ETL - We have the extract, we have the Load as visual elements but we lack the Transform. Understandable with the age of the platform.
Relying on js to do this limits the audience of the tool. implement the transformation as js behind the scenes though as boomi and mule and jitterbit and workato have there own proprietary scripting languages which always felt limiting as ‘everyone’ knows js to a point
BUT there are no excuses for you not to investigate and borrow from Jitterbit (two products), Mulesoft (3 Products), ̶D̶e̶l̶l̶ Boomi, Workato, Integromat, Zapier as they offer trial tiers and they have already invested millions on the decisions you have to make.
First question is who is the audience…develoeprs, low-coders, no-coders…look at mule with their 3 anypoint products anypoint studio, flow designer, orchestrator…all using the same underlying tech but presented differently…
Secondly design time testing is soooooooooooooo important. Be able to test the transformation at design time instead of running it and hitting and hoping…this keeps n8n support lean as transforming data is a huge drag on support calls, secondly introduces self learning at a different level. I cant emphases this point more.
And the bad thing i feel is i know for example, map function i know the logic how it behaves and the theory but i can’t write the damn thing correctly to run successfully when i want to manipulate some data. Sorry to bother you and come here asking for solutions.
@Mulen have you seen the (newish) Item Lists node? Granted it doesn’t cover all cases but helps with various transform utilities related to arrays <> n8n items. By using multiple Item Lists nodes, it’s possible to achieve things like “Output a single random item from a bigger input list”.