Single-pass file reading

[Jelinek-J]

Consider a simple script test.R:
#! /usr/bin/env Rscript
args = commandArgs(trailingOnly=TRUE)
Biostrings::readDNAStringSet(args[1])

and a simple fasta file test.fa:
>test
ACGT


If I run it directly $ ./test.R test.fa, it works correctly and prints this:
DNAStringSet object of length 1:
  width seq names
[1] 4 ACGT test


But if I use process substitution $ ./test.R <( cat test.fa ) (I need to preprocess the file), it only prints:
DNAStringSet object of length 0
Warning messages:
1: In file(fp) : using 'raw = TRUE' because '/dev/fd/63' is a fifo or pipe
2: In file(fp) : using 'raw = TRUE' because '/dev/fd/63' is a fifo or pipe

resp. throws an error if the fasta file is at least 4096 bytes long:
Error in .Call2("fasta_index", filexp_list, nrec, skip, seek.first.rec, :
reading FASTA file /dev/fd/63: ">" expected at beginning of line 1
Calls: -> .read_XStringSet -> fasta.index -> .Call2
In addition: Warning messages:
1: In file(fp) : using 'raw = TRUE' because '/dev/fd/63' is a fifo or pipe
2: In file(fp) : using 'raw = TRUE' because '/dev/fd/63' is a fifo or pipe
Execution halted


From the warnings I gather that Biostrings loads files in two passes, which is not possible with process substitution. Is there any combination of arguments that will force loading files in single-pass? If not, would be possible to implement such an option?