Chembrows is slow when data base contains too many references

[JPFrancoia]

When the database grows, it becomes longer and longer to calculate the hot paperness for all the articles.

Actually, what's time consuming is writing all the percentages in the database.

Possible solutions:

• Delete old articles from database (option to set the time limit?). But which articles should be deleted? Liked? Unliked? No discrimination? Deleting read articles (read or not) would affect the quality of the predictions

• Not write anything in the database and calculate the percentages on the fly when articles are displayed. Calculations would be performed in the delegate. Percentages need to be calculated to be sorted by relevance.