Climatology

[rgleason]

Sean,
Just to stress test the Isobar function, I took most recent install.exe compile (v3.2 opencpn.lib) for Climatology and used it in 3.3.1824 most recent.
Checked sea level pressure and it ran fine.
Then set "steps" to "very fine".
Initially the progress bar ran fast the firstt 30 seconds or so, then it slowed to a creep and the time clock did not update but froze. So what I did was periodically hit the minimize / maximize button to rewrite the Opencpn screen, thus I could see that the time clock changed and the progress bar continued. The cpu did not seem to be working excessively and the bar was progressing. So I waited and did min - max to be sure it was continuing. 1.45 minutes....2.05 minutes ...2.25 minutes and it finally completed, but I did not know this until I hit MIN - MAX to shrink the program and rewrite the screen, then the last time I did that the progress bar was gone and the isobars were all written.

That is what happens. There are a couple of things that need to be done I think.
1. Perhaps "fine" and "very fine" need to be dropped or the amount of detail reduced so as not to overburden the cpu.
2. The isobar routine is working I think, but for some reason it looses contact with refreshing the screen as to its status, and it does not indicate when it is finished. These are little details, but I think they would be very helpful to the user, because most users would not think to hit MINimize - Maximize to refresh what is on the screen.

Hope that helps.

[seandepagnier]

Which exact datatype and what spacing setting is very slow?

Do the other isobar settings work faster? Can you cancel the generation?

I think the solution is to update more frequently, and/or abort if it takes more than a few seconds.

[rgleason]

Sean,

In climatology configure, when you tab out of a field isobar starts rebuild....currently.

I think we need an "Apply" in the "Climatology Configuration" Panel so that a user configures and then selects "apply" so the program does not get confused or start doing its work before the user intends. This will reduce freezeups when user changes something, the isbar gets started building, then user is not finished and trys to change something else. While the isobar is building, perhaps all fields are greyed out.... while it builds. Then when it is finished the fields become available for change...

Also there does not seem to be a "Cancel" button in the isobar progress bar that I am using. ...

I am using climatology with v3.2 right now. Ah, the screen refresh is lost if you double click on the up arrow for example under isobars spacing. Then you have to hit min / max and the progress bar is completed and screen is refreshed. If user does not know to do that, it will look like the progress bar is frozen.

So if the user has changed something, isbar operation is put on hold to let user make changes, multiple changes. Then user must hit "apply" and the boxes go grey while isobar runs.... when isobar is complete... the configuration fields un grey and can be selected again. I think the grey of fields only needs to be for Isobar operation...but I leave that to your judgement. Also I think the isobar popup needs a "Cancel" in it.

I think this will help a lot, but I'll check on the other questions later about limits and data types.

Pressure works fine down to 1 and in fine, perhaps very fine is not needed?
With a setting of 4 and very fine it takes 40 seconds and I have to hit min-max.
When I have other data up like sea temp it takes longer.

[rgleason]

Need to have a note that shows somehow, when user has turned off Opengl and has checked data-types that show shades, because shades do not show. Don't know how to do this so it does not become a bother though... with popups. Perhaps there could be a message area in the Climatology Configuration screen at the bottom when features are checked (shades) that require Opengl be turned on.

[rgleason]

Seadepth Isobars checked with 5000 spacing, normal step.
Isobar ran 1/4 across with the clock still at 0:0:0 and seemed to stop responding, then I hit min / max at about 20secs the isbar had completed and was gone with the isobars drawn.

On a separate matter, perhaps uncheck the isobar check box when a user starts changing fields...., thus the isobar routing will not start until user checks the box again.

Still in v3.2

[seandepagnier]

I took the cancel out on windows only because I thought it might help. Apparently not.

It looks like the progress dialog is a bit broken on windows. I don't know what you mean by "shades"

I can make it disable isobars on windows only if you change the settings. Using apply is very annoying as I want to see updates instantly.

I think maybe you need to build the plugin in release mode. I made some changes which should speed things up a bit.

[rgleason]

From recent compiled 3.3.1824 (release version)
with the very recent git pull of climatology 0.11
See attached.

Created an isobar on almost every data-type and in a number of cases set an increment of 1 and at level "fine".

1. What I did was make sure in the Climatology Display Control that no data types were checked.

2 .Then I went into "Config" and configured isobars for each datatype (with no isobar being run at this time because it is not selected for display.)

3. Then I closed config.

4. Then I checked one data-type at a time and let the isobar progress bar run and finish. Then once finished checked the next date type.

