Ok, now i'm curious about something.. I know its possible to designate multiple keys to a single sort using the form.. sort lines of plist by item 1 of each & item 2 of each & item 3 of each & item 4 of each which works fine. But changing it to.. sort lines of plist ascending numeric by item 1 of each & item 2 of each & item 3 of each & item 4 of each -- reverse this for your desired sort doesn't sort numerically (even if each key is forced to numeric with a + 0)
BTW, I had to go back to comments in an older version of LC to remember how to do the multiple-key sorts. Since it ignores the "numeric" part, if one limits ips to all 3 digit numbers, the sort works as expected despite sorting as alpha. So my question is this.. Would it be a reasonable feature request to adjust sort so that its multiple key format respects the numeric sort type? On Fri, Jun 29, 2018 at 10:46 AM Bob Sneidar via use-livecode < [email protected]> wrote: > A realistic expectation of IP addresses for a given network might only be > around 16,535 (class B network) assuming every address was in use, an > unlikely scenario. I thought of way to do this for an extremely large > number of addresses by reading a large file in 1000 line chunks into 4 > columns of a SQL memory database, then querying using concatenation and > sorts on the 4 columns, and using limit 1,1000 1001,1000 2001,1000 etc. and > writing back to another file. The time to do this of course would be much > longer, but it would avoid any memory constraints. > > Bob S > > > > On Jun 29, 2018, at 09:35 , Mike Bonner via use-livecode < > [email protected]> wrote: > > > > ## was writing this when your response showed up.. > > Just did some tests. For reasonable size lists of IP numbers, your > method > > is clearly faster. As lines increase the disparity shrinks (due to 4 > > processings of the same lines rather than 1) until around 35000 lines at > > which point positions reverse and the disparity grows in favor of the > > function(each) method. > > > > After just a bit more testing, the fastest method (for anything over > 20000 > > lines) is to run through the whole list first converting it to numeric > and > > then do a single simple sort, but that leaves you with a straight list of > > numbers that would then have to be re-divided into triads, so the post > > processing needed kills the whole idea. > > > > As for your response, > > Yes the function is called by sort for each line and returns the numbers > > to use to give that line its sort value. It isn't as fast as I had hoped > > most likely because it has to call a function for each line in the list. > > With the first method I posted, on my machine, the crossover point is > right > > around 35000 lines (at least on my system, at 35000 lines it takes 475 > > millisec to 612 millisec vs the 4xsort) At 200,000 lines the disparity > > grows to over 3 seconds difference, but i'm unsure what max length list > > might be reasonably expected. The 4 sort method is by far the simplest > to > > code and is plenty fast for list under 100000. > > The nicest part of your method is that if processing a huge list, its > easy > > to give visual feedback between each sort if need be. But again, all > this > > is likely moot unless the ip list is huge. > > > _______________________________________________ > use-livecode mailing list > [email protected] > Please visit this url to subscribe, unsubscribe and manage your > subscription preferences: > http://lists.runrev.com/mailman/listinfo/use-livecode > _______________________________________________ use-livecode mailing list [email protected] Please visit this url to subscribe, unsubscribe and manage your subscription preferences: http://lists.runrev.com/mailman/listinfo/use-livecode
