of tags might have the same problem?
• Anywhere else??

Specific use-cases would be appreciated. JUnit test-cases would be even better!

By the way, why is consistent ordering important for Generama test-cases? Can you provide an example of a test that fails?

generama test cases use ast parser to compare generated java files.
well, it's not generama which uses them, of course it's unit test
for xdoclet-2 plugins which happen to generate java code.

There is manually written java source, which serves as "correct"
code, and it is compared with generated one.

If I iterate over properties inside some method - test fails.

-------%<---------
setFoo("blurge");
setBar("bang");
-------%<---------

is not equal to:

-------%<---------
setBar("bang");
setFoo("blurge");
-------%<---------

I will raise dependency on xdoclet plugins and look where similar problem may arise.

Okay, I'm convinced Thanks.

I've switched tag named-parameters from HashMap to TreeMap, which will hopefully fix your problem for now.

Stay tuned, though, as I intend to change this again: order will be based on relative position in the source, rather than alphabetical.

well, if you do it after 1.6 ( on 1.7 ) everything would be fine,
unless some unlucky soul raises dependency on generama / xdoclet-2 / xdoclet plugins

I will certainly alerted on the spot by failed unit tests...

Done; I added a dumb Map implementation that returns things in insertion order.

Konstantin, please let me know if problems remain.

No complaints. Case closed.