A problem you might encounter is that certain MGET tools access numpy’s binary interface. These tools utilize Python extension modules (which are just Windows DLLs) written in C++ that bind against numpy’s own extension modules. Because numpy has introduced breaking changes into their binary interface before, we adopted the policy of always recompiling our extension modules each time a new version of numpy comes out, and then detecting at run time which compiled version should be loaded. Because 10.5 appears to use numpy 1.9.3 and MGET only includes extension modules compiled up through numpy 1.9.2, MGET will likely complain that an unsupported version of numpy is installed when you try to run a tool that accesses the binary interface.

 

It is unlikely that numpy broke the binary interface between 1.9.2 and 1.9.3. (If they were going to do that, they would have revved the version number to 1.10.) Therefore it is probably possible to reuse MGET’s 1.9.2-versioned extension modules with numpy 1.9.3. Basically, you’d need to recursively find files in C:\Python27\ArcGIS10.5\Lib\site-packages\GeoEco named *_1_9_2.pyd and copy them to *_1_9_3.pyd.

 

But before you try this hack, I suggest you wait for this problem to occur, then contact me to confirm that is the issue.

 

Best,

 

Jason

 

 

Whoops, yes, that is the problem. I forgot that ArcGIS changes the Python installation directory when the version number is changed. You should be able to do this:

 

1.    Copy the entire directory tree C:\Python27\ArcGIS10.4\Lib\site-packages\GeoEco on the 10.4 machine to C:\Python27\ArcGIS10.5\Lib\site-packages\GeoEco on the 10.5 machine. This directory should not exist yet on 10.5. This is where MGET’s code really lives. If you look at the scripts referenced by the toolbox .tbx file you’ll see they are just short stubs that call stuff in the GeoEco package.

 

2.    Copy the file C:\Python27\ArcGIS10.4\Lib\site-packages\ _GeoEcoArcGISHelper.pyd on the 10.4 machine to the C:\Python27\ArcGIS10.5\Lib\site-packages directory on the 10.5 machine. The fact that this file is missing on the 10.5 machine is what is causing the scripts to execute in 0.00 seconds successfully without doing anything.

 

3.    Maybe necessary: On the 10.5 machine, recursively delete *.pyc and *.pyo from C:\Python27\ArcGIS10.5\Lib\site-packages\GeoEco. These are compiled versions of Python modules that are created when MGET is installed. The installer creates them to speed execution of Python code. Because you potentially have a different version of Python with 10.5 and a different processor or operating system architecture than the 10.4 machine, these compiled files might not work. (There might not be enough differences, though; the existing files might just work.)

 

Let me know if that works.

 

Jason

 

 

Hi Jason,

 

Thanks for the reply. I do not see an interactive plot as you describe. That’s what I see on my other machine which has an installation with 10.4.1 and works fine. What I’ve also noticed is that the installer does not install or compile anything to the Python27 folder. This may be the real issue since that is where the business code lives. I doubt I can copy/paste from my successful MGET installation on my 10.4.1 machine’s Python27 folder to my 10.5 machine… but if that would work I’d try it.

 

 

-Joe

 

 

Hi Joseph,

 

Thanks for your interest. Sorry we have not released a 10.5 compatible installer yet. We not likely to get to it until January. Sorry for the inconvenience.

 

The only step of the installer that I expect to fail is the part where it tries to add the toolbox automatically. This part of the installer depends on some .NET ArcObjects APIs that were not documented, at least at the time the installer code was developed. In general, Esri changes the .NET assemblies sufficiently between minor releases (e.g. 10.4 to 10.5) that we have to rebuild this part of the installer. It may be possible to accomplish it without recompilation using a .NET publisher policy file but I have not attempted that...

 

In any case, if only that part of the installation failed, then MGET should be fully installed except for the adding of the toolbox to the default ArcToolbox list, which of course can be done manually, as you have done. Assuming 10.5 does not contain any changes that would break MGET, then you should be able to use MGET normally.

 

It does not sound like the Density Histogram tool is working correctly, however. By default, that tool should create an interactive plot. The tool should stay running until the plot is closed. A successful run should look like this:

 

Executing: RExploratoryPlotsDensityHistogramForArcGISField C:\Temp\Estelle\simu_dir_test\ReefData\reef_geometry.dbf AREA # # # # # 1000 3000 3000 10 white

Start Time: Wed Dec 21 15:57:06 2016

Running script RExploratoryPlotsDensityHistogramForArcGISField...

Querying ArcGIS table "C:\Temp\Estelle\simu_dir_test\ReefData\reef_geometry.dbf".

Finished retrieving rows: 0:00:00 elapsed, 444 rows retrieved, 0:00:00.000466 per row.

Waiting for R windows to be closed...

Completed script RExploratoryPlotsDensityHistogramForArcGISField...

Succeeded at Wed Dec 21 15:57:18 2016 (Elapsed Time: 11.56 seconds)

 

The Elapsed Time will be based on how quickly you close the window.

 

Are you seeing that? If not, there could be some other problem to investigate. Let me know and I can advise you on what to check next.

 

Jason

 

 

Hello and Happy Holidays!

 

I understand there is no distributed installer for the newest release of ArcGIS 10.5, however, I figured I’d give it a go anyway. The installer fails saying that ArcGIS is not installed, presumably because it is doing a check on the ArcGIS version and 10.5 is not in the domain which is being checked. However, the toolbox and scripts have been extracted to the GeoEco folder in my Program Files directory so I figured I’d add the toolbox manually and go from there. I was able to do that but upon running a quick tool (Density Histogram for a point field) I see that the histogram does not display or save to a file, but the tool runs successfully.

 

Has anyone else tried MGET tools, as currently distributed, with ArcGIS 10.5? I realize it has just been released and I work with Esri, but figured I would ask anyway.

 

 

Thanks!

Joe

 

----------------------------------------------------------------------

Joe M^cGlinchy | Imagery Scientist

Esri Professional Services

ESRI || 1 International Court || Broomfield, CO 80021 || USA

T 909-793-2853, ext. 4783 LinkedIn