I have a csv file with data similar to this:

"1b579a5e-9701-40eb-bd36-2bc65169da99","week14_Friday-019","6907eaad-1aff-4d26-9088-ba20374b67c0","2181-019","f20af5bb-c716-42e0-9b9d-cbolf5bfecea","15-BIO-2001","COLLEGE Bio 1","d39330be-df56-4365-8fb4-37e68d040c52","Which engine has the smaller efficiency?","{choices:[a","b","c","d],"type:MultipleChoice}","{solution:[0],"selectAll:false,{"selectMultiple:false",}"type:MultipleChoice}","2016-04-25 00:30:19.000","1922ac5a-6ff6-4ea4-9078-6df4d85d294f","{solution:[0],"type:MultipleChoice}","1","1116911f-8ee5-45c3-b173-a6be681bb15a","FakeLastName","FakeFirstName","[email protected]","Student"

I want to remove the double-quotes ", but only if they are inside of the curly braces {}, preferably with sed or awk. Desired output is:

"1b579a5e-9701-40eb-bd36-2bc65169da99","week14_Friday-019","6907eaad-1aff-4d26-9088-ba20374b67c0","2181-019","f20af5bb-c716-42e0-9b9d-cbolf5bfecea","15-BIO-2001","COLLEGE Bio 1","d39330be-df56-4365-8fb4-37e68d040c52","Which engine has the smaller efficiency?","{choices:[a,b,c,d],type:MultipleChoice}","{solution:[0],selectAll:false,{selectMultiple:false,}type:MultipleChoice}","2016-04-25 00:30:19.000","1922ac5a-6ff6-4ea4-9078-6df4d85d294f","{solution:[0],type:MultipleChoice}","1","1116911f-8ee5-45c3-b173-a6be681bb15a","FakeLastName","FakeFirstName","[email protected]","Student"

Any help would be greatly appreciated. Thanks!

  • @Thephrastus - regexp is almost always applied by a proper stateful language such as sed, ed, ex, awk, perl, and others.
    – mikeserv
    Aug 7, 2018 at 19:37

3 Answers 3


It would be easier with sed:

sed -e :1 -e 's/\({[^}]*\)"\([^}]*}\)/\1\2/g; t1'

Or perl:

perl -pe 's{\{.*?\}}{$& =~ s/"//gr}ge'

Note that it assumes there's no nested {...}.

To handle nested {...}, you can use perl's recursive regexp capabilities:

perl -pe 's(\{(?:[^{}]++|(?0))*\})($& =~ s/"//gr)ge'

With sed, working our way outwards to escape the inner {...}s before removing the "s:

sed 's/_/_u/g
     s/\({[^{}]*\){\([^{}]*\)}/\1_<\2_>/g; t1
     s/\({[^}]*\)"\([^}]*}\)/\1\2/g; t2
     s/_</{/g; s/_>/}/g;s/_u/_/g'
  • 1
    Stéphane Chazelas, This worked great, except I do actually have nested {...} I updated the description input/desired output. Please update when you have a chance? Thanks! Aug 7, 2018 at 16:36
  • 1
    @BrockWinfrey, see edit. Aug 7, 2018 at 19:04

Try this using sed:

$ sed -r ' :L; s/(\{[^"}]*)"(([^"}]*")*)([^"}]*\})/\1\2\4/g; tL; ' file

awk for nested braces:

    awk -F{ '
            {for (i=2; i<=NF; i++)  {if (1 == n = split ($i, T, "}")) n++ 
                                     for (j=1; j<n; j++)    {RS = index ($i, T[j])
                                                             L = gsub (/"/, _, T[j])
                                                             $i = substr ($i, 1, RS-1)  T[j]   substr ($i, RS + length (T[j]) + L)
    1' OFS="{" file

Clumsy, but yields the output specified, at least for singly nested braces.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .