yauzl

yet another unzip library for node. For zipping, see yazl.

Design principles:

• Follow the spec. Don't scan for local file headers. Read the central directory for file metadata. (see No Streaming Unzip API).
• Don't block the JavaScript thread. Use and provide async APIs.
• Keep memory usage under control. Don't attempt to buffer entire files in RAM at once.
• Never crash (if used properly). Don't let malformed zip files bring down client applications who are trying to catch errors.
• Catch unsafe file names. See validateFileName().

Usage

var yauzl = require("yauzl");
 
yauzl.open("path/to/file.zip", {lazyEntries: true}, function(err, zipfile) {
  if (err) throw err;
  zipfile.readEntry();
  zipfile.on("entry", function(entry) {
    if (/\/$/.test(entry.fileName)) {
      // Directory file names end with '/'.
      // Note that entires for directories themselves are optional.
      // An entry's fileName implicitly requires its parent directories to exist.
      zipfile.readEntry();
    } else {
      // file entry
      zipfile.openReadStream(entry, function(err, readStream) {
        if (err) throw err;
        readStream.on("end", function() {
          zipfile.readEntry();
        });
        readStream.pipe(somewhere);
      });
    }
  });
});

See also examples/ for more usage examples.

API

The default for every optional callback parameter is:

function defaultCallback(err) {
  if (err) throw err;
}

open(path, [options], [callback])

Calls fs.open(path, "r") and reads the fd effectively the same as fromFd() would.

options may be omitted or null. The defaults are {autoClose: true, lazyEntries: false, decodeStrings: true, validateEntrySizes: true, strictFileNames: false}.

autoClose is effectively equivalent to:

zipfile.once("end", function() {
  zipfile.close();
});

lazyEntries indicates that entries should be read only when readEntry() is called. If lazyEntries is false, entry events will be emitted as fast as possible to allow pipe()ing file data from all entries in parallel. This is not recommended, as it can lead to out of control memory usage for zip files with many entries. See issue #22. If lazyEntries is true, an entry or end event will be emitted in response to each call to readEntry(). This allows processing of one entry at a time, and will keep memory usage under control for zip files with many entries.

decodeStrings is the default and causes yauzl to decode strings with CP437 or UTF-8 as required by the spec. The exact effects of turning this option off are:

• zipfile.comment, entry.fileName, and entry.fileComment will be Buffer objects instead of Strings.
• Any Info-ZIP Unicode Path Extra Field will be ignored. See extraFields.
• Automatic file name validation will not be performed. See validateFileName().

validateEntrySizes is the default and ensures that an entry's reported uncompressed size matches its actual uncompressed size. This check happens as early as possible, which is either before emitting each "entry" event (for entries with no compression), or during the readStream piping after calling openReadStream(). See openReadStream() for more information on defending against zip bomb attacks.

When strictFileNames is false (the default) and decodeStrings is true, all backslash (\) characters in each entry.fileName are replaced with forward slashes (/). The spec forbids file names with backslashes, but Microsoft's System.IO.Compression.ZipFile class in .NET versions 4.5.0 until 4.6.1 creates non-conformant zipfiles with backslashes in file names. strictFileNames is false by default so that clients can read these non-conformant zipfiles without knowing about this Microsoft-specific bug. When strictFileNames is true and decodeStrings is true, entries with backslashes in their file names will result in an error. See validateFileName(). When decodeStrings is false, strictFileNames has no effect.

The callback is given the arguments (err, zipfile). An err is provided if the End of Central Directory Record cannot be found, or if its metadata appears malformed. This kind of error usually indicates that this is not a zip file. Otherwise, zipfile is an instance of ZipFile.

fromFd(fd, [options], [callback])

Reads from the fd, which is presumed to be an open .zip file. Note that random access is required by the zip file specification, so the fd cannot be an open socket or any other fd that does not support random access.

options may be omitted or null. The defaults are {autoClose: false, lazyEntries: false, decodeStrings: true, validateEntrySizes: true, strictFileNames: false}.

