Loading of manual results into the DB

[mdettweiler]

Originally Posted by kar_bon

What about the user 'Unknown'?

There're still 18 primes (300<k<400) found by this user and ~400,000 pairs in the database!

That's still pending Dave's coding the "upsert" procedure we'll need to insert those pairs under the correct usernames and have them replace the "Unknown"s, rather than be rejected as duplicates. From what I understand that's a somewhat complex task.

Dave, how's that coming? Do you by chance have an ETA on when that will be ready?

[AMDave]

Acknowledged.
The retrospective aggregations have just now been refreshed after the manual loads which included Max's updates of 'who-did-what'.
/ed - except for the 3 primes under investigation which unfortunately is still pending testing & I have to check the other 15 -ed/

[mdettweiler]

Originally Posted by AMDave

Acknowledged.
The retrospective aggregations have just now been refreshed after the manual loads which included Max's updates of 'who-did-what'.
/ed - except for the 3 primes under investigation which unfortunately is still pending testing & I have to check the other 15 -ed/

Cool--I see that "Unknown" has been accordingly knocked down to 26th place (from somewhere in the earlier 1x's). However, I also see that he still has 87427 pairs to his name--is there by chance still something we missed that the upsert didn't cover? I don't have time right now but I can check later today to see which drive and range the remaining "Unknown" results are from.

[henryzz]

Would it help if I submitted some results with the username Unknown? Then we could find them easily.

[mdettweiler]

Originally Posted by henryzz

Would it help if I submitted some results with the username Unknown? Then we could find them easily.

Hmm...actually, I think that would more serve to complicate matters since if we query the database for a list of all Unknown results, yours would come up as well. What we'd be looking for instead is the large, essentially contiguous bunches of results that would surely be produced by such a query; those indicate entire server results files that were never imported the first time around and were therefore caught in the manual import.

[kar_bon]

If your're looking at the prime list on NPLB-stats using participant='Unknown' as filter, there're 18 primes shown:

15 primes are of the range 300<k<400 and 340k<n<384k so it's the range from Drive #3 of LLRnet port IronBits 5000! These are 76,999 pairs!

The other 3 primes are from same drive, IB5000 too, range 566k<n<576k. This makes 19,177 pairs.

So all over 96,176 pairs, a bit too muchas the current 87,427, but these should be the ranges to look further!

[henryzz]

Originally Posted by mdettweiler

Hmm...actually, I think that would more serve to complicate matters since if we query the database for a list of all Unknown results, yours would come up as well. What we'd be looking for instead is the large, essentially contiguous bunches of results that would surely be produced by such a query; those indicate entire server results files that were never imported the first time around and were therefore caught in the manual import.

I was joking

[mdettweiler]

The first individual-k drive (k=300-400, n=600K-1M) has now been processed for importing to the DB and is awaiting Dave's triggering of the process.

BTW: yes, I didn't forget about those remaining "Unknown" results. I'll try to track those down as soon as I get the chance.

[mdettweiler]

I've queried the DB for a list of all "Unknown" results and dumped the output to http://www.noprimeleftbehind.net/max...s_20100521.txt.

The ranges currently claimed by "Unknown" appear to be as follows:
1st Drive, n=510K-510.5K
Ergh, this was the one that was done by me manually and which I accidentally imported under "Unknown". I pointed this out in my previous itemization of "Unknown" ranges but never got around to fixing it. Dave, quick question: would there be any negative side effects if I remedied the problem with the following command?
update results set username = 'mdettweiler' where (k between 400 and 1001) and (n between 510001 and 510500);

1st Drive, n=~543038-~543749
This was in the middle of LLRnet C443's range, n=542.6K-544.0K. It would appear that one or more results files from in here were never imported (probably an accidental blooper, since importing of C* server results was a manual process). Karsten, do you have results files that cover this range? If so, please send them to Dave and I and we can get them in the DB using the "upsert" process.

3rd Drive, n=340K-~383274
This rather large range was done by LLRnet IB5000. My guess is that it was right around here that IronBits started copying off his results files automatically; indeed, the very first results file by date that we have in our master directory starts at about n=383274 (not counting a few assorted stragglers). Again, Karsten, if you have results files for this range, please send them to Dave and I.

Max

[kar_bon]

Originally Posted by mdettweiler

Again, Karsten, if you have results files for this range, please send them to Dave and I.

Sent to Max! Please check them if these all of the needed.

[mdettweiler]

The first individual-k drive (k=300-400, n=600K-1M) has now been imported into the DB. Since this drive was almost entirely manual, this represents a rather large jump in the stats (especially for Ian).

From here on out, the remaining drives to be imported represent a comparatively small amount of points. Therefore, we shouldn't have any more enormous jumps in the stats.