Unix tool to reduce file chunks to a string (--delim=, --pre=[ --suf=] < file)

Question

I am looking for a program, which concatenates strings with a char, prepends one char and appends another. I think I have used the wrong keywords for my search, but I was not able to find the perfect unix tool for that issue.

Suppose I have a file (note the starting empty lines):

file in.txt:

{
"some": "json",
"with_different": "intendation",
  "which": [],
  "has": 2
}

{
 "json":"objects"
}

and generate out.txt

[
{
"some": "json",
"with_different": "intendation",
  "which": [],
  "has": 2
}
,
{
 "json":"objects"
}
]

Basically, I want a JSON-array from that, meaning:

1. get rid of first empty lines ( uniq | tail --lines=+2),
2. replace empty lines with comma (sed -e 's/^$/,g/') and
3. prepend/append it with [ and ] (awk 'BEGIN {print "["} {print $1} END {print "]"}).

uniq <in.txt | tail --lines=+2 | sed -e 's/^$/,/g' | awk 'BEGIN {print "["} {print $1} END {print "]"}' is giving me what I want, but I sure think, that this is not elegant.

I have found paste, xargs, join, but they do not help me. Also I know about the OFS variable in awk, which may replace the sed part, but I don't know how to convince awk to treat all 'non-empty' lines as $1 (probably using IFS, but IFS='^$' is surely not working.) And then we still have the other boilerplate around it.

I am hoping that someone can point me to magic-program like magic -d"," -s"[" -e"]" <in, provided I have cleaned the empty lines above, or the objects are one-liners

file in:

{"some":"json",  "which":[],  "has": 2}

{ "json":"objects"}

to file out:

[
{"some":"json",  "which":[],  "has": 2}
,
{ "json":"objects"}
]

Other example would be echo "a b c" | magic -d',' -s'[' -e']' returns [a,b,c].

Or, to not only give JSON examples: echo "my new component" | magic -d'-' -s'<' -e'>' returns <my-new-component>.

Notes:

• jq -s . would work for this json-problem (cf. How to combine the sequence of objects in jq into one object?) but if the start/end/delim chars are something else it wouldn't.
• I am fine with line breaks being removed.
• I would really like to have a shorter one-liner than my own attempt

Answer 1

$ awk -v RS= 'BEGIN{sep="[\n"} {printf "%s%s", sep, $0; sep="\n,\n"} END{print "\n]"}' in.txt
[
{
"some": "json",
"with_different": "intendation",
  "which": [],
  "has": 2
}
,
{
 "json":"objects"
}
]

.

$ awk -v RS= 'BEGIN{sep="[\n"} {printf "%s%s", sep, $0; sep="\n,\n"} END{print "\n]"}' in
[
{"some":"json",  "which":[],  "has": 2}
,
{ "json":"objects"}
]

Answer 2

Using GNU sed:

sed '/./,$!d; s/^/[\n/; :a; n; s/^$/,/; $s/$/\n]/; ba' in.txt

with the assumption that there is no trailing blank line in the input (leading ones are discarded).

Alternatively:

sed -n '/./{s/^/[\n/; :a; p; n; s/^$/,/; $s/$/\n]/; ba}' in.txt