fread with gunzip: What's the more memory efficient way?

Question

If I have a large data file which is zipped with gzip, say dat.gz, what is more memory efficient?

mydat <- fread("gunzip -c dat.gz")

or, first unzip/uncompress the file to say, dat, and then do

mydat <- fread("dat")

I'm concerned with memory rather than speed, to prevent R from crashing.

Answer 1

I wrote a 5000x5000 matrix to temp.csv and profiled the memory usage of the two approaches using profvis:

profvis({system("gunzip -c temp.csv.gz > temp.csv"); mat <- fread("temp.csv")})

Memory usage: 190.9 MB

profvis({fread("gunzip -c temp.csv.gz")})

Memory usage: 190.8 MB

I ran it several times, and the memory usage fluctuated between 190-191 for both commands. So I conclude that the memory usage is the same.