sed convert double quoted csv comma delimited to non quoted pipe delimited

The input file 'input.file' is:

"col one",,,"col, two",,"col, three"
,"col one",,"col, two",,"col, three"
,,"col one","col, two",,"col, three"

The desired output file is:

col one|||col, two||col, three
|col one||col, two||col, three
||col one|col, two||col, three

This is what is in place so far:

sed -r 's/"([^"])/\1/g; s/,/|/g'  ./input.file

The objective is the first substitution

s/"([^"])/\1/g

would parse arbitrary fields delineated by " and copy them to the output and the second substitution

s/,/|/g

would replace the ',' not included in double quoted fields with '|'.

Solution

$ cat ip.txt 
"col one",,,"col, two",,"col, three"
,"col one",,"col, two",,"col, three"
,,"col one","col, two",,"col, three"

$ perl -pe 's/"[^"]+"(*SKIP)(*F)|,/|/g; s/"//g' ip.txt
col one|||col, two||col, three
|col one||col, two||col, three
||col one|col, two||col, three

"[^"]+"(*SKIP)(*F) skip the pattern "[^"]+" and look for any other alternate match provided
- (*F) is short-form for (*FAIL) and (?!) can also be used instead
|, alternate pattern to match ,
|/g replace all , with |
s/"//g then delete all "

Further Reading: