Automatically focus on frequently used entries about jabref

[koppor]

@ThiloteE The idea is that JabRef automatically tracks readings - without any additional user input. Explanation of the general idea in German: https://de.wikipedia.org/wiki/Mylyn

Scientific background:

• Focusing knowledge work with task context. PhD thesis. Available at: https://www.researchgate.net/publication/235350419_Focusing_knowledge_work_with_task_context
• What You See Is What You Need: WYSIWYN: Using Task Focus to Ease Collaboration. [PDF]
• Beyond Integrated Development Environments: Adding Context to Software Development. This paper reasons on an "Automated Assistant" and calls for a deeper investigation of the concept of "context". In contrast, the ContextSwitcher assumes that the user knows about his context and defines queries for the context by hirself.

[ehehela]

To be exact, JabRef should count the number of times that the entry is accessed in a period, e.g., the number of times the attachments (PDF files or others) were accessed in a month.
Or perhaps, JabRef could add a new column named ViewCount to log the number of times in a period set by users. In this case, users can quickly find entries by sorting this column.

[ehehela]

I was struggling to articulate my thoughts, but thankfully, @koppor accurately and professionally expressed what I wanted to say.

Take myself for instance, there are many literatures regarding one research topic. Unread and low-quality entries are automatically moved into "Inbox" group, in which priority is used to mark entries to be read and quality-improved first. Other entries are in specific groups according to their research fields. When I decide to do research in one certain direction, I usually first locate a group then scroll down and up to query related papers which may well be the key references and will be looked up many times in the future. However, these papers are scattered in different (sub-)groups. How to quickly get these needed and strongly correlated papers? Maybe Jabref could guess according to user behavior, e.g., times of opening attachments. Maybe some smart groups will promote this. Or like the following drafts (illustrative, not necessary).

[ilippert]

I guess, it might be possible for JabRef to be aware if an entry was engaged with - though I would want too disable it. For enabling it, an additional timestamp might be used that is updated whenever an entry is shown in the entry editor, or a click on it in the entry table happens.

[jiahengguo123]

Hello, I am a new contributor. And I will carefully read the contribution guidelines. Could you please assign this issue to me?

[ThiloteE]

@ehehela could you please describe your workflow where you need this? Why do you want to know this information? Is this a shared library or do you work alone? Have you ever thought about simply adding frequently used entries to groups or labeleling them as important? We even have a priority label.

You also could create automatic keyword groups that capture entries with certain keywords. E.g.

If we were to implement what you suggest, would this not introduce performance issue for large libraries? Also, how to reset the counter? Where to store the counter? - on pc or in library file? What happens, if multiple users have access to the library? Do you want to have separate counters for each user or multiple counters for each? ....

[koppor]

To be exact, JabRef should count the number of times that the entry is accessed in a period, e.g., the number of times the attachments (PDF files or others) were accessed in a month.

This is a nice thing. Implementation: Use a MVStore provided by h2 stored in the directory retrieved with

Path.of(AppDirsFactory.getInstance()
                                         .getUserDataDir(
                                                 OS.APP_DIR_APP_NAME,
                                                 "views",
                                                 OS.APP_DIR_APP_AUTHOR))

The database knows tracks for each citation key the number of views. Granularity: one hour (that means, if accessed at 10:05 and 10:43, it counts as one).

Or perhaps, JabRef could add a new column named ViewCount to log the number of times in a period set by users. In this case, users can quickly find entries by sorting this column.

This is the second thing to implement so that the data can be accessed.

8 years ago, we had explicitly writable fields (refs #574). I think, we should have a org.jabref.model.entry.field.FieldProperty adding INMEMORY stating it should not be written and not be synchronized to a server.

Hints

• On load of the library the field viewcount has to be filled based on the MVStore content
• The MVStore is updated on each view
• Use the obersable JavaFX lists to "listen" to changes (personal remark: I like the event bus more, but we decided to focus on one technology)
• In case the citation key is renamed, it has to be renamed in the count, too

[koppor]

A new intelligent group should be introduced. This is comparable to BibDesk's "Smart" groups.

• Add preference as outlined at screenshot at #10418 (comment)
  • Use "enable internal user behavior analysis" (to indicate that no data is shared to an external server)
  • 
• Add group "Popularity"
  • 
  • Name: "Popularity" (capital "P")
  • Enabled only if above preference is enabled
  • Do not show "Ignore zero times of read". The group always shows entries with read > 0. For the other feature, we
• The views need to be tracked in a MVStore
• Introduce new method getOtherDataDir in org.jabref.gui.desktop.os.NativeDesktop
• Store the MVStore in tracking.mv in getOtherDataDir. Hashtable <BibEntry, Integer>. Hashtable name: hashCode() of fileName of library (or BackupFileUtil.getUniqueFilePrefix(file) + "--" + file.getFileName()). In this way, each library tab has its own hash table
• Hook into org.jabref.gui.LibraryTab#showAndEdit to update the counter