Concatenating lines between braces

Question

I need to concatenate a variable number of lines between braces, the braces are nested so only braces that begin with a specific prefix pattern to the brace need to be concatenated:

header {

category1 (a) {
field1 : value1 ;
field2 : value2 ;
...
fieldn : valuen ;
}

new cat1 (b) {
newfield1 : newvalue1 ;
newfield2 : newvalue2 ;
....
newfieldn : newvaluen ;
}

...

}

The category and field names are variable, the trailing ";" may or may not have a preceding space. Any line may have spaces or tabs in front of it or between words/delimiters/etc.

The output should look like:

header {
category1 (a) { field1 : value1 ; field2 : value2 ; ... fieldn : valuen ; }
new cat1 (b) { newfield1 : newvalue1 ; newfield2 : newvalue2 ; .... newfieldn : newvaluen ; }
}

I'm using CentOS, so sed/awk/perl are available.

Thanks!

Answer 1

Try

$ awk '
/^ *$/          {next
                }
/cat.*{/        {while (! /}/)  {getline X
                                 $0 = $0 X
                                }
                }
1
' file
header {
category1 (a) {field1 : value1 ;field2 : value2 ;...fieldn : valuen ;}
new cat1 (b) {newfield1 : newvalue1 ;newfield2 : newvalue2 ;....newfieldn : newvaluen ;}
...
}

Answer 2

With GNU awk for multi-char RS and RT this will do the bit in the middle:

$ awk -v RS='[^\n]+{[^{}]+}' '{$0=RT; $1=$1} RT' file
category1 (a) { field1 : value1 ; field2 : value2 ; ... fieldn : valuen ; }
new cat1 (b) { newfield1 : newvalue1 ; newfield2 : newvalue2 ; .... newfieldn : newvaluen ; }

Printing the first and last lines is as simple as head -1 and tail -1:

$ head -1 file; gawk -v RS='[^\n]+{[^{}]+}' '{$0=RT; $1=$1} RT' file; tail -1 file
header {
category1 (a) { field1 : value1 ; field2 : value2 ; ... fieldn : valuen ; }
new cat1 (b) { newfield1 : newvalue1 ; newfield2 : newvalue2 ; .... newfieldn : newvaluen ; }
}

or figure out how to do it in that awk script too if you care, it's not particularly hard jut requires a bit more thought than I'm up for right now!

Answer 3

This will do mostly what you want, though it won't add spaces between values and semicolons if they're left out...

Assuming your input is in file.txt:

tr "\n" " " < file.txt | tr "\t" " " | tr -s " " | sed "s/}/}\n/g" | sed "s/header {/header {\n/" | sed "s/^ //g"

1) Replace all the existing newlines with spaces

tr "\n" " "

2) Replace all tabs with spaces

tr "\t" " "

3) Replace all runs of spaces with a single space

tr -s " "

4) Replace all close-brackets with close-bracket followed by newline

sed "s/}/}\n/g"

5) Add a newline after the initial "header {"

sed "s/header {/header {\n/"

6) Remove the leading space on subsequent lines

sed "s/^ //g"