How to parse this string into kv in awk in a simple way - awk

Question

Now I have a str in awk like this:
str = "a='abc',b=1,c='http://xxxx,http://yyyy,http://zzz'"
How can I parse it to get this result:
(a abc)(b 1)(c http://xxxx,http://yyyy,http://zzz)
By now I still implement it in such an ugly way:
result = ""
while (match(str, /[^=]*=('[^']*'|[^,]*),/) != 0) {
subs = substr(str, RSTART, RLENGTH)
str = substr(str, RSTART + RLENGTH, length(str) - RSTART - RLENGTH + 1)
split(subs, vec, "=")
gsub(/'/, "", vec[1])
gsub(/'/, "", vec[2])
if (substr(vec[2], length(vec[2]), 1) == ",") {
vec[2] = substr(vec[2], 0, length(vec[2]) - 1)
}
result = result"("vec[1]" "vec[2]")"
}
I wonder if there exist some more elegant way.

Answer 1

Using awk
The trick here is that we need to treat quoted commas differently from unquoted commas. That can be done as follows:
$ echo "$str" | awk -F"'" -v OFS="" '{for (i=1;i<=NF;i+=2) gsub(",", ")(", $i)} {gsub("=", " "); print "("$0")"}'
(a abc)(b 1)(c http://xxxx,http://yyyy,http://zzz)
How it works
-F"'" -v OFS=""
This sets the input field separator to a single quote and the output separator to an empty string.
{for (i=1;i<=NF;i+=2) gsub(",", ")(", $i)}
This replaces unquoted commas (odd fields) with )(.
Even numbered fields represent the quoted strings and they are left unchanged here.
gsub("=", " ")
This replaces equal signs with spaces.
print "("$0")"
This adds parens to the beginning and end and prints the line.
Using sed
$ echo "$str" | sed -r ":a; s/^(([^']*'[^']*')*[^']*'[^,']*),/\1\n/; ta; s/,/)(/g; s/^/(/; s/$/)/; s/\n/,/g; s/'//g; s/=/ /g"
(a abc)(b 1)(c http://xxxx,http://yyyy,http://zzz)
How it works
First, remember that sed processes input line-by-line. That means that, unless we put one in it, no line in sed's pattern space will contain a newline character.
This command works by replacing all quoted commas with newline characters. It then adds ( to the beginning of the line, ) to the end of the line, and replaces the remaining commas with )(. The newline characters are changed back to commas. Next the single-quotes are removed. Finally, the = signs are then replaced with spaces and we are done.
We can tell whether a comma is quoted or unquoted by whether is it is preceded by an odd or an even number of single-quotes.
In more detail:
sed -r
This starts sed with extended regular expressions.
:a; s/^(([^']*'[^']*')*[^']*'[^,']*),/\1\n/; ta
This converts all quoted commas into newline characters. The regex ^(([^']*'[^']*')*[^']*'[^,']*) matches, starting at the beginning of the line, any odd numbers of single-quotes and the text surrounding them up to the first comma afterward. The substitution command s/^(([^']*'[^']*')*[^']*'[^']*),/\1\n/ consequently replaces the first quoted comma found with a newline, \n.
:a is a label. ta is a test: it branches back to label a if a substitution was made. Thus, as many substitutions are made as needed to replace all the quoted commas with newline characters.
s/,/)(/g; s/^/(/; s/$/)/
These three substitution commands puts parens everywhere that we want one.
s/\n/,/g
Now that we have parens where we need them, this converts the newline characters that we added back to commas.
s/'//g
This removes all the single quotes.
s/=/ /g
This replaces the equal signs with spaces.