Skip to main content
Jump to: navigation, search

Difference between revisions of "Recommenders/FAQ"

m (Can't download any recommendation models. What's wrong?)
m (Can't download any recommendation models. What's wrong?)
Line 9: Line 9:
  
 
== Can't download any recommendation models. What's wrong? ==
 
== Can't download any recommendation models. What's wrong? ==
If your .metadata/.plugins/org.eclipse.recommenders.rcp/ folder stays almost empty either your repository URL is incorrect or your company network requires you to go through an internal proxy. 1.0.0 does not support proxies. A patch is in progress. See http://www.eclipse.org/forums/index.php/mv/msg/366941/894052/#msg_894052 for details how to patch Recommenders manually and build it using maven tycho.
+
If your .metadata/.plugins/org.eclipse.recommenders.rcp/ folder stays almost empty either your repository URL is incorrect or your company network requires you to go through an internal proxy. 1.0.0 does not support proxies. A patch that respects the Eclipse proxy settings is available on the HEAD update sites. See http://www.eclipse.org/forums/index.php/mv/msg/366941/894052/#msg_894052 for details.
  
 
== Error "Total path and file name length must not exceed 260 characters" when Unziping Java Package ==
 
== Error "Total path and file name length must not exceed 260 characters" when Unziping Java Package ==

Revision as of 01:03, 11 July 2012

Does Chain completion work on default content assist list?

No. Computing call chains completion performs an exhaustive search on the API graph which would slow down completion to much in most occasions. Thus, it can only be put on subsequent content assist lists (2nd, 3rd etc.). Actually, it deactivates itself when it found itself being used on the default content assist list.

Does Dynamic Templates completion work on default content assist list?

No. Computing code snippets takes some time which would slow down completion to much in most occasions. Thus, it can only be put on subsequent content assist lists (2nd, 3rd etc.). Actually, it deactivates itself when it found itself being used on the default content assist list.

Does Subwords work with Mylyn and JDT completion on default content assist list?

No. Subwords deactivate itself as soon as it detects its put on the default content assist list together with Mylyn or JDT's Java completion. Only one of these engines can be used on the default tab.

Can't download any recommendation models. What's wrong?

If your .metadata/.plugins/org.eclipse.recommenders.rcp/ folder stays almost empty either your repository URL is incorrect or your company network requires you to go through an internal proxy. 1.0.0 does not support proxies. A patch that respects the Eclipse proxy settings is available on the HEAD update sites. See http://www.eclipse.org/forums/index.php/mv/msg/366941/894052/#msg_894052 for details.

Error "Total path and file name length must not exceed 260 characters" when Unziping Java Package

Please use Winzip or 7Zip to extract files. In addition, these files are generated by Maven. As they are not used by Eclipse they can be ignored. Also see https://bugs.eclipse.org/bugs/show_bug.cgi?id=384529

How can I build Code Recommenders from source?

See Recommenders/BuildingFromSource for details.

I'd like to participate. How?

There are different ways how to participate. Have a nice idea? Get in contact with us in the forum or developer mailing list. If you have already existing code (from a former project or research project) you may be interested in Recommenders/Incubator project.

Can I generate my own models ?

Generally yes but there is yet no API to support for this. To build and provide you own models, you have to (i) provide your own model repository and (ii) provide your own model archives. Models for, for instance, call completion can be generated by yourself by specifying the patterns for you libraries explicitly using BayesianNetwork class in org.eclipse.recommenders.bayes.serialization. Some more documentation about the network structure will follow.

Back to the top