See open() for the meaning of the options and callback.

fromBuffer(buffer, [options], [callback])

Like fromFd(), but reads from a RAM buffer instead of an open file. buffer is a Buffer.

If a ZipFile is acquired from this method, it will never emit the close event, and calling close() is not necessary.

options may be omitted or null. The defaults are {lazyEntries: false, decodeStrings: true, validateEntrySizes: true, strictFileNames: false}.

See open() for the meaning of the options and callback. The autoClose option is ignored for this method.

fromRandomAccessReader(reader, totalSize, [options], [callback])

This method of reading a zip file allows clients to implement their own back-end file system. For example, a client might translate read calls into network requests.

The reader parameter must be of a type that is a subclass of RandomAccessReader that implements the required methods. The totalSize is a Number and indicates the total file size of the zip file.

options may be omitted or null. The defaults are {autoClose: true, lazyEntries: false, decodeStrings: true, validateEntrySizes: true, strictFileNames: false}.

See open() for the meaning of the options and callback.

dosDateTimeToDate(date, time)

Converts MS-DOS date and time data into a JavaScript Date object. Each parameter is a Number treated as an unsigned 16-bit integer. Note that this format does not support timezones, so the returned object will use the local timezone.

validateFileName(fileName)

Returns null or a String error message depending on the validity of fileName. If fileName starts with "/" or /[A-Za-z]:\// or if it contains ".." path segments or "\\", this function returns an error message appropriate for use like this:

var errorMessage = yauzl.validateFileName(fileName);
if (errorMessage != null) throw new Error(errorMessage);

This function is automatically run for each entry, as long as decodeStrings is true. See open(), strictFileNames, and Event: "entry" for more information.

Class: ZipFile

The constructor for the class is not part of the public API. Use open(), fromFd(), fromBuffer(), or fromRandomAccessReader() instead.

Event: "entry"

Callback gets (entry), which is an Entry. See open() and readEntry() for when this event is emitted.

If decodeStrings is true, entries emitted via this event have already passed file name validation. See validateFileName() and open() for more information.

If validateEntrySizes is true and this entry's compressionMethod is 0 (stored without compression), this entry has already passed entry size validation. See open() for more information.

Event: "end"

Emitted after the last entry event has been emitted. See open() and readEntry() for more info on when this event is emitted.

Event: "close"

Emitted after the fd is actually closed. This is after calling close() (or after the end event when autoClose is true), and after all stream pipelines created from openReadStream() have finished reading data from the fd.

If this ZipFile was acquired from fromRandomAccessReader(), the "fd" in the previous paragraph refers to the RandomAccessReader implemented by the client.

If this ZipFile was acquired from fromBuffer(), this event is never emitted.

Event: "error"

Emitted in the case of errors with reading the zip file. (Note that other errors can be emitted from the streams created from openReadStream() as well.) After this event has been emitted, no further entry, end, or error events will be emitted, but the close event may still be emitted.

readEntry()

Causes this ZipFile to emit an entry or end event (or an error event). This method must only be called when this ZipFile was created with the lazyEntries option set to true (see open()). When this ZipFile was created with the lazyEntries option set to true, entry and end events are only ever emitted in response to this method call.

The event that is emitted in response to this method will not be emitted until after this method has returned, so it is safe to call this method before attaching event listeners.

After calling this method, calling this method again before the response event has been emitted will cause undefined behavior. Calling this method after the end event has been emitted will cause undefined behavior. Calling this method after calling close() will cause undefined behavior.

openReadStream(entry, [options], callback)

entry must be an Entry object from this ZipFile. callback gets (err, readStream), where readStream is a Readable Stream that provides the file data for this entry. If this zipfile is already closed (see close()), the callback will receive an err.

options may be omitted or null, and has the following defaults:

{
  decompress: entry.isCompressed() ? true : null,
  decrypt: null,
  start: 0,                  // actually the default is null, see below
  end: entry.compressedSize, // actually the default is null, see below
}

If the entry is compressed (with a supported compression method), and the decompress option is true (or omitted), the read stream provides the decompressed data. Omitting the decompress option is what most clients should do.

The decompress option must be null (or omitted) when the entry is not compressed (see isCompressed()), and either true (or omitted) or false when the entry is compressed. Specifying decompress: false for a compressed entry causes the read stream to provide the raw compressed file data without going through a zlib inflate transform.

If the entry is encrypted (see isEncrypted()), clients may want to avoid calling openReadStream() on the entry entirely. Alternatively, clients may call openReadStream() for encrypted entries and specify decrypt: false. If the entry is also compressed, clients must also specify decompress: false. Specifying decrypt: false for an encrypted entry causes the read stream to provide the raw, still-encrypted file data. (This data includes the 12-byte header described in the spec.)

The decrypt option must be null (or omitted) for non-encrypted entries, and false for encrypted entries. Omitting the decrypt option (or specifying it as null) for an encrypted entry will result in the callback receiving an err. This default behavior is so that clients not accounting for encrypted files aren't surprised by bogus file data.

The start (inclusive) and end (exclusive) options are byte offsets into this entry's file data, and can be used to obtain part of an entry's file data rather than the whole thing. If either of these options are specified and non-null, then the above options must be used to obain the file's raw data. Speficying {start: 0, end: entry.compressedSize} will result in the complete file, which is effectively the default values for these options, but note that unlike omitting the options, when you specify start or end as any non-null value, the above requirement is still enforced that you must also pass the appropriate options to get the file's raw data.

It's possible for the readStream provided to the callback to emit errors for several reasons. For example, if zlib cannot decompress the data, the zlib error will be emitted from the readStream. Two more error cases (when validateEntrySizes is true) are if the decompressed data has too many or too few actual bytes compared to the reported byte count from the entry's uncompressedSize field. yauzl notices this false information and emits an error from the readStream after some number of bytes have already been piped through the stream.

This check allows clients to trust the uncompressedSize field in Entry objects. Guarding against zip bomb attacks can be accomplished by doing some heuristic checks on the size metadata and then watching out for the above errors. Such heuristics are outside the scope of this library, but enforcing the uncompressedSize is implemented here as a security feature.

It is possible to destroy the readStream before it has piped all of its data. To do this, call readStream.destroy(). You must unpipe() the readStream from any destination before calling readStream.destroy(). If this zipfile was created using fromRandomAccessReader(), the RandomAccessReader implementation must provide readable streams that implement a .destroy() method (see randomAccessReader._readStreamForRange()) in order for calls to readStream.destroy() to work in this context.

close()

Causes all future calls to openReadStream() to fail, and closes the fd, if any, after all streams created by openReadStream() have emitted their end events.

If the autoClose option is set to true (see open()), this function will be called automatically effectively in response to this object's end event.

If the lazyEntries option is set to false (see open()) and this object's end event has not been emitted yet, this function causes undefined behavior. If the lazyEntries option is set to true, you can call this function instead of calling readEntry() to abort reading the entries of a zipfile.

It is safe to call this function multiple times; after the first call, successive calls have no effect. This includes situations where the autoClose option effectively calls this function for you.

If close() is never called, then the zipfile is "kept open". For zipfiles created with fromFd(), this will leave the fd open, which may be desirable. For zipfiles created with open(), this will leave the underlying fd open, thereby "leaking" it, which is probably undesirable. For zipfiles created with fromRandomAccessReader(), the reader's close() method will never be called. For zipfiles created with fromBuffer(), the close() function has no effect whether called or not.

Regardless of how this ZipFile was created, there are no resources other than those listed above that require cleanup from this function. This means it may be desirable to never call close() in some usecases.

isOpen

Boolean. true until close() is called; then it's false.

entryCount

Number. Total number of central directory records.

comment

String. Always decoded with CP437 per the spec.

If decodeStrings is false (see open()), this field is the undecoded Buffer instead of a decoded String.

