Iterate over a list of attribute names to compare in 2 datasets

Dear All,

I am facing the following issue. I have 2 datasets (which are quite large) that contain a series of common attributes. I have a 3d table that stores the attribute names that I would like to compare (no geometry comparison).

What I would like to do sounds very simple (sorry, I am newbie to FME...) :

1. loading the 2 data, add prefix, merging them based on a key attribute (they will then have attributes like source1_col1, source2_col1, source1_col2, source2_col2, etc.)
2. load the data that contain the column names I want to compare (e.g. col1, col2, etc.)
3. iterate over col1, col2, etc. and for each of this value use TestFilter testing sth like "@Value(source1_$(column)) = @Value(source2_$(column))", write the unfiltered in an excel file on "$column" sheet)

I have tried thousand things (custom transformer, ChangeDetector, workspace runner), but could not achieve exactly what I would like to do.

One solution works rather well : I have a parent workspace that loads the column names, and I use a WorkspaceRunner that takes as input each of the feature (=column name) and in which I load the 2 datasets and performs the comparison. The issue here is that it is not efficient at all -> for each column name, the 2 datasets are loaded each time.

Ideally, I would like to pass the 2 (already) loaded dataset to the WorkspaceRunner but so far as I know, this is not possible ??

Alternatively, I've tried to use a CustomTransformer which takes as input the 2 already loaded datasets, but then I don't know how to pass the column names inside the CustomTransformer ??

I was thinking about using a loop inside the CustomTransformer (and pass the column names as input to the CustomTransformer for the loop), but I did not succeed. Would it be possible to do it in this way ??

Or maybe my way to approach the problem is not a good one at all ?

In brief, I am quite desperate because I was initially thinking that it would be very simple to do, but I spend hours without obtaining significant results... So if any of you would have suggestions, I will be very grateful 🙂

Thanks in advance

NB : I have attached some of my trials and toy data