Something broken in 2.062

[jsc]

David,

I've been working with some large files and having no trouble with 2.060. However, I just tried 2.062 and when I try to load the csv file DPlot starts using 100% of the CPU and seems to hang, ie, a long wait and the file is not read.

I reverted to 2.060 and the file reads without trouble.

I'll send you the example.

Jon

(Yes, yes, why am I playing with this Saturday morning! (should go on a bike ride!))

[DPlotAdmin]

Jon,
Thanks for the file (and yes, you should definitely be on a bike ). But your file worked fine for me.

The guestimation might still need a little work: initial max number of points was 100; after reading 100 values DPlot guessed that 236544 were required. Then it read 236544 values only to find that that wasn't quite enough. Next guess was 237568. Actual number of points in the file is 236800.

My initial thought was that there is some combination of initial max and the actual required number that might result in an endless loop, e.g. Not enough points, set max to 236544 and restart. Not enough points, set max (again) to 236544 and restart, etc. But I've stared at the code and don't see how that would be possible.

What happens if you double-click the "Max Size" area of the status bar and set the max number of points to something > 236800 before reading the file?

[jsc]

David,

Yes, I did get that ride in. It was hot today! (Of course MI can't compete with MS on that account, but it's hot for Michigan!)

I did what you suggested - actually, I didn't realize that you could set the array size, or needed to - and set the array to 2x236800 and the file read into 2.062 as easily as it did in 2.060, taking about 15-17 seconds on my 1.73 gHz Pentium M with 1 gig of memory.

Then I reset the array back to the default - had not paid attention, but thought it was 9x65536 - and tried again with 2.062. This time I carefully watched the status bar report the array size. It quickly jumped from 9x65536 to 1x197632 and sat there for quite a few seconds (maybe a minute?). Then eventually it increased to 198656, then slowly to 199680, and continued to do that, in regular 1024 jumps, until it eventually hit 237568 several minutes later. Then it plotted the graph.

Yes, looks like the initial guess is close, but if it is not enough, the algorithm seems to give up and goes into a very conservative 1024 step mode like it is creeping up to the the right number. I watched the status bar every step of the way, and it was always a constant 1024 jump.

Now here's the interesting part. I tried a couple of files with 1x452200 points. I left the initial array size at 9x65536, and loaded the file into 2.062. DPlot quickly guessed the array size and jumped it to 1x459776 and read the file in without any difficulty!

So this is odd!

[DPlotAdmin]

So this is odd!

Yes it is. This obviously needs work. With the two large files I worked with the "guess" was fairly close and large enough on the first pass, but that was apparently just dumb luck. Thanks for your detailed reply, that will be very helpful. I'll post an update as soon as this is fixed, fixed in this case meaning "works better".

[DPlotAdmin]

Fixed a silly mistake. You can get the update with Help>Check for Updates. Thanks for your input on this.

[jsc]

Thanks Dave. Works great now.

Jon