KTp/libkpeople
About
LibKPeople is a contact aggregation library for KDE.
It is *not* a KTp specific library, we are simply the first adopters as well as the developers so the wiki page is here for the time being. This page is about the libkpeople changes needed to make a release of KTp with libkpeople support.
Big discussions about libkpeople should CC the pim mailing list.
Setting up
Install libkpeople (https://projects.kde.org/projects/playground/network/libkpeople) Install ktp-nepomuk-service (https://projects.kde.org/projects/playground/network/telepathy/ktp-nepomuk-service)
start the nepomuk service either by logging out/in or by running "nepomukservicestub nepomuktelepathyservice"
Install and run ktp-contactlist branch mklapetek/kpeople
You may also want to install PersonViewer a tool to view all contact information from all sources (http://quickgit.kde.org/?p=scratch%2Fmklapetek%2FPersonViewer.git)
Tasks
libkpeople
| Status | Summary | Notes | Claimed By |
|---|---|---|---|
| TODO | Add KPeople namespace | 3 votes to 1 on kpeople vs people | |
| IN PROGRESS | Update Roles as per discussion | [mailto:MCK182
<MCK182 >] | |
| TODO | Restore updates on presence change | ||
| TODO | Remove unused / commented out code | ||
| IN PROGRESS | Add dialog for finding duplicates | [mailto:franck
<franck >] | |
| DONE | Add Dialog to view PersonData | (see PersonViewer) in mck182 scratch | [mailto:[email protected]
<[email protected] >] |
| TODO | Add widget to PersonViewer to view associated websites | ||
| TODO | Port away from QStandardItemModel | It's slow and won't be accepted by PIM. PersonItem and ContactItem are fine, we just need to just not inherit from QStandardItem and do model stuff ourselves (apparently) | |
| DONE | Rename examples | [mailto: Aleix Pol
< Aleix Pol >] | |
| IN PROGRESS | Fix tests | PersonData 80% done. PersonsModel to do | [mailto:[email protected]
<[email protected] >] |
| DONE | Cache sub-resources in PersonsData | Vishesh to implement in Nepomuk | [mailto: vhanda
< vhanda >] |
PIM Demos
| Status | Summary | Notes | Claimed By |
|---|
KTp Stuff
| Status | Summary | Notes | Claimed By |
|---|---|---|---|
| TODO | Move KTp plugins to k-ci-i | ||
| TODO | Add PersonData method to KTp::ContactsModel runtime switch from libkpeople or not. | ||
| TODO | Add PersonData method to KTp::Contact | ||
| TODO | KTp Contact has helper methods to get name + avatar for person | ||
| TODO | KTp ContactFactory can get personData as a feature on load, otherwise use Tp stuff. | ||
| TODO | Fix the singleton AccountManager problem | I don't know how many times I've shouted at people about the problem with using AM as a singleton.. and now we've done it in libkpeople. Brilliant. | |
| DONE | Restore Grouping | [mailto:David
<David >]
|
Public API Review
Below are a list of issues, please resolve (either by fixing or discussing) then remove from this list
The list of public classes are:
- PersonsModel
- AbstractPersonsPlugin
- BasePersonsDataSource
- DuplicatesFinder
- Match
- MatchesSolver
- PersonActionsModel
- PersonsPluginManager
- PersonData
Anything else in the main lib is private and thus I don't care.
PersonsModel
Why is addContactsToPerson async but removeContactsFromPerson syncronous?
The code says: "//FIXME: for some reason we get most of the contacts twice in the resultset," What's happening is we run a query "SELECT Distinct ?contactURI ?emailAddressForThisContact" If we have a contact with 2 email addresses, we get the same contact URI twice once for each new email address. (email addresses are just one example, it applies to all things fetched) Current code as far as I can tell, will ignore/hide the second email address. If we're going to do that, we can just simplify the query and make it faster too..or we should try and fix it properly
Why does indexForUri go searching recursively through all the indexes comparing them in turn when we have a frickin' hash table of everything indexed by Uri!
Why is there duplicate detection stuff in the middle of this? It wraps DuplicatesFinder and turns the result URIs into indexes. but why? All the core methods work on URIs anyway, so this just ends up with us having an option to go to indexes and back for no reason at all.
Why is query() a slot invoked through QMetaObject? I assume there's a clever sensible reason, I don't know what it is, so it needs documenting please
I /think/ if you call removeContactsFromPerson that contact is now not in the model, not inserted as a contact anywhere.
Misc thought: createPersonFromContacts et al, don't actually use the model. Can/Should these be moved elsewhere? Is there a case for calling it without the model?