How can I get the column in BASH? - bash

Question

I have a large .xml file like that:
c1="a1" c2="b1" c3="cccc1"
c1="aa2" c2="bbbb2" c3="cc2"
c1="aaaaaa3" c2="bb3" c3="cc3"
...
I need the result like the following:
a1 b1 cccc1
aa2 bbbb2 cc2
aaaaaa3 bb3 cc3
...
How can I get the column in BASH?
I have the following method in PL/SQL,but it's very inconvenient:
SELECT C1,
TRIM(BOTH '"' FROM REGEXP_SUBSTR(C1, '"[^"]+"', 1, 1)) c1,
TRIM(BOTH '"' FROM REGEXP_SUBSTR(C1, '"[^"]+"', 1, 2)) c2,
TRIM(BOTH '"' FROM REGEXP_SUBSTR(C1, '"[^"]+"', 1, 3)) c3
FROM TEST;

Answer 1

Use cut:
cut -d'"' -f2,4,6 --output-delimiter=" " test.txt
Or you can use sed if the number of columns is not known:
sed 's/[a-z][a-z0-9]\+="\([^"]\+\)"/\1/g' < test.txt
Explanation:
[a-z][a-z0-9]\+ - matches a string starting with a alpha char followed by any number of alphanumeric chars
"\([^"]\+\)" - captures any string inside the quotes
\1 - represents the captured string that in this case is used to replace the entire match
A perl approach (based on the awk answer by #A-Ray)
perl -F'"' -ane 'print join(" ",#F[ map { 2 * $_ + 1} (0 .. $#F) ]),"\n";' < test.txt
Explanation:
-F'"' set input separator to "
-a turn autosplit on - this results in #F being filed with content of fields in the input
-n iterate through all lines but don't print them by default
-e execute code following
map { 2 * $_ + 1} (0 .. $#F) generates a list of indexes (1,3,5 ...)
#F[map { 2 * $_ + 1} (0 .. $#F)] takes a slice from the array, selecting only odd fields
join - joins the slice with spaces
NOTE: I would not use this approach without a good reason, the first two are easier.
Some benchmarking (on a Raspberry Pi, with a 60000 lines input file and output thrown away to /dev/null)
cut - 0m0.135s no surprise there
sed - 0m5.864s
perl - 0m8.218s - I guess regenerating the index list every line isn't that fast (with a hard coded slice list it goes to half, but that would defeat the purpose)
the read based solution - 0m52.027s

Answer 2

You can also look at the built-in substring replacement/removal bash offers. Either in a short script or One-Liner:
#!/bin/bash
while read -r line; do
new=${line//c[0-9]=/} ## remove 'cX=', where X is '0-9'
new=${new//\"/} ## remove all '"' (double-quotes)
echo "$new"
done <"$1"
exit 0
Input
$ cat dat/stuff.xml
c1="a1" c2="b1" c3="cccc1"
c1="aa2" c2="bbbb2" c3="cc2"
c1="aaaaaa3" c2="bb3" c3="cc3"
Output
$ bash parsexmlcx.sh dat/stuff.xml
a1 b1 cccc1
aa2 bbbb2 cc2
aaaaaa3 bb3 cc3
As a One-Liner
while read -r line; do new=${line//c[0-9]=/}; new=${new//\"/}; echo "$new"; done <dat/stuff.xml

Answer 3

awk -F '"' '{ for(i=2; i<=NF; i+=2) { printf $i" " } print "" }'
Explanation
-F '"' makes Awk treat quotation marks (") as field delimiters. For example, Awk will split the line...
c1="a1" c2="b1" c3="cccc1"
...into fields numbered as...
1: 'c1='
2: 'a1'
3: ' c2='
4: 'b1'
5: ' c3='
6: 'cccc1'
7: ''
for(i=2; i<=NF; i+=2) { printf $i" " } starts at field 2, prints the value of the field, skips a field, and continues. In this case, fields 2, 4, and 6 will be printed.
print outputs a string following by a newline. printf also outputs a string, but doesn't append a newline. Therefore...
printf $i" "
...outputs the value of field $i followed by a space.
print ""
...simply outputs a newline.