Repeated phrases should be done on corpora

[benel]

... because getting it for the whole database can be far too long.

Tests done in Liège: more than one minute for a 50 Mb phrases list.

[christophe-lejeune]

Moreover, in its current state, the 50 Mb phrases is completely reloaded each time "repeeted phrases" is hit (even when no changes occur in the database).

[benel]

Yes, I don't understand why it's not cached.

[benel]

Another complementary optimization would be to call this view asynchronously (with jquery get for example).

[benel]

The cache seems to work on smaller corpora. The problem could be due to the size of the view...

[benel]

When it is restricted to a single corpus (e.g. 2009_ordinateur), the cost is 5.4 Mb in 6 s.

If a list filters phrases to keep only repeated ones, the cost is 730 kb in 32 s.

Note: Tests done

• with curl because REST Client has problems with the raw view,
• on localhost (it would be interesting to test that on a distant server, when the payload size is more important).