I have been using AWK at work and found it cumbersome for my use-case (run-once analysis/manipulation of 200GB csv datasets). I found out about Miller[1] a year or so back and have been using that instead. I don't know how it stacks up in terms of performance, but for my money, named arguments and one-shot statistics is all I need.
For example analysing the number of people per-year over multiple differently formatted files is as easy as `mlr uniq -f pid,year then count -g year`. It has filtering, very extensive manipulation capabilities and a nice documentation.
John has also reacted very fast on feature requests I had (so fast that I've not yet implemented using them).