The problem was that an automatically triggered 'idle' completion result
could come in while we were waiting for actual 'triggered' completion
and then the real completion result would be discarded.
To avoid that:
* Only accept proposals from the currently running completion.
* Increase the idle completion timeout to 400ms so it doesn't get
triggered during regular typing.
Change-Id: Ibb2fa8907a2b67a94e092dfdc5a48fb50be06b65
Reviewed-by: Leandro Melo <leandro.melo@nokia.com>
Even if the completion popup is already visible is should
respect the behavior expected for explicit invocation.
Change-Id: Ib4635fead060b90f677bfd7b2751a20c030f67a1
Reviewed-on: http://codereview.qt.nokia.com/571
Reviewed-by: Thorbjørn Lindeijer <thorbjorn@lindeijer.nl>
This is necessary for the very first characters typed on the editor
in the case the activation sequence length is greater than one.
Change-Id: I9ec611f816278795b7ac3f75913fccf5a52165bf
Reviewed-on: http://codereview.qt.nokia.com/47
Reviewed-by: Roberto Raggi <roberto.raggi@nokia.com>
This is a re-work of our completion engine. Primary goals are:
- Allow the computation to run in a separate thread so the GUI is not locked.
- Support a model-based approach. QStrings are still needed (filtering, etc), but
internal structures are free to use more efficient representations.
- Unifiy all kinds of *assist* into a more reusable and extensible framework.
- Remove unnecessary dependencies on the text editor so we have more generic
and easily "plugable" components (still things to be resolved).
