Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Tools like

https://www.alteryx.com/

are really limited in my view because they are based on a relational model that requires joins to do things that are easily expressed without joins in the object-relational (basically JSON) model.

"Boxes and lines" data pipelines that involve joins are a bear to maintain because of the structural instability: what looks like a little change to management can turn a very simple pipeline that goes in one direction to a pipeline that splits into four streams that have to get reassembled at the end.

If you were passing JSON documents between the nodes you can keep the stuff that would have been split out and then joined inside the document and the pipelines are a lot easier to maintain.

The people who develop those tools, however, are in love with the relational model because it is SIMDable, not branchy, easy to implement at hyper-speed so there is going to be an opportunity for people to make one that gets much better customer satisfaction.

To do that though somebody has to ask the question of "what algebra does this thing work over?"



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: