On 12/11/2021 17:05, Rodolfo Aramayo wrote:
Dear Coreutils Maintainers,
First, thank you for your work. I use coreutils daily both for my research
and teaching. It is a great set of tools.
Second, I recently needed to extract Coding Sequences information from a
GenBank file. GenBank files are used in Computational
Genomics/Bioinformatics extensively. I used csplit, and it works like a
charm.
The command I used is:
csplit -sz -n 5 --prefix=02_ 01_00001
/[[:space:]][[:space:]][[:space:]][[:space:]][[:space:]]CDS[[:space:]][[:space:]][[:space:]][[:space:]][[:space:]][[:space:]][[:space:]][[:space:]][[:space:]][[:space:]][[:space:]][[:space:]][[:space:]]/
{*};
I was unable to declare: "[[:space:]]\+" as I expected for POSIX aware code.
My question is: Is csplit POSIX compatible? and if it is not, can we make
it POSIX compatible?
Well POSIX defines BRE and ERE, with csplit supporting the former.
From the code we have:
re_syntax_options =
RE_SYNTAX_POSIX_BASIC & ~RE_CONTEXT_INVALID_DUP & ~RE_NO_EMPTY_RANGES;
Generally one can replace '+' functionality from ERE, with '\{1,\}' in BRE.
So you'd be using something like:
[[:space:]]\{1,\}CDS[[:space:]]\{1,\}
We might add an option to use ERE, though there isn't a big need
for that I think for csplit use cases.
cheers,
Pádraig