5. Then I tried unchecking several and rechecking and it was faster display.

6. There were no clitches during this process. See the attached showing the amount of isobar with all checked. Seems pretty stable using it in this way.
====

Earlier I had tried the MSVC debug with climatology, had some bombs for some reason so I went to the above process. I will try again, after I get this to bomb out on me.

*** I think the key is that if someone is changing a datatype config, the display for that datatype should be turned off first until the user is finished adjusting and goes to a new datatype or hits "refresh" or "apply" to turn on the display of the data type again.

I understand you want it to display immediately after a change with no "apply" needed and I think that would be ideal too, but windows seems to get screwed up if the user makes more than one change and does not wait for completion of the calculation. --Perhaps you need a user override on the isobar calculation when the user tries to make a second or third or fourth change to the config, such that the isobar calculation is cancelled and the second or third change is then entered, and then after a brief pause (user selected?) the isobar tries to run the progress bar again and completes. The other way to do it is to prevent any further user configuration changes until the isobar has completed.

--Just thinking about it. here. Not finished yet testing.

[rgleason]

Sean I just tried making sure Sea depth was unchecked in the display control. Then we and changed sea depth to very fine in config and closed config. Then checked sea depth in display control. Isobar ran efficiently and took about 15 seconds.

After that I tried something simple like air temp and had that checked in display control, went into config under air temp, and changed a setting once. Isobar instantly opened and ran through completely. I did not interrupt it by hitting another spacing or step. Everything worked fine.

Then with air temp checked under display control, and isobar checked under config, with spacing at 3, I tried changing step from Normal to Fine, then while isobar progress is still running and popped up, I tried changing step from Fine to Very fine. The time clock stopped at 2 seconds and the progresss on the progress bar appeared to freeze, but I now know (most users do not) that it will complete eventually. I min the program, then max it, several times, and I see the progress bar has popped up again and is at 13 seconds. It is still in a frozen state, but we can take snapshots at sequential times by using min - max. Now it is at 1:46 to do the "very fine" steps for air temp. Min-max again, now the time clock say 2:32 and the bar and time are still frozen. We must be patient because there is no "Cancel" bar.

I hope this fully explains the problem now. If you are patient it will complete but it it very much slower. Say 3 minutes when it normally takes 11 seconds.

I think the user should be allowed one change, if the data-type is checked in the display control. The isobar comes up and runs with "cancel" showing. The Config fields are greyed and do not allow input. To interrupt the progress bar and make any further changes, the "Cancel" Isobar progress bar must be hit. The greyed out Config fields become active again and user can make changes.

The problem appears to be user's making two changes, the second one after the progress bar has started running. What happens is the screen clock and progress bar disconnect from the display interface so it seems frozen, and the processing gets longer. The user is never advised when it is completed. Prevent the user from making two or more successive changes if the data-types display control is checked. There are several ways of doing that.

1. In config, regardless of the state of checkbox in display control, if user makes a move to change something under the ISOBAR, the check is turned off automatically first. Then after they get done fooling around with settings. They just hit the check box again.

2. In config, if checkbox in display control is checked, and if user makes a move to change spacing or step under isobar, then the display control check is turned off, user makes his changes, closes "config" and the check is put back into that data-type and the isobar runs.

I prefer #1 alternative.

[rgleason]

In Opencpn 3.2.0 - I just installed Climatology 0.11 and ran it. It worked great when a "display control" data-type was checked, and in "Config" same data-type isobar was checked, and you DID NOT select a second settting when the isobar progress bar was running! It worked like a champ. Also when I first loaded climatology it inherited the setting from the 3.3.1824 setup earrlier, so it had a lot of isobars to calc. As I turned them off and then on, the progress bar came up and the longest one was 15 seconds. The key is not to interrupt it.

Is there a way to make a "cancel" that works?

Note: this ran in Opencpn 3.2.0 ! and I compiled it with the opencpn-3-2.lib.
Have not tried v3.2.2 but I expect it to work.

Do I need to run in debug stilll? Please advise.

[seandepagnier]

I tried very hard but cannot seem to get problems no matter what I do. Everything works perfectly. The cancel button works.


Are you sure this is the latest git?


The way you have a problem is if it is building the isobars and while the progress is displayed, you enable, say the overlay map correct? Currently it grays out the climatology dialog but not the configuration dialog. Would greying out the configuration fix this?

