read/write bits

[MonteShaffer]

Hello, Thank you for your efforts. It is very fast. I tried it on the PRIME sieve where the idx are the prime numbers.

For n=1000 primes, I overshoot using gn = ceiling( n * log(n) + n * log(log(n)) )

Here is the format of the output... I truncate the function result to exactly n = 1000, but keep the full calculation; in this case it looks like there are 1102 primes with gn as upper bound.

bitwhich: 1102/8841 occupying only 1102 int32 in 1 representation
    1     2     3     4     5     6     7     8        8834  8835  8836  8837  8838  8839  8840  8841 
FALSE  TRUE  TRUE FALSE  TRUE FALSE  TRUE FALSE    .. FALSE FALSE FALSE  TRUE FALSE  TRUE FALSE FALSE 

My question is about storage. From my understanding, if I just save/load or saveRDS/readRDS on the 8841 elements, the filesize should only be about 1.1KB ... Trying both methods (save/saveRDS), I get 2.6KB ... it throws an error when I try writeBin.

I didn't see any SPECIAL read/write functions in the DOCS. Maybe I missed it.

Any ideas or suggestions?

Again thank you. I wonder if DNA researchers could use a two-bit version of this for "AGCT"? Very powerful.

[MichaelChirico]

I'm afraid there's some context lost here. But I think the overall point is we'd like an equivalent of readBin()/writeBin() for I/O of bit objects.

Could you give a reprex on your saveRDS() experiment? I don't find the same result, actually I get a lot of compression:

x=bit(1e8)
x[sample(1e8, 1e6)] <- TRUE

saveRDS(x, tmp<-tempfile())

file.size(tmp)
# [1] 1601882
as.double(object.size(x))/file.size(tmp)
# [1] 7.804265