better-fd-slicer

better-fd-slicer

Safe fs.ReadStream and fs.WriteStream using the same fd.

Let's say that you want to perform a parallel upload of a file to a remote server. To do this, we want to create multiple read streams. The first thing you might think of is to use the {start: 0, end: 0} API of fs.createReadStream. This gives you two choices:

1. Use the same file descriptor for all fs.ReadStream objects.
2. Open the file multiple times, resulting in a separate file descriptor for each read stream.

Neither of these are acceptable options. The first one is a severe bug, because the API docs for fs.write state:

Note that it is unsafe to use fs.write multiple times on the same file without waiting for the callback. For this scenario, fs.createWriteStream is strongly recommended.

fs.createWriteStream will solve the problem if you only create one of them for the file descriptor, but it will exhibit this unsafety if you create multiple write streams per file descriptor.

The second option suffers from a race condition. For each additional time the file is opened after the first, it is possible that the file is modified. So in our parallel uploading example, we might upload a corrupt file that never existed on the client's computer.

This module solves this problem by providing createReadStream and createWriteStream that operate on a shared file descriptor and provides the convenient stream API while still allowing slicing and dicing.

This module also gives you some additional power that the builtin fs.createWriteStream do not give you. These features are:

• Emitting a 'progress' event on write.
• Ability to set a maximum size and emit an error if this size is exceeded.
• Ability to create an FdSlicer instance from a Buffer. This enables you to provide API for handling files as well as buffers using the same API.

Usage

import {createFromFd} from 'better-fd-slicer';
import fs from 'node:fs';

fs.open("file.txt", 'r', function (err, fd) {
    if (err) throw err;
    const slicer = createFromFd(fd);
    const firstPart = slicer.createReadStream({start: 0, end: 100});
    const secondPart = slicer.createReadStream({start: 100});
    const firstOut = fs.createWriteStream("first.txt");
    const secondOut = fs.createWriteStream("second.txt");
    firstPart.pipe(firstOut);
    secondPart.pipe(secondOut);
});

You can also create from a buffer:

import {createFromBuffer} from 'better-fd-slicer';
import fs from 'node:fs';

const slicer = createFromBuffer(someBuffer);
const firstPart = slicer.createReadStream({start: 0, end: 100});
const secondPart = slicer.createReadStream({start: 100});
const firstOut = fs.createWriteStream("first.txt");
const secondOut = fs.createWriteStream("second.txt");

firstPart.pipe(firstOut);
secondPart.pipe(secondOut);

API Documentation

fdSlicer.createFromFd(fd, [options])

import {createFromFd} from 'better-fd-slicer';

fs.open("file.txt", 'r', function (err, fd) {
    if (err) throw err;
    const slicer = createFromFd(fd);
    // ...
});

Make sure fd is a properly initialized file descriptor. If you want to use createReadStream make sure you open it for reading and if you want to use createWriteStream make sure you open it for writing.

options is an optional object which can contain:

• autoClose - if set to true, the file descriptor will be automatically closed once the last stream that references it is closed. Defaults to false. ref() and unref() can be used to increase or decrease the reference count, respectively.

fdSlicer.createFromBuffer(buffer, [options])

import {createFromBuffer} from 'better-fd-slicer';
const slicer = createFromBuffer(someBuffer);
// ...

options is an optional object which can contain:

• maxChunkSize - A Number of bytes. see createReadStream(). If falsey, defaults to unlimited.

Properties

fd

The file descriptor passed in. undefined if created from a buffer.

Methods

createReadStream(options)

Available options:

• start - Number. The offset into the file to start reading from. Defaults to 0.
• end - Number. Exclusive upper bound offset into the file to stop reading from.
• highWaterMark - Number. The maximum number of bytes to store in the internal buffer before ceasing to read from the underlying resource. Defaults to 16 KB.
• encoding - String. If specified, then buffers will be decoded to strings using the specified encoding. Defaults to null.

The ReadableStream that this returns has these additional methods:

• destroy(err) - stop streaming. err is optional and is the error that will be emitted in order to cause the streaming to stop. Defaults to new Error("stream destroyed").

If maxChunkSize was specified (see createFromBuffer()), the read stream will provide chunks of at most that size. Normally, the read stream provides the entire range requested in a single chunk, but this can cause performance problems in some circumstances. See thejoshwolfe/yauzl#87.

createWriteStream(options)

Available options:

• start - Number. The offset into the file to start writing to. Defaults to 0.
• end - Number. Exclusive upper bound offset into the file. If this offset is reached, the write stream will emit an 'error' event and stop functioning. In this situation, err.code === 'ETOOBIG'. Defaults to Infinity.
• highWaterMark - Number. Buffer level when write() starts returning false. Defaults to 16KB.
• decodeStrings - Boolean. Whether or not to decode strings into Buffers before passing them to _write(). Defaults to true.

The WritableStream that this returns has these additional methods:

• destroy() - stop streaming

And these additional properties:

• bytesWritten - number of bytes written to the stream

And these additional events:

• 'progress' - emitted when bytesWritten changes.

read(buffer, offset, length, position, callback)

Equivalent to fs.read, but with concurrency protection. callback must be defined.

write(buffer, offset, length, position, callback)

Equivalent to fs.write, but with concurrency protection. callback must be defined.

ref()

Increase the autoClose reference count by 1.

unref()

Decrease the autoClose reference count by 1.

Events

'error'

Emitted if fs.close returns an error when auto closing.

'close'

Emitted when FdSlicer closes the file descriptor due to autoClose. Never emitted if created from a buffer.

Credits

This library is heavily based on an excellent fd-slicer library, and uses some of it (modernized and improved) code.