Dear All,
I have pushed to the branch 'luca' a new version of the code that implements the fast page-analysis mode we discussed.
This is the mode that minimizes the writing of blobs during page analysis, leading to potentially much more efficient analysis of one page at a time.
I have tested the code in the "batch" mode only, but I have not tested it from the dispatcher.
The only way the new code is exercised is from the dispatcher, when the dispatcher asks
updater.ml to update a page.
Ian, Bo, can you give it a try and let me know whether it works? There are two cases that need testing:
- When the page is seen for the first time, that is, updater.ml is updating a page that has no revisions that have been colored yet (so that no entry in wikitrust_page exists for the page, either).
- When the page already contains some colored revisions, and we are using updater.ml to color some more pages.
Can you please give the code a spin and let me know if you find any problems?
Also, FYI, I changed the Makefile; now "make all" makes both what used to be the batch and online parts, and similarly, make allopt.
It is just a simpler organization this way, and the fact is that the batch and online portions of the code are becoming more and more intertwined.
Best,
Luca