Delete multiline blocks with sed non-greedy

Question

How can one remove code blocks starting with /** START */ and ending with /** END */ using sed non-greedily considering that code blocks are multiline and empty lines may occur within START-END blocks?

START marker single line comment

SOLUTION

INPUT:

class MyClass {
    keepField;
    /** START */
    deleteField;
    /** END */

    construct() {
        /** START */
        this.deleteField = 'delete';
        /** END */
        this.keepField = 'keep';
        /** START */
        this.deleteFunc();
        /** END */
    }
    
    /** START */
    deleteFunc() {
        this.keepField = 'delete';

        if (true) {
            console.debug('Line before if statement is empty.');
        }
    } /** END */
}

OUTPUT:

class MyClass {
    keepField;

    construct() {
        this.keepField = 'keep';
    }
    
}

I have tried the following sed '/./{H;$!d} ; x ; s/START.*END//' MyClass.js as described in the sed manual > Multiline techniques section.

However, the above command is greedy in blocks when there are no empty lines and multiple START-END blocks (like in the constructor) and also empty lines of code are not taken into account inside START-END blocks (like in the deleteFunc function).

Any idea how the above can be solved with sed or any other command line tool such as awk?

START marker is block comment

SOLUTION

INPUT:

class MyClass {
    /**
     * same code as above only this time the START block is 
     * multiline like below.
     */

    /**
     * START
     */
    deleteFunc() {
        this.keepField = 'delete';

        if (true) {
            console.debug('Line before if statement is empty.');
        }
    } /** END */
}

OUTPUT should also be:

class MyClass {
    keepField;

    construct() {
        this.keepField = 'keep';
    }

}

Answer 1

sed is an excellent tool for doing simple s/old/new/ operations. For anything else, just use awk for clarity, efficiency, robustness, portability, maintainability, etc. For example, using any POSIX awk:

$ cat tst.awk
{ rec = rec $0 ORS }
END {
    while ( match(rec,/\/\*\*[[:space:]*]*END[[:space:]*]*\*\//) ) {
        toEnd = substr(rec,1,RSTART+RLENGTH-1)
        sub(/(\n[[:blank:]]*)?\/\*\*[[:space:]*]*START[[:space:]*]*\*\/.*/,"",toEnd)
        printf "%s", toEnd
        rec = substr(rec,RSTART+RLENGTH)
    }
    printf "%s", rec
}

$ awk -f tst.awk file
class MyClass {
    keepField;

    construct() {
        this.keepField = 'keep';
    }

}

class MyClass {
    /**
     * same code as above only this time the START block is
     * multiline like below.
     */

}

If you don't have a POSIX awk then change every [:space:] to \t\n and [:blank:] to \t (first char of each string is a literal blank char) and then it'll work in any awk.

The above was run on this input file:

$ cat file
class MyClass {
    keepField;
    /** START */
    deleteField;
    /** END */

    construct() {
        /** START */
        this.deleteField = 'delete';
        /** END */
        this.keepField = 'keep';
        /** START */
        this.deleteFunc();
        /** END */
    }

    /** START */
    deleteFunc() {
        this.keepField = 'delete';

        if (true) {
            console.debug('Line before if statement is empty.');
        }
    } /** END */
}

class MyClass {
    /**
     * same code as above only this time the START block is
     * multiline like below.
     */

    /**
     * START
     */
    deleteFunc() {
        this.keepField = 'delete';

        if (true) {
            console.debug('Line before if statement is empty.');
        }
    } /** END */
}

but consider also this pathological case where the whole input is on a single line:

$ cat file
class MyClass { keepField; /** START */ deleteField; /** END */ construct() { /** START */ this.deleteField = 'delete'; /** END */ this.keepField = 'keep'; /** START */ this.deleteFunc(); /** END */ } /** START */ deleteFunc() { this.keepField = 'delete'; if (true) { console.debug('Line before if statement is empty.'); } } /** END */ }

and note that the above script handles it correctly (as it also would many other unstated cases I could imagine except where your start/end strings could be inside literal strings or themselves inside comments - you can't handle such cases with pattern matching as we're doing):

$ awk -f tst.awk file
class MyClass { keepField;  construct() {  this.keepField = 'keep';  }  }

Answer 2

Using GNU sed

$ sed -Ez 's~ +/?\*+ START( \*)?([^*]*\*+)([^\n]*\n[^*]*\*+)? END[^\n]*\n~~g' input_file

class MyClass {
    keepField;

    construct() {
        this.keepField = 'keep';
    }
    
}

class MyClass {
    /**
     * same code as above only this time the START block is 
     * multiline like below.
     */

    /**
}