If you like PowerShell, try using jq in your Unix shell pipelines. Ingesting data into json is still a bit finicky, but once you know the patterns you need to use with cut and a csv-to-json tool things become much faster.
Hadn't thought of chaining that with a csv-to-json workflow, that's a neat idea -- but if that's the easiest way out, that's clearly a symptom of the problem I'm describing, isn't it?
Quite possible. I will say that I'd much prefer structured text that I can massage to my liking over an actual object model. Having things print and read JSON streams is just as good as having cmdlets push and pull .net object, except that we can write our programs in languages that aren't .net-aware. And it makes it way easier for the system to interact with systems that don't speak interchange data format, including things like filesystems and websites.
One of the things I've been trying to figure out for a while is a way to build a trivial adapter for the csv-to-json | jq pipelines I've been building, something sufficiently easy to use that I could bake it into a |-equivalent extension to something like zsh.