Class: Entry

Objects of this class represent Central Directory Records. Refer to the zipfile specification for more details about these fields.

These fields are of type Number:

• versionMadeBy
• versionNeededToExtract
• generalPurposeBitFlag
• compressionMethod
• lastModFileTime (MS-DOS format, see getLastModDateTime)
• lastModFileDate (MS-DOS format, see getLastModDateTime)
• crc32
• compressedSize
• uncompressedSize
• fileNameLength (bytes)
• extraFieldLength (bytes)
• fileCommentLength (bytes)
• internalFileAttributes
• externalFileAttributes
• relativeOffsetOfLocalHeader

fileName

String. Following the spec, the bytes for the file name are decoded with UTF-8 if generalPurposeBitFlag & 0x800, otherwise with CP437. Alternatively, this field may be populated from the Info-ZIP Unicode Path Extra Field (see extraFields).

This field is automatically validated by validateFileName() before yauzl emits an "entry" event. If this field would contain unsafe characters, yauzl emits an error instead of an entry.

If decodeStrings is false (see open()), this field is the undecoded Buffer instead of a decoded String. Therefore, generalPurposeBitFlag and any Info-ZIP Unicode Path Extra Field are ignored. Furthermore, no automatic file name validation is performed for this file name.

extraFields

Array with each entry in the form {id: id, data: data}, where id is a Number and data is a Buffer.

This library looks for and reads the ZIP64 Extended Information Extra Field (0x0001) in order to support ZIP64 format zip files.

This library also looks for and reads the Info-ZIP Unicode Path Extra Field (0x7075) in order to support some zipfiles that use it instead of General Purpose Bit 11 to convey UTF-8 file names. When the field is identified and verified to be reliable (see the zipfile spec), the the file name in this field is stored in the fileName property, and the file name in the central directory record for this entry is ignored. Note that when decodeStrings is false, all Info-ZIP Unicode Path Extra Fields are ignored.

None of the other fields are considered significant by this library. Fields that this library reads are left unalterned in the extraFields array.

fileComment

String decoded with the charset indicated by generalPurposeBitFlag & 0x800 as with the fileName. (The Info-ZIP Unicode Path Extra Field has no effect on the charset used for this field.)

If decodeStrings is false (see open()), this field is the undecoded Buffer instead of a decoded String.

Prior to yauzl version 2.7.0, this field was erroneously documented as comment instead of fileComment. For compatibility with any code that uses the field name comment, yauzl creates an alias field named comment which is identical to fileComment.

getLastModDate()

Effectively implemented as:

return dosDateTimeToDate(this.lastModFileDate, this.lastModFileTime);

isEncrypted()

Returns is this entry encrypted with "Traditional Encryption". Effectively implemented as:

return (this.generalPurposeBitFlag & 0x1) !== 0;

See openReadStream() for the implications of this value.

Note that "Strong Encryption" is not supported, and will result in an "error" event emitted from the ZipFile.

isCompressed()

Effectively implemented as:

return this.compressionMethod === 8;

See openReadStream() for the implications of this value.

Class: RandomAccessReader

This class is meant to be subclassed by clients and instantiated for the fromRandomAccessReader() function.

An example implementation can be found in test/test.js.

randomAccessReader._readStreamForRange(start, end)

Subclasses must implement this method.

start and end are Numbers and indicate byte offsets from the start of the file. end is exclusive, so _readStreamForRange(0x1000, 0x2000) would indicate to read 0x1000 bytes. end - start will always be at least 1.

This method should return a readable stream which will be pipe()ed into another stream. It is expected that the readable stream will provide data in several chunks if necessary. If the readable stream provides too many or too few bytes, an error will be emitted. (Note that validateEntrySizes has no effect on this check, because this is a low-level API that should behave correctly regardless of the contents of the file.) Any errors emitted on the readable stream will be handled and re-emitted on the client-visible stream (returned from zipfile.openReadStream()) or provided as the err argument to the appropriate callback (for example, for fromRandomAccessReader()).

