FeatureReader playback woes

Hi Kaitlin.

Your understanding is correct. But the rabbit hole is apparently deeper than expected.

I'm running 2020.2.5, as our FME server hasn't been upgraded to 2022 yet (two earlier attempts has been thwarted by bugs in FME, long story).

So I made a simple workspace with just a Creator, a FeatureReader, and a dynamic writer.

And I first tried it in 2022.1.1, with upgraded transformers, but this fails altogether, even when running the complete workspace !

I then made an identical workspace in 2020.2.5, and it works when running the complete workspace.

However, the funny thing is, when running them incrementally, i.e. first the FeatureReader, and then playback to the dynamic writer, they _both_ fail in the same manner.

I don't know what you've done with 2022 (and 2021), but 2020.2.5 is my very best friend for now.

And just for fun, I ran the 2022 workspace in 2023.0 beta, and updated the FeatureReader.

And it failed to run the complete workspace, just as in 2022.

The playback also failed, unsurprisingly.

Hello @Lars I Nielsen​, I have a feeling this is somewhat expected behavior. The ordering of features really matters for dynamic writing and as you suspect, the schema features must get to the writer before the data features. While I'm not 100% sure, It seems like something in feature caching run (or in other words, the "playback") may not work because something (eg. not an actual feature, but some information) is not passed from the FeatureReader to the writer, that is passed during a non-feature cached complete run. 

There are a few different methods you can use to control the order of features. By using a Sorter transformer, you can ensure the schema features are always read first, even if you use feature caching. Add an 'order' attribute (AttributeCreator) and assign a value of 0 to the schema features and 1 to all the other features, ordering in ascending numeric order to ensure the schema attributes are received first.

Other methods I tried that also worked: FeatureHolder, Decelerator, etc... but if you'll be working with an unknown number of features each run, I might stick to sorting!

Lastly, in order to get any of these methods to work, had to remove the value set for the 'Schema Definition Name:' parameter on the MSSQL_SPATIAL writer. It looks like that parameter is looking for the attribute on the data features and not the schema features. If I add an AttributeCreator after the Gadelys_lyspunkt port and set the attribute fme_feature_type_name to Gadelys_lyspunkt, it works (and then fails for the second feature type because those data features don't have the fme_feature_type_name attribute). 

I’ve attached an updated workspace made in 2023.0, please take a look and let me know if you have any other questions! Thank you for your patience on this! Happy to help, Kailin.

Hello @Lars I Nielsen​, yes, absolutely! I've filed this issue for development to consider (FMEENGINE-76158). I've made a note to have this thread updated as well. I think this may be the parameter that matches the schema feature to the data features. I can look further into this. Happy to help, Kailin.