CODEWITHSUNDEEP
javascriptnode.js
xis
xis

Reputation: 24850

Efficiently read file line-by-line in node

I have already learned that readline can be used to read the file line by line, e.g.

readline
    .createInterface({input: fs.createReadStream('xxx')})
    .on('line', (line) => { apply_regexp_on_line })
    .on('close', () => { report_all_regexps });

However, this is pretty slow, since I compared the performance of grep and JavaScript regexp, and the latter has better performance on the regexps I tested. (see benchmark) So I think I have to blame the node async readline.

In my situation, I do not care async at all, I just need to exploit the fast regexp from JavaScript to process very large log files (typically 1-2GB, sometimes up to 10GB). What is the best way of doing this? My only concern is speed.

Bonus points: some of the log files are gzipped, so I need to uncompress them. If someone can recommend me a fast line-by-line reader for both plain text and gzipped text exists, I would be really appreciated.

Upvotes: 13

Views: 3166

Answers (1)

Tomalak
Tomalak

Reputation: 338208

How does this hold up against your data?

// module linegrep.js
'use strict';
var through2 = require('through2');
var StringDecoder = require('string_decoder').StringDecoder

function grep(regex) {
    var decoder = new StringDecoder('utf8'),
        last = "",
        lineEnd = /\r?\n/;

    var stream = through2({}, function transform(chunk, enc, cb) {
        var lines = decoder.write(last + chunk).split(lineEnd), i;
        last = lines.pop();
        for (i = 0; i < lines.length; i++) {
            if (regex.test(lines[i])) this.push(lines[i]);
        }
        cb();
    }, function flush(cb) {
        if (regex.test(last)) this.push(last);
        cb();
    });
    stream._readableState.objectMode = true;
    return stream;
}

module.exports = grep;

and 

// index.js

'use strict';
var fs = require('fs');
var zlib = require('zlib');
var grep = require('./linegrep');

function grepFile(filename, regex) {
    var rstream = fs.createReadStream(filename, {highWaterMark: 172 * 1024});
    if (/\.gz$/.test(filename)) rstream = rstream.pipe(zlib.createGunzip());
    return rstream
        .pipe(grep(regex));
}

// -------------------------------------------------------------------------

var t = Date.now(), mc = 0;
grepFile('input.txt', /boot\.([a-z]+)_head\./).on('data', function (line) {
    mc++;
    console.log(line);
}).on('end', function () {
    console.log( mc + " matches, " + (Date.now() - t) + " ms" );
});

This turns a file stream into an object stream of lines, maps them through your regex and returns only the matching lines.

Upvotes: 1

Related Questions