"To Case or not to Case" was the follow on question and after sorting/tallying millions of results, it came out that the order in which to test worked out to be [for the particular mix of results]

190     Select Case valst 'Case may be unsorted, numerical and string!
            Case 1: Statuses = "Valid"
            Case 0: Statuses = "Pending Validation"
            Case 4: Statuses = "Pending Verification"
            Case 2: Statuses = "Invalid"
            Case 3: Statuses = "Too Many Errors on Wingmen"
            Case 5: Statuses = "Too Late"
195     End Select

Hardcoded now, as why look something up that has not changed in the past 15 years, or however long BOINC has been with the crunching world (bar that Case 4, Pending Verification, originally was 'Inconclusive'). At any rate, the number of Valid (1) on the result status pages far outstrip PV (0+4), making this the top item to show in the Result 'Status' field, sort of a present ratio of 94835 : 3635 ATM, or IOW, test the case and most of the times the prog can stop looking at Case 1 and not waste nanoseconds on first passing by Case 0. If that is repeated many many, many many times, the nanoseconds become a serious peak-shaving event... when you're waiting for it :D

Have a nice day

(The technique is only build in the 'biggest' version of the hunting tool... the small and medium account size versions just do not benefit in any form or fashion noticeable... no need to spend time on porting back ;)

This morning was merge day, fruit way up in the tree, but did it pay off to make the climb... Various complex helper field formulas previously converted to UDFs, integrated, courtesy of VBA, into a single UDF that produces the result mask, the 6 bit string such as 111111, which means no bad exit code, outcome ok, credit granted, validation state great, modtime in present stats period and not yet on archive, as when it is, the newly fetched result record can be ignored, which is a mask of 111100... easy peesy. Sum of the 3 UDFs into a single is 25% less code, 4 less If/Case tests, 4 less fields to fetch and up to 2x134,000 life results to replicate the UDFs to, and compute on days the FAHV shorts are flowing heavy. All in a matter of hours pensing on the how to (climb the tree safely) and 45 minutes coding.

Made my day.
Have a good one

While cycling, and let WCG be WCG until the last mile, which encompasses a half mile stretch at 18% (18 meters up per 100 meters, to get that pulse racing) the final switch was decided on... no more formula storing, just the output and we have ludicrous speed power 2 with turbo... the yellow with previous runtime on right and new runtime at left, and this is with 143*250 current live results and 177K prior period results on active archive.



Most shocking was the save file time got slashed by a factor 10, the resultonly pull and calc full halved again.

Surprisingly the actual saved file size did not shrink... seems Excel is pretty good in compressing series of identical [complex] formula by keeping one and then tokenize, which is the piece that makes saving and reopening slow.

Time for some vino rosso, doc.

saluti tutti

btw, this is the final. done, dusted, 3.0K build 33-7

For laughs, sometime on entering your account there was Log In and when leaving Sign Out, and was testing on first scrape for Sign Out left top, yes we still prior century scrape for lack of xml stats pages, but Bill visited and changed one of the 2 states, so now it is Log In and Log Out. If the test was <> Sign Out, the Log In procedure was run and the page was scraped a second time... so that started failing "As it so happened", and round and round and round it went endlessly scraping.

Another discovery, Avast is rejecting security of the WCG site, because...



Under Firefox, that's not a problem, there's just the yellow padlock notice, but under IE it is, and MS requires Office to use IE. Think I asked about secure linking of signature picks and what not, as they can harbor iffyness, but think the question was not understood.

The bike's front suspension gave up, a Suntur they frown upon at the bike garage, so flipped the dial from open to lock, and there I had not only a hard tail, but a hard top as well :O)))

While rattling back the last third, some gray cells woke up and produced some synaptic action of the oh-yes output type on late validations, our good techs deem to still needing to go into the last stats period, like 15 results this time, and never seen 'that' late...

Last ModTime
17-04-29 12:06:13
17-04-29 12:06:14
17-04-29 12:06:14
17-04-29 12:06:13
17-04-29 12:06:13
17-04-29 12:06:14
17-04-29 12:06:14
17-04-29 12:06:15
17-04-29 12:06:13
17-04-29 12:06:15
17-04-29 12:06:15
17-04-29 12:06:13
17-04-29 12:06:15
17-04-29 12:06:15

worth -15 0:02:51:30 WCG ran late

The cut-off officially was at 4/29/17 12:06:03 (UTC) [2 hour(s) ago, and sure enough, just needed to extend the chronological modtime selection with these and bingo. Now, when these are post-period identified, usually fully automatic on post-archive run, they are also moved to archive and would have disappeared in the pile, currently some 200,000 results, that 'oh yes, why not?' event came up with an additional code line to have these late validations suffixed in the Archive Notes column with "Missed", and sure enough, from Total Recall to Total Recon, and back... simple db filter for last period with a 'missed' note, and we have the 15 broken back for reminiscence.

Whilst on a short roll, churned out a half dozen more pop-up comments for the users who wonder 'what's that?', along the line of 'It's this, but DON'T do that then and then!' and away we are.

Time for late late pranzo.