binary-split-streams2

This module is a simple streams2 implementation of a stream tokenizer. It takes a stream in, and emits the chunks between the specified delimiter, in the same way as String.prototype.split makes arrays from a string.

This module operates entirely on buffers and is implemented as a Transform stream. There are no dependencies. This could cause problems with certain delimiters depending on the character encoding of the input; be aware of this. For my purpose (splitting JSON on newlines), it will work just fine; newlines can't be embedded in UTF-8 characters.

Supported Node versions: 0.10 through latest/8.x

Usage

var split = require('binary-split-streams2');
require('fs').createReadStream(__filename)
    .pipe(split())
    .on('data', function (chunk) {
        console.log(chunk);
    });

You can specify the empty string, '' to split between every character, just like with String.prototype.split. The default delimiter is require('os').EOL.

Truncation

As of version 1.1.0, you may specify a max size of the internal buffer. The buffer is truncated to be no greater than this size after emitting new chunks, so it is possible that this package will emit a chunk of size greater than maxBuffer when it receives that entire chunk plus the splitter in a single write. You can supply the option strictTruncation: true if you want all chunks to be forcibly truncated for you, which will guarantee that your reads will be <= maxBuffer in size.

When the buffer is truncated, the stream will emit a truncated event with the amount of bytes that were dropped. If strictTruncation is enabled, it will also emit a truncated event when a chunk it emits that would have been too long is truncated. The sum of the values passed to the truncated event and the sum of the lengths of your read chunks will equal the size of your input minus all splitters.

var split = require('binary-split-streams2');
require('fs').createReadStream(__filename)
    .pipe(split({maxBuffer: 1000, strictTruncation: true}))
    .on('data', function (chunk) {
        assert(chunk.length <= 1000);
        console.log(chunk);
    })
    .on('truncated', function (amount) {
        console.log('dropped %d bytes from the input stream', amount);
    });

Other modules

split is a useful module and more full-featured than this, but it uses through which seems to only provide old-style (streams1) streams. Streams2 streams can be lazy, and that's helpful. There is also a module called binary-split which does more or less the same thing as this module, but also uses through and seems abandoned.

binary-split makes a bit of a case for performance, so I tested this module to make sure I wasn't putting out something "worse" than that module; this module runs about 150% as fast as that one in a simple test utilizing a similar scenario to the one described there (2.4 gb file with ~ 500 byte lines, split on newline). Recent versions of binary-split have been equal to or faster than this module, however as of version 1.1.0, in node 4.0 or higher, this module is again much faster.