The same is true in many other data science branches (such as physics). Perl was the perfect glue language in the 90s but then got more and more replaced by pyt

The same is true in many other data science branches (such as physics). Perl was the perfect glue language in the 90s but then got more and more replaced by python. I think one reason for that is Numpy/Pylab, which makes python a perfect replacement for proprietary Matlab codes. All at once, scientists can write most of their codes in a single language/software ecosystem which also feels relatively close to mathematical notation, at least you can write equations similarly readable as in (modern) Fortran. The perl sigils really got out of fashion.