righthandabacus/filesnitch

Snitch up multipart files as one read-only file objects

FileSnitch

This is a module allows you to concatenate multiple files on the fly and present as a single read-only file object.

Introduction

In Python Standard Libarary, there is a module fileinput which provides similar function but useful only for text input, supporting only the readline() function. That module does not support rewinding as well.

This module, filesnitch allows one to seek() and read() files. A typical use case is a multipart zip file. For example, we have multiple files: foo.zip.001, foo.zip.002, foo.zip.003. To extract some file from it, we should combine them with the following shell command

cat foo.zip.001 foo.zip.002 foo.zip.003 > foo.zip

Afterwards we can use it, for example using zipfile module in Python

import zipfile
with zipfile.ZipFile("foo.zip") as zipfp:
    for zipinfo in zipfp.infolist():
        print("File {} size {}".format(zipinfo.filename, zipinfo.file_size))

With this module, we can avoid creating a combined file but use the chunks directly:

import zipfile
import filesnitch

with filesnitch.input(files=["foo.zip.001", "foo.zip.002", "foo.zip.003"]) as bigfile:
    with zipfile.ZipFile(bigfile) as zipfp:
        for zipinfo in zipfp.infolist():
            print("File {} size {}".format(zipinfo.filename, zipinfo.file_size))

Generally, multiple files can be concatenated as a file object and use in everywhere that a read-only file object is expected.

Getting the latest code

To get the latest code using git, simply type:

git clone git://github.com/righthandabacus/filesnitch.git

Install

You can use pip to install from pypi:

pip install filesnitch

or, from the source directory:

python setup.py install

Dependencies

filesnitch has no dependencies besides standard Python libraries. Python 3 required.

Examples

Create a file object of multiple files, tell the total file size, and read the last 20 bytes:

import io
import filesnitch

filenames = ["foo.zip.001", "foo.zip.002", "foo.zip.003"]
fileobj = filesnitch.FileSnitch(filenames)  # by default in binary mode
fileobj.seek(0, io.SEEK_END)
filesize = fileobj.tell()
fileobj.seek(-20, io.SEEK_END)
lastbytes = fileobj.read(20)
print("Total file size {}. Last 20 bytes:\n{}".format(filesize, repr(lastbytes))

Above we use the standard function seek() and tell() to move within the concatenated file as if the file object is a standard io.IOBase. The read() call accepts the size argument for the number of bytes to read. If no size argument provided, it will read until EOF.

Read all files in one shot as a large blob:

import io
import filesnitch

filenames = ["foo.zip.001", "foo.zip.002", "foo.zip.003"]
with filesnitch.FileSnitch(filenames) as fileobj:
    blob = fileobj.readall()

The readall() call is same as read() without argument, both will read until EOF. FileSnitch class can be used as a context manager.

We can also pass in file objects into FileSnitch() as long as those file objects supports seek and read correctly. An example is as follows: The diabetic-retinopathy-detection.zip file is a 83GB zip file from Kaggle (with correct tools, we can download it using the command kaggle competitions download -c diabetic-retinopathy-detection). In it, there are seven files, named test.zip.001 to test.zip.007, that are parts of a zip file of images. To extract one image from it without deflating any zip file onto disk, we can do the following:

import zipfile
import filesnitch

# this call to zipfile needs `allowZip64=True` because of the size is huge
with zipfile.ZipFile("diabetic-retinopathy-detection.zip", allowZip64=True) as zip83G:
    # open each file inside the 83GB zip file, then snitch them up as one file object
    zipparts = [zip83G.open(f, mode="r") for f in [
        "test.zip.001",
        "test.zip.002",
        "test.zip.003",
        "test.zip.004",
        "test.zip.005",
        "test.zip.006",
        "test.zip.007",
    ]]
    with filesnitch.input(files=zipparts, mode='rb') as testzip:
        # use the snitched file as another zip file, then read a file inside it
        with zipfile.ZipFile(testzip) as zfp:
            with zfp.open("test/10000_right.jpeg", mode="r") as fp:
                with open("10000_right.jpeg", "wb") as outfp:
                    outfp.write(fp.read())

API

FileSnitch class derived from FileInput class and the methods filename(), fileno(), lineno(), filelineno(), isfirstline(), isstdin(), nextfile() and close() are same as the parent class. Contrary to FileInput, files opened by FileSnitch by default are in mode 'rb' instead of 'r'. Random access and readline() should not be mixed and it is not suggested to use readline() call with FileSnitch.

FileSnitch.read(size=-1)

Read the specified number of bytes/characters or until the end of the concatenated file. Calling read() with no argument or with size=-1 will read until the concatenated file is exhaustted. Depends on the mode that the files are opened, it returns a str or bytes object.

FileSnitch.readall()

Same as read(size=-1). It read from the current position until the end of the concatenated file.

FileSnitch.tell()

Returns the current position, in number of bytes from the beginning of the concatenated file.

FileSnitch.seek(offset, wherence=io.SEEK_SET)

Reset the current position in the concatenated file. It has the same syntax as the seek() function in other file object.

FileSnitch.seekable()

Returns a boolean for whether the concatenated file is seekable. It is a best effort guess for whether all the component files are seekable.

FileSnitch.readable()

Always returns True.

FileSnitch.isatty()

Same as FileSnitch.seekable()

filesnitch's Issues

Wheel does not contain code

The wheel (.whl) file for version 0.1 of this project on PyPI does not contain any code. This is due to the fact that filesnitch.py is a flat module, rather than a package, and so setuptools.find_packages() does not find it. The correct way to declare a flat-module project in setup.py is by replacing the packages=... argument to setup() with a py_modules argument, like so:

py_modules=['filesnitch']