remove last occurrence of ',' in a file - but not the whole line-CodePudding

If I have a file containing this:

[
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]},
]

I want to remove that very last comma so I get:

[
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]}    <----- no comma
]

I can't find a simple way to do this so far - most solutions remove the whole line or do line searching which does not match what I am trying. I feel like it would be a sed or awk command.

CodePudding user response：

rev file | sed -z 's/,//' | rev

Reverse the file, remove the first comma in zero separated stream, so the first comma in the whole file, and than reverse it again.

CodePudding user response：

Using ed, which, unlike sed, is intended for editing files and thus allows for things like searching backwards in a file:

printf "%s\n" '?,$?s/,$//' w | ed -s file

will find the last line ending with a comma and remove that comma, then save the file.

I agree with @Barmar in the comments though that a better solution is fixing whatever is generating your file to not include that trailing comma in the first place.

CodePudding user response：

With GNU sed:

sed -z 's/},\n]/}\n]/' file

Output:

[
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]}
]

CodePudding user response：

With shell parameter expansion, to blindly remove the last comma:

contents=$(< file)
fixed=${contents%,*}${contents##*,}
echo "$fixed"

Using text tools: reverse the file, remove the ending comma on the first line that has one, then re-reverse the file

tac file | awk '!p && /,$/ {sub(/,$/, ""); p = 1} 1' | tac

CodePudding user response：

A little different take that should allow for the desired output even if the ',' is not the last character of the line.

Find the last line containing a comma and capture the line number:

line_no=$(grep -n ',' t.dat | tail -n 1 | cut -f1 -d:)

Then remove the last comma from that specific line no matter where it occurs:

awk -v li="${line_no}" 'BEGIN{ FS=OFS="," } NR==li{ for(i=1; i<NF-1; i  ). printf "%s", $i OFS; printf "%s\n", $i "" $NF; next }1' t.dat

This can be combined as:

awk -v li="$(grep -n ',' t.dat | tail -n 1 | cut -f1 -d:)" 'BEGIN{ FS=OFS="," } NR==li{ for(i=1; i<NF-1; i  ) printf "%s", $i OFS; printf "%s\n", $i "" $NF; next
}1' t.dat

t.dat contents:

[
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]},
]

Process t.dat:

$ awk -v li="$(grep -n ',' t.dat | tail -n 1 | cut -f1 -d:)" 'BEGIN{ FS=OFS="," }
      NR==li{ for(i=1; i<NF-1; i  ) printf "%s", $i OFS;
            printf "%s\n", $i "" $NF; next
}1' t.dat

Output:

[
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]}
]

CodePudding user response：

if the entire file ends with that chunk, then just do

gawk FS  RS='[]][}],\n[]]\n' ORS=']}\n]\n'
nawk $$  RS='[]][}],\n[]]\n' ORS=']}\n]\n'
mawk //_ RS='[]][}],\n[]]\n' ORS=']}\n]\n'

[
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]}
]

CodePudding user response：

 awk '/,/ &&   count==1{gsub(/,$/,"")}1' <(tac file)|tac

CodePudding user response：

Using sed

$ sed -z 's/\(.*\),/\1/' input_file

$ sed 'N;s/,\(\n]\)/\1/' input_file

Output

[
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]},
{[ ...,..., ]}
]