CDT/calls/Indexing23102006

Agenda (Proposed)

• Do we need to co-ordinate unit test writing?
• Performance testing
• 74433 (offline/composite indexes) update
• Discussion of index model (IndexFactory)
  • Should an indexer attached to project A write to other projects
  • Do we want to model static functions, unnamed namespaces in the index?
• How to present elements from the index in the UI.
• Proposal: Use String instead of IPath in IIndex and IIndexFragment to reduce dependencies for remote clients

Attendees

• Doug (QNX)
• Markus (WindRiver)
• Chris (IBM)
• Vivian (IBM)
• Jason (IBM)
• Mike (IBM)
• Andrew (Symbian)

Content

• Co-ordination of unit tests
  • We're happy for duplication to happen - more tests is better than fewer
• Performance testing
  • Andrew to try to get this working locally
• 74433 update
  • proof on concept is working, by having new PDOMBinding records used instead of placeholders
  • work has begun on an CIndexProvider extension point. Proposed draft patch to follow
  • a IPDOMVisitor impementation for traversing multiple pdom has been tested, we might not need this
  • question over whether we need a composite implementation for bindings at all - it seemsnamespaces are the only type of binding which surface substructure that is alterable outside of a translation unit. Andrew to raise a bugzilla about whether these methods could be removed
• static functions
  • Markus to comment in bugzilla
• anonymous namespaces
  • Markus to remove them from index (they go in under the empty name)
• Finding the enclosing function for a reference - this is not currently stored in the index
  • Suggestion of storing additional location information in a special definition class
  • Markus to raise a bugzilla
• Standardization of displaying objects from the index in a UI
  • Lifecycle issue - objects returned from index queries can expire
  • Markus suggests using ICElements to store information
• IBM propose using String rather than IPath in index API
  • Discussion of whether IIndexFragment can be public long term. Doug says ways of organizing this to come from guidelines on managing APIs
  • We need to support EFS also - which uses URI (part of the JDK - so may benefit IBM)
  • Solving relocatable PDOM's is probably orthogonal to EFS vs IPath
• Talk about resolving bindings from the AST against the PDOM (rather than the DOM)
  • Andrew to write unit tests as this overlaps with offline indexing
  • Markus to spend some time looking at code, and maybe adding comments
  • Doug says Andrew Niefer is probably still happy to donate time (as the original author) to help with questions