regex - How to remove all except the first 3 and last of a specific character with sed -

i've looked on place can't find answer. i've used sed before i'm familiar syntax - 1 has me stumped.

i want remove except first 3 instances , last instance of specific character. here specific example:

input.csv:

"first", "some text "quote" blaw blaw", 1 "second", "some more text "another quote" blaw blaw", 3 

i want remove quotes (") except first 3 , last 1 looks this:

output.csv:

"first", "some text quote blaw blaw", 1 "second", "some more text quote blaw blaw", 3 

any pointers? thanks.

$ sed -r ':a; s/([^"]*"[^"]*"[^"]*")([^"]*)"([^"]*")/\1\2\3/; ta' input.csv "first", "some text quote blaw blaw", 1 "second", "some more text quote blaw blaw", 3 

how works

the code works looking first 5 quotes. removes fourth. process repeated looping until there 4 quotes left.

• :a

  this defines label a.

• s/([^"]*"[^"]*"[^"]*")([^"]*)"([^"]*")/\1\2\3/

  this looks the first 3 quotes , text that precedes them group 1. looks next set of non-quote characters group 2. looks following double quote. looks non-quote characters followed fifth quote group 3. replaces 3 groups, omitting fourth quote.

  let's break down more explicitly:

  • ([^"]*"[^"]*"[^"]*")

    this looks the first 3 quotes , text that precedes them. saved group 1.

  • ([^"]*)

    this looks next set of non-quote characters. saved group 2.

  • "

    this matches fourth quote on line.

  • ([^"]*")

    this matches next group of non-quote characters followed fifth quote on line. saved group 3.

  the replacement text \1\2\3 has effect of removing fourth quote of 5 quotes found.

• ta

  if substitution made, loops label a. if not, done line.

bsd or mac osx

try:

sed -e -e ':a' -e 's/([^"]*"[^"]*"[^"]*")([^"]*)"([^"]*")/\1\2\3/' -e 'ta' input.csv