The returned stream must implement a method .destroy() if you call readStream.destroy() on streams you get from openReadStream(). If you never call readStream.destroy(), then streams returned from this method do not need to implement a method .destroy(). .destroy() should abort any streaming that is in progress and clean up any associated resources. .destroy() will only be called after the stream has been unpipe()d from its destination.

Note that the stream returned from this method might not be the same object that is provided by openReadStream(). The stream returned from this method might be pipe()d through one or more filter streams (for example, a zlib inflate stream).

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

Subclasses may implement this method. The default implementation uses createReadStream() to fill the buffer.

This method should behave like fs.read().

randomAccessReader.close(callback)

Subclasses may implement this method. The default implementation is effectively setImmediate(callback);.

callback takes parameters (err).

This method is called once the all streams returned from _readStreamForRange() have ended, and no more _readStreamForRange() or read() requests will be issued to this object.

How to Avoid Crashing

When a malformed zipfile is encountered, the default behavior is to crash (throw an exception). If you want to handle errors more gracefully than this, be sure to do the following:

• Provide callback parameters where they are allowed, and check the err parameter.
• Attach a listener for the error event on any ZipFile object you get from open(), fromFd(), fromBuffer(), or fromRandomAccessReader().
• Attach a listener for the error event on any stream you get from openReadStream().

Minor version updates to yauzl will not add any additional requirements to this list.

Limitations

No Streaming Unzip API

Due to the design of the .zip file format, it's impossible to interpret a .zip file from start to finish (such as from a readable stream) without sacrificing correctness. The Central Directory, which is the authority on the contents of the .zip file, is at the end of a .zip file, not the beginning. A streaming API would need to either buffer the entire .zip file to get to the Central Directory before interpreting anything (defeating the purpose of a streaming interface), or rely on the Local File Headers which are interspersed through the .zip file. However, the Local File Headers are explicitly denounced in the spec as being unreliable copies of the Central Directory, so trusting them would be a violation of the spec.

Any library that offers a streaming unzip API must make one of the above two compromises, which makes the library either dishonest or nonconformant (usually the latter). This library insists on correctness and adherence to the spec, and so does not offer a streaming API.

Here is a way to create a spec-conformant .zip file using the zip command line program (Info-ZIP) available in most unix-like environments, that is (nearly) impossible to parse correctly with a streaming parser:

$ echo -ne '\x50\x4b\x07\x08\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00' > file.txt
$ zip -q0 - file.txt | cat > out.zip

This .zip file contains a single file entry that uses General Purpose Bit 3, which means the Local File Header doesn't know the size of the file. Any streaming parser that encounters this situation will either immediately fail, or attempt to search for the Data Descriptor after the file's contents. The file's contents is a sequence of 16-bytes crafted to exactly mimic a valid Data Descriptor for an empty file, which will fool any parser that gets this far into thinking that the file is empty rather than containing 16-bytes. What follows the file's real contents is the file's real Data Descriptor, which will likely cause some kind of signature mismatch error for a streaming parser (if one hasn't occurred already).

By using General Purpose Bit 3 (and compression method 0), it's possible to create arbitrarily ambiguous .zip files that distract parsers with file contents that contain apparently valid .zip file metadata.

Limitted ZIP64 Support

For ZIP64, only zip files smaller than 8PiB are supported, not the full 16EiB range that a 64-bit integer should be able to index. This is due to the JavaScript Number type being an IEEE 754 double precision float.

The Node.js fs module probably has this same limitation.

ZIP64 Extensible Data Sector Is Ignored

The spec does not allow zip file creators to put arbitrary data here, but rather reserves its use for PKWARE and mentions something about Z390. This doesn't seem useful to expose in this library, so it is ignored.

