As you say, working with Interlis isn't always easy and it can be quite frustrating, the format is anything but trivial.

That said, I'm surprised that you get slightly different results on FME Cloud compared to FME Server, they should be identical.

I would say that the first thing to do is to thoroughly compare the job logs from FME Cloud and FME Server to see if there are some clues there as to the different results. I'm not certain that it's the writer in itself, it could even be some of processing done before the writer.

Remember that FME Cloud runs on Linux and that there might be some slight differences to FME Server on Windows when dealing with external resources etc.

Hi Maria,

Hi Maria (@gpt_executer ),

if you can share a simple workspace with log files that runs successfully on FME Desktop but fails on FME Cloud, please feel free to share and I'll give it try on a cloud instance.

The format was also updated for 2017 so it might be worth to run it on a 2017.0 or 2017.1 cloud instance. Can you share the exact version of FME Desktop & FME Server (build number) the workspace ran successfully with?

Hi David, Stefan and Gerhard,

thanks for your hints! As my affected workspace is quite huge/complex/"mazy" it's hard to read and compare the log files.

I compared the versions of the ili-Reader/Writer. They are the same. So it cannot be due to the version.

I've created a simple workspace: read xtf, offset the geometry and write xtf. As expected: it works in both, Desktop and Cloud.

But I came to another stumbling block, which I couldn't solve yet: Within the workspace some additional geometries are generated/calculated. One feature can carry one or more geometries within attributes besides its own geometry. I set the "Geometry Encoding" parameter to "FMEBIN", in reader and writer. Even within the workspace (ie GeometryReplacer) I chose "FMEBIN" as the Geometry Encoding Type. Still the same: Cloud writes only one geometry whereas Desktop writes all the geometries.

Might that be a hint I should pursue?

I would suggest to add an FFS writer to your workspace and to send all your output data to both the ili2fme and the FFS writer at once. Then upload to FME Cloud and execute.

Download both the .ffs and the .xtf and compare the contents. If the .ffs exhibits the same problems as the .xtf geometries, I would say it's probably a problem with the workspace rather than the ili2fme writer.

You should also triple-check that the input data is exactly the same on FME Server and FME Cloud.

I don't get it: I added an FFS-writer but my output zip-file doesn't contain that *.ffs-file. The log file says it wrote it to FFS. Did I miss something?

In the Desktop version the FFS file looks good. It contains all missing attributes with its geometry saved as FMEBIN.

Desktop and Server does work a bit differently. When republishing the workspace to FME Cloud you'll have to tell FME that you want the FFS writer result rather than the ili2fme one.

I got it. Thanks!

Now I see that the geometries exist, but they are not written to Interlis.

My current conclusion: An Interlis feature can carry more than just one geometry. The first geometry is the one I can see in the Inspector. All other geometries are carried as attributes. With the parameter "Geometry Encoding" of the Reader I can control the way the geometry is stored (FMEBIN, FMEXML, ...). In the output of FFS it look like this:

The writer then should recognize this on its own and write the correct geometries.

As my desktop installation runs on windows and the FME Cloud on Linux, might there be a problem there?

Ok guys,

I've created a simple workspace. It runs fine on desktop and even on our server installation (both windows 64bit), but fails on cloud. The only difference I can see is the operating system: windows vs. linux.

I'm still in contact with the support of safe. @XiaomengAtSafe and her team are on it.

Any hint is appreciated!

We have been in contact with the developer and were able to identify the problem that caused different translation results on Unix based operating systems compared to Windows. The plugin has been updated and the newest version is available here: http://www.eisenhutinformatik.ch/interlis/ili2fme/ili2fme-6.2.0.zip

A request has been filed to upgrade the ili2fme plugin, that ships with our products, to the latest version.