Very large values in numerical columns can cause a crash due to overflow

[Schaechtle]

This line can crash with an overflow error for very large values of x becasue Clojure doesn't automatically upgrade Ints to BigInts. This happend with the Gapminder dataset and is reproducible in the Clojure repl:

$ clojure
Clojure 1.11.1
user=> (def a 12756500000000)
#'user/a
user=> (* a a)
Execution error (ArithmeticException) at java.lang.Math/multiplyExact (Math.java:892).
long overflow

The value 12756500000000 is the problematic cell value in the Gapminder dataset. It's unclear whether this problem is common.

[pratmeha]

For what it's worth, I ran into this when working with a dataset of executable file characteristics, which contained values like memory addresses or Unix timestamps. I expect some of these, especially the timestamps, could be removed from the data without harm but sometimes that's hard to tell a priori.

If this is troublesome to fix, I wonder if it could be worked around by approximating large integers with scientific notation? I'm not sure if that's supported, though.

[sritchie]

@Schaechtle you can fix this by using *', with an extra apostrophe at the end; I'm not sure where in the code the bare * was hit, but that would be one fix.