No Multi-Disk Archive Support

This library does not support multi-disk zip files. The multi-disk fields in the zipfile spec were intended for a zip file to span multiple floppy disks, which probably never happens now. If the "number of this disk" field in the End of Central Directory Record is not 0, the open(), fromFd(), fromBuffer(), or fromRandomAccessReader() callback will receive an err. By extension the following zip file fields are ignored by this library and not provided to clients:

• Disk where central directory starts
• Number of central directory records on this disk
• Disk number where file starts

Limited Encryption Handling

You can detect when a file entry is encrypted with "Traditional Encryption" via isEncrypted(), but yauzl will not help you decrypt it. See openReadStream().

If a zip file contains file entries encrypted with "Strong Encryption", yauzl emits an error.

If the central directory is encrypted or compressed, yauzl emits an error.

Local File Headers Are Ignored

Many unzip libraries mistakenly read the Local File Header data in zip files. This data is officially defined to be redundant with the Central Directory information, and is not to be trusted. Aside from checking the signature, yauzl ignores the content of the Local File Header.

No CRC-32 Checking

This library provides the crc32 field of Entry objects read from the Central Directory. However, this field is not used for anything in this library.

versionNeededToExtract Is Ignored

The field versionNeededToExtract is ignored, because this library doesn't support the complete zip file spec at any version,

No Support For Obscure Compression Methods

Regarding the compressionMethod field of Entry objects, only method 0 (stored with no compression) and method 8 (deflated) are supported. Any of the other 15 official methods will cause the openReadStream() callback to receive an err.

Data Descriptors Are Ignored

There may or may not be Data Descriptor sections in a zip file. This library provides no support for finding or interpreting them.

Archive Extra Data Record Is Ignored

There may or may not be an Archive Extra Data Record section in a zip file. This library provides no support for finding or interpreting it.

No Language Encoding Flag Support

Zip files officially support charset encodings other than CP437 and UTF-8, but the zip file spec does not specify how it works. This library makes no attempt to interpret the Language Encoding Flag.

Change History

• 2.10.0
  • Added support for non-conformant zipfiles created by Microsoft, and added option strictFileNames to disable the workaround. issue #66, issue #88
• 2.9.2
  • Removed tools/hexdump-zip.js and tools/hex2bin.js. Those tools are now located here: thejoshwolfe/hexdump-zip and thejoshwolfe/hex2bin
  • Worked around performance problem with zlib when using fromBuffer() and readStream.destroy() for large compressed files. issue #87
• 2.9.1
  • Removed console.log() accidentally introduced in 2.9.0. issue #64
• 2.9.0
  • Throw an exception if readEntry() is called without lazyEntries:true. Previously this caused undefined behavior. issue #63
• 2.8.0
  • Added option validateEntrySizes. issue #53
  • Added examples/promises.js
  • Added ability to read raw file data via decompress and decrypt options. issue #11, issue #38, pull #39
  • Added start and end options to openReadStream(). issue #38
• 2.7.0
  • Added option decodeStrings. issue #42
  • Fixed documentation for entry.fileComment and added compatibility alias. issue #47
• 2.6.0
  • Support Info-ZIP Unicode Path Extra Field, used by WinRAR for Chinese file names. issue #33
• 2.5.0
  • Ignore malformed Extra Field that is common in Android .apk files. issue #31
• 2.4.3
  • Fix crash when parsing malformed Extra Field buffers. issue #31
• 2.4.2
  • Remove .npmignore and .travis.yml from npm package.
• 2.4.1
  • Fix error handling.
• 2.4.0
  • Add ZIP64 support. issue #6
  • Add lazyEntries option. issue #22
  • Add readStream.destroy() method. issue #26
  • Add fromRandomAccessReader(). issue #14
  • Add examples/unzip.js.
• 2.3.1
  • Documentation updates.
• 2.3.0
  • Check that uncompressedSize is correct, or else emit an error. issue #13
• 2.2.1
  • Update dependencies.
• 2.2.0
  • Update dependencies.
• 2.1.0
  • Remove dependency on iconv.
• 2.0.3
  • Fix crash when trying to read a 0-byte file.
• 2.0.2
  • Fix event behavior after errors.
• 2.0.1
  • Fix bug with using iconv.
• 2.0.0
  • Initial release.