Quickmark completion

[claudehohl]

gnah, I created a new branch on top of my importer commits... I guess I totally fucked it up this time.

Here is the quickmark completer. With lousy performance, any tips how to access the marks values more 'directly'? :>

---

This change is 

[The-Compiler]

No worries - mind rebasing it though?

From the top of my head:

git checkout master
git pull
git checkout quickmark-completion
git rebase master
git push --force

This looks like the most performant way it can currently be done, I'd have to test and profile it to say more about why it's slow.

[The-Compiler]

I did a profiler run to find out why it's slow:

I guess I could gain some time by not calling sourceModel() and caching it instead (no idea why it takes so long...). Also maybe there's some way to optimize lessThan a bit. I'll take a look at it.

[claudehohl]

Something is kaputt now.. When I scroll through the completions, the completer "recompletes" itself with one result - the highlighted quickmark. Hmm...

[The-Compiler]

Thanks, I fixed this now I hope.

Is it intended that the user can complete the URLs instead of the names? I'd have expected to complete the names with :quickmark-load. What about doing a QuickmarkCompletionModel and a OpenCompletionModel, and the one for :open at the moment just contains the URLs for the quickmarks, while :quickmark-load completes names?

[claudehohl]

Excellent idea.

[The-Compiler]

Also the model currently won't update when the user adds/removes quickmarks.

However doing this would mean we'd have to convert the stuff in quickmarks.py to a class inheriting from QObject, and adding pyqtSignals when the data is changed.

Is it okay if I do that, and then you can update your part accordingly?

[claudehohl]

Of course. Thx!

[The-Compiler]

You can remove this import now after 74839d7

[The-Compiler]

Done now - rough guide to get it auto-updated:

• Move the initializing back into its own function (sorry!)
• Add a on_quickmarks_changed function decorated with @pyqtSlot() which does self.clear() and calls the initializing function again.
• Make the _models attribute in utils/completer.py:Completer public (s/_models/models/g)
• In app.py:_connect_signals, add something like:

quickmark_model = objreg.get('completer').models[usertypes.Completion.quickmarks]
quickmark_manager = objreg.get('quickmark-manager')
quickmark_manager.changed.connect(quickmark_model.on_quickmarks.changed)

Though the question is how long it takes to re-build the whole completion model... if it should take too long, a solution would be to do the changes twice in quickmarks.py (once in the model, once in the marks dict), but I'd like to avoid that.

[claudehohl]

Dear Der-Kompilator, it's redi to mörtsch!

[The-Compiler]

Since we don't use this as a slot anymore, can you move the stuff below into __init__ directly?