convert table into comma separated in text file using bash - bash

Question

I have a text file like this:
+------------------+------------+----------+
| col_name | data_type | comment |
+------------------+------------+----------+
| _id | bigint | |
| starttime | string | |
+------------------+------------+----------+
how can i get a result like this using bash
(_id bigint, starttime string )
so just the column names and type
#remove first 3 lines
sed -e '1,3d' < columnnames.txt >clean.txt
#remove first character from each line
sed 's/^.//' < clean.txt >clean.txt
#remove last character from each line
sed 's/.$//' < clean.txt >clean.txt
# remove certain characters
sed 's/[+-|]//g' < clean.txt >clean.txt
# remove last line
sed '$ d' < clean.txt >clean.txt
so this is what i have so far, if there is a better implementation let me know!

Answer 1

Something similar, using only awk:
awk -F ' *[|]' 'BEGIN {printf("(")} NR>3 && NF>1 {printf("%s%s%s", NR>4 ? "," : "", $2, $3)} END {printf(" )\n")}' columnnames.txt

Answer 2

# Set the field separator to vertical bar surrounded by any number of spaces.
# BEGIN and END blocks print the opening and closing parens
# The line between skips the header lines and any line starting with '+'
$ awk -F"[[:space:]]*[|][[[:space:]]*" '
BEGIN { printf "%s", "( "}
NR > 3 && $0 !~ /^[+]/ { printf("%s%s %s", c, $2, $3); c = ", " }
END { print " )" }' file
( _id bigint, starttime string )

Answer 3

$ awk -F'[| ]+' 'NR>3 && NF>1{v=v s $2" "$3; s=", "} END{print "("v")"}' file
(_id bigint, starttime string)

Answer 4

I would do this :
cat input.txt \
| tail -n +4 \
| awk -F'[^a-zA-Z_]+' '{ for(i=1;i<=NF;i++) { printf $i" " }}'
Its a little bit shorter.

Answer 5

Another way to implement Diego Torres Milano's solution as a stand-alone awk program:
tableconvert
#!/usr/bin/env -S awk -f
BEGIN {
FS="[[:space:]]*[|][[[:space:]]*"
printf "%s", "( "
}
{
if (FNR <= 3 || match($0, /^[+]/))
next
else {
printf("%s%s %s", c, $2, $3)
c = ", "
}
}
END {
print " )"
}
Make tableconvert an executable:
chmod +x tableconvert
Run tableconvert on intablefile.txt
./tableconvert intablefile.txt
( _id bigint, starttime string )
With added bonus that using FNR instead of NR allow the awk program to process multiple input files as arguments:
./tableconvert infille1.txt infile2.txt infile3.txt ...

Answer 6

A variation on the other answers using awk with the field-separator being the '|' with optional spaces on either side as GNU awk allows, then taking fields 2 and 3 as the fields wanted in each record, and formatting the output as described in the question with the closing " )" provided in the END rule:
$ awk -F' *\\| *' '
NR>3 && $1~/^[+]/{exit} # exit condition first line w/^+
NR==4{$1=$1; printf "(%s %s", $2,$3} # 1st data record is 4
NR>4{$1=$1; printf ", %s %s", $2,$3} # process all remainng records
END{print " )"} # output closing " )"
' table
(_id bigint, starttime string )
(note: if you don't want the two-spaces before the closing ")", just remove them from the print in the END rule)
Rather than using a BEGIN the first record of interest (4) is used to provide the opening "(". Look things over and let me know if you have questions.