cps

A CPS (Continuation Passing Style) library to ease the event-driven/asynchronized coding style in node.js. There seems to be enough node.js libs (e.g. async) doing the same thing, why yet another? This lib is notably different from other libs in exception handling. Using cps:

• Any "throw" statements in the procedures will be transformed to an application of the callback onto the error object.
• A "rescue" function is provided, which can be used to catch such exceptions pass through the callbacks.

This consistently recovers the try/catch functionality in continuation passing style programming.

Install

npm install cps

Terminologies

Callback

We call a function of the following form a callback:

function(err, res) {
  // process the err and res
}

A callback is a function that takes two arguments, "err" and "res". Semantically, a non-null "err" corresponds to a program exception; while a null "err" corresponds to normal return without any exceptions.

Procedure

We call a function of the following form a procedure:

function(arg1, arg2, ..., callback) {
  // do some work with the arguments and then invoke the callback to continue
}

A procedure is a function that takes a callback as the last argument. Semantically, a procedure does some work with the input arguments and at some point, call the callback to continue. Note that a call to the "callback" argument MUST always be a tail call. In particular, the following is a procedure:

function(callback) {
  // do some work and then invoke the callback to continue
}

API Document

• seq
• pwhile
• peach
• pmap
• pfor
• rescue
• parallel

Sequence a list of procedures. Note that the result of each procedure is fed into the next procedure in the listed order.

Example

var concatFile = function(f1, f2, resFile, cb) {
    var contentOfA, contentOfB;
    
    cps.seq([
        function(_, cb) {
            fs.readFile(f1, cb);
        },
        function(res, cb) {
            contentOfA = res;
            fs.readFile(f2, cb);
        },
        function(res, cb) {
            contentOfB = res;
            fs.writeFile(resFile, contentA + contentB, cb);
        }
    ], cb);
};

An asynchronized version of while loop.

Example

Consider a world in which arithmatic operations do not exists and must be accomplished through alien technology. Then the Fibonacci function needs to be written in the following way:

var alienAdd = function(a, b, cb) {
    setTimeout(function() {
        cb(null, a + b);
    }, 0);
};

var asyncFib = function(n, cb) {
    if (n < 0) {
        throw new Error('fib input error');
    }
    if (n == 0) {return cb(null, 1);}
    if (n == 1) {return cb(null, 1);}

    var a = 1, 
    b = 1, 
    i = 2;
    cps.seq([
        function(_, cb) {
            cps.pwhile(
                function(cb) {
                    cb(null, i <= n);
                },
                function(cb) {
                    cps.seq([
                        function(_, cb) {
                            alienAdd(a, b, cb);
                        },
                        function(res, cb) {
                            a = b;
                            b = res;
                            alienAdd(i, 1, cb);
                        },
                        function(res, cb) {
                            i = res;
                            cb();
                        }
                    ], cb);
                },
                cb
            );
        },
        function(_, cb) {
            cb(null, b);
        }
    ], cb);
};

Apply a procedure on an array sequentially.

Example

Then in the same "arithmetic-less" world, print out the first 10 Fibonacci numbers.

cps.peach(
    [1,2,3,4,5,6,7,8,9,10],
    function(el, cb) {
        cps.seq([
            function(_, cb) {
                asyncFib(el, cb);
            },
            function(res, cb) {
                console.log(res);
                cb();
            }
        ], cb);

    },
    cb
);

Apply a procedure on an array sequentially, and record the results in another array, which is pass to the callback.

Example

You can also map it out first and then log the result list.

cps.seq([
    function(_, cb) {
        cps.pmap(
            [1,2,3,4,5,6,7,8,9,10],
            function(el, cb) {
                asyncFib(el, cb);
            },
            cb
        );
    },
    function(res, cb) {
        console.log(res);
        cb();
    }
], cb);

Apply a procedure on a sequence of consecutive indices, starting of 0.

Example

var sum = 0;
cps.seq([
    function(_, cb) {
        console.log('here');
        cps.pfor(10, function(i, cb) {
            setTimeout(function() {
                sum += i;
                console.log(sum);
                cb();
            }, 1000);
        }, cb);
    },
    function(_, cb) {
        console.log(sum);
        cb(null, 'ok');
    }
], cb);

An asyned version of try/catch. It take two procedures as arguments. If the first one fails, the second is executed to rescue.

Example

What if there's some invalid input? Let's catch it without disturbing the overall flow.

cps.seq([
    function(_, cb) {
        cps.pmap(
            [1,2,3,4,5,6,7,8,9,10, -1],
            function(el, cb) {
                cps.rescue({
                    'try': function(cb) { // try clause
                        asyncFib(el, cb);
                    },
                    'catch': function(err, cb) { // catch clause
                        console.log(err);
                        cb(null, -1);
                    },
                    'finally': function(cb) { // finally
                        console.log("always print this whether it's good or bad.");
                        cb(); 
                    }
                }, cb);
            },
            cb
        );
    },
    function(res, cb) {
        console.log(res);
        cb();
    }
], cb);

Parallel a list of procedures. The top level callback is only called after each parallel procedure finishes, regardless the procedure succeeds or fails. The callback will never take a non-null error parameter; the result parameter is an array of the following form:

[
   {"status": "ok", "data": res},  /* in case the procedure in the 
                                      corresponding position succeeds 
                                      with result res
                                   */
   {"status": "error", "error": err}  /* in case the procedure in the 
                                         corresponding position fails 
                                         with error err
                                      */
   // ...
]

Example

See "thread b" being printed out before "thread a":

var parallelTest = function(cb) {
    cps.parallel([
        function(cb) {
            setTimeout(function() {
                console.log('3');
                cb(new Error('kaz'));
            }, 3000);
        },
        function(cb) {
            setTimeout(function() {
                console.log('1');
                cb(null, 'ok');
            }, 2000);
        },
        function(cb) {
            setTimeout(function() {
                console.log('2');
                cb(new Error('foobar'));
            }, 1000);
        }
    ], cb);
};

Running this procedure will yield the following output:

2
1
3
[ { status: 'error', error: [Error: kaz] },
  { status: 'ok', data: 'ok' },
  { status: 'error', error: [Error: foobar] } ]