I have been using the OmniBroser framework to extend the Lint browser
(OR2LintBrowser) and i found some strange behavior i would like to comment:

1) When you select an item in the browser, the metagraph define some edges
leading to other metanode types.  This is translated sending the edge names
to the real nodes and getting all the following nodes, and then, those nodes
are displayed on the following column at the browser, and if there isn't a
next column it is created.  The strange issue here is that, if the messages
for edge taking on the concrete nodes retrieves an empty collection of
nodes, the next column is created anyway, because for the browser there is a
next metagraph node.  Is this a problem or not? (I had to define much more
complex metagraphs because of this, just to avoid an empty column).

2) When you use filters on the metagraph, they are applied over each taken
edge leading to the next metagraph.  The filters are usually used for node
ordering and here is a problem.  If you have more than one edge from the
current metanode, you will get a order filter applied over two different
edges and then, the order will be not ok, since there will be some partially
ordered concatenated node lists, and not just a fully ordered list.

3) Another issue is not about omnibrowser, but i think is worth to be noted. 
If you extend any class in another package using *package-category protocol
for new or modified methods, you will see the extended class on the package
root, but you won't see it at the cathegory (i think it would be nice and a
little more clear seeing the extensions at the specific package-category). 
Also i found that TestRunner do not put the unitTest classes under a package
unless the class itself where defined at the selected package (even if i
added some test using the *package-category protocol name)

I hope some of you found this intresting

cheers


-- 
View this message in context: 
http://forum.world.st/Some-questions-about-omnibroser-framework-tp3173047p3173047.html
Sent from the Pharo Smalltalk mailing list archive at Nabble.com.

Reply via email to