Add set key to data.table operations.

[ricardonovaes]

setkey make joins extremely faster in data.tables, the codes over join benchmark are not setting the keys properly and can affect the main results.

It is also important in other kinds of data manipulation such as deduce. for instance:
setkey(DT, key)
unique(DT, by = 'key')

is very much faster than
unique(DT, by 'key')

This can go from 15 minutes to seconds for 100GB+ datasets

Joins work the same way:

setkey(DTA, key)
setkey(DTB, key)

DTA[DTB, on = .(key)]

I hope it can make the benchmar better!!

[Tmonster]

Hi,

Thank you for the suggestion. I am mainly focused on maintaining the benchmark, and not on improving every system. I am happy to review PRs however and kick off some tests if needed