While trying to script safely handling filenames while handling newlines safely, I came across a difficult test case.
Given the input
a.b.c
.d.staging
where this input represents a single filename, I want to strip the .staging suffix. I would normally use something akin to | rev | cut -d. -f2- | rev for this, but this fails:
echo -ne "a.b.c\\n.d.staging" | rev | cut -d. -f2- | rev
yields
a.b
.d
In addition to having lost the c component in addition to the staging suffix, there's also a lone newline at the end there Markdown is hiding.
The best solution I've come up with so far is to use sed -e ':a' -e 'N' -e '$!ba' -e 's/\(.*\)\..*/\1/', which appears to work:
echo -ne "a.b.c\\n.d.staging" | sed -e ':a' -e 'N' -e '$!ba' -e 's/\(.*\)\..*/\1/'
yields
a.b.c
.d
which is the correct output.
This seems an inelegant solution, as it's hammering sed into handling newlines, which is something sed isn't great at doing.
Is there a more elegant solution? Ideally a POSIX-compatible one.
If you have the name in a variable, the newline is not an issue.
$ fname=$'a.b.c\n.d.staging'
$ echo "$fname"
a.b.c
.d.staging
$ echo "${fname%.*}"
a.b.c
.d
$