node.js dynamic javascript execution performance

Question

I'd like to execute an arbitrary - user supplied - piece of javascript code on each line of a file. I created a simple node command line application, that will - as an example - just take each line and will output its length:

#!/usr/bin/env node
// eachline.js - execute some js per line in a file

var fs = require('fs'),
    readline = require('readline');
    vm = require('vm');

var args = process.argv.slice(2);

if (args.length < 1) {
    console.log("Usage: eachline.js FILENAME")
    process.exit(code=1)
}

var rd = readline.createInterface({
    input: fs.createReadStream(args[0]),
    output: process.stdout,
    terminal: false
});

var context = vm.createContext({});
// this snippet should be user supplied, arbitrary javascript
var script = vm.createScript('output = input.length');

rd.on('line', function(line) {
    context.input = line;
    script.runInContext(context);
    console.log(context.output);
});

The performance is not optimal. For a file with 1 million lines, it takes several minutes:

$ time ./eachline.js 1M.txt > /dev/null

real    4m14.366s
user    4m12.200s
sys     0m4.545s

With awk this simple operation takes about five seconds. With a similar program , that executes arbitrary javascript a bit more than ten seconds.

I haven't worked with node before, so the program above is my first uneducated guess at a solution. How could I improve performance?

Edit: Above source code in a gist: https://gist.github.com/miku/31864156938fcd0c8430

---

PS. Goal is to have a program that can run arbitrary javascript on each line of a file fast.

Answer 1

Try this:

var context = vm.createContext({});
// this snippet should be user supplied, arbitrary javascript
var script = vm.createScript('(function (input) { return input.length; })');
var fn = script.runInContext(context);

rd.on('line', function(line) {
    console.log(fn(line));
});

The idea is to wrap provided expression in a function, evaluate the script and then use the function. I think v8 doesn't optimize code inside script because it is not supposed to run more than once.