[rgleason]

Sean, I git pull and new files came down. Ran cmake... only one warning:

Build succeeded.
"C:\Data-Dart\Up-Soft\Navigation\o-plugin\climatology_pi\build\ALL_BUILD.vcxproj " (default target) (1) ->
"C:\Data-Dart\Up-Soft\Navigation\o-plugin\climatology_pi\build\climatology_pi.vcxproj" (default target) (4) ->
(ClCompile target) -> ..\src\ClimatologyConfigDialog.cpp(269): warning C4800: 'long' : forcing value to bool 'true' or 'false' (performance warning) [C:\...\o-plugin\climatology_pi\build\climatology_pi.vcxproj]
1 Warning(s)
    0 Error(s)

Will check it now.
FYI -Sean, I am building this with my Opencpn-3-2.lib (opencpn.lib from version 3.2.0, as I have done before). --Have not heard an answer to my question if opencpn.lib for v3.2.0 and v3.2.2 is the same, but the file is the same size.

[rgleason]

Sean, Using current climatology with v3.3.1824 recent compiled as release. This works much much much better! It behaves very well. When I have a data-type checked to display, and hit Config, and set the dropdown to that data-type and then proceed to change step or spacing, the display of the isobar is unchecked just before the value is finally changed by the program... very good. This occurs whether or not the Data-type Display check is checked or not.

Thought: Program could also check if the data-type "Display" is unchecked and if it is unchecked, NOT uncheck the Config Isobar check, because isobar progress will not run. But if this is too complicated, don't do it. It works now and is a consistent user interface.

Problem: Of course after exercising this a lot. I went for the jugular... sea depth.
Display checked, Config Isobar checked, Highlighted spacing "5000" changed to 1000, the config isbar unchecked, under step selected "fine". Then checked the config isobar. The progress bar came up and for some reason it froze on 0:0:0 clock. I minimized the program and then maximized it and it bombed and I sent in a report. This only happened once!

I tried to make it happen again bringing sea depth spacing down to 200 feet and set on fine then back to normal and back to fine, and each time it ran beautifully! I am puzzled. It must have been a fluke. I will send log file and settings file so maybe you can see what the problem was, but I think it is largely fixed and much much better. Congratulations that was a tough one. Good job. Thanks!!

PS: The above was done with opengl off. I then turned Opengl ON and it worked even better, using sea depth, down to 200 spacing and fine and back up..awesome.

I am going to upload this to Opencpn Beta File Thingie now for other BETA USERS.

[rgleason]

Isobar also works in Opencpn 3.2.0 now! Just fine.

Maybe that one bomb on sea depth above was just a data glitch? Is there a check and an error message that pops up for that? I've never seen one, but perhaps that would slow the machine processing down.

Anyway for the WINDOWS BETA TESTERS please give this a final spin and let us know if there are any problems, beat on it and try to make it fail. Sean wants to push this out to release NOW. Please post your results here, positive or neg (with log file if crash). THANKS.

Opencpn Beta File Thingie rguser, rgpass

Directory: 7.1 Plugins v3.3.1824
File: climatology_pi-0.11-win32.exe install file with the data directory

also
Directory: 7.1 Plugins v3.2
File: climatology_pi-0.11-win32.exe install file with the data directory

Look for date 7/4/2014 6:58 am on the file in File Thingie.

These are the same file. Post your results please.

[seandepagnier]

Your log of the crash did not include a backtrace.

If you can reproduce this condition repeatedly there is a chance to fix it. In any case, if you cannot repeat it more than once, then it maybe isn't worth bothering with at the moment.

[rgleason]

Sean,

With climatology in opencpn 3.2.2
With opengl not checked it seems fine.
With opengl check it also does shading and there is more processing, so it is more important how much isobar is drawn...
In this opengl mode with overlay map still checked on most everything... I tried turning all isobars (I will give setting below), on and got several situtions where the isobar progress screen popped up and the clock was stopped. When I hit min/max the isobar had finished calcs and was gone.
See attached artic artifacts and progress bar.

Then under opengl with all display types checked and with config open I checked isobar and turned it off on each dropdown data-type with no problems... see the this list of settings I think is reasonable for my machine.
Wind 5 normal
Currents 4 normal
Sea Level Pressure 2 normal
Sea Surf Temp 5 normal
Air Temp 5 normal
Cloud Coverr 10 normal
Precip 1 normal
RH 5 normal
Lightning 30 normal
Sea Depth 5000 normal


With everything on in the Display Panel, everything checked, and if Isbars is checked on for each datatype, plus the overlay is on for each datatype, that is a lot of calcs, so it is slower and isbar regen is more of a problem.

If you turn off opengl, the overlay is turned off so if all isbars are on, it will be faster, but still slow.

Opencpn 3.3.1824 seems to be a little better about this due to the opengl and optimizing.