Slow first run loading a tiny csv file

[xgdgsc]

When first run CSV.read to load a tiny csv file (less than 20 lines) into a dataframe, it is unbelievably slow. And allocations seems too much. Afterwards it only takes 0.0001 second to load other similar files.

[nalimilan]

Unfortunately that's expected due to compilation of functions. Hopefully at some point it will be possible to do this during package precompilation.

[quinnj]

Yes, this is more a function of the language itself rather than something specific to CSV.jl. There are probably some very small improvements we could make by rearranging functions a bit and doing some precompile, but honestly, those efforts have never paid off very much in my experience. I'll take a quick look though and see if there is any low-hanging fruit.