pyckup 0.1.1

Tools to provide easy access to prepared data to data scientists that can't be asked.

pyckup

Tools to provide easy access to prepared data to data scientists that can't be asked.

They just want to get on with the fun -- not get stuck in data access and data preparation concerns. And they should want that!

Of course, someone needed to do the work of getting the data from where and how it is, to where and how it needs to be (for a particular problem and context).

What we believe is that this work should not only be less tedious and less time-consuming (see py2store and related for that!), but also, once it's done, it shouldn't have to be re-done every time someone wants to kick the data around.

So we made pyckup.

We hope it helps.

install

pip install pyckup

Usage

A protocol is what's to the left of :// in a url. You've seen'em. For example, http and https. Well, if you try grabbing a valid url, you'll get bytes out of it, without having to make an actual http request yourself.

>>> b = grab('https://raw.githubusercontent.com/i2mint/pyckup/master/LICENSE')
>>> isinstance(b, bytes)
True

Now let's put those bytes in a file:

>>> from tempfile import mktemp
>>> filepath = mktemp()
>>> with open(filepath, 'wb') as fp:
...     n_bytes_written = fp.write(b);
>>> assert n_bytes_written == len(b)

A filepath is a url. So you should be able to grab it's contents too.

>>> file_bytes = grab(filepath)
>>> assert file_bytes == b

There wasn't really a protocol there, but when grab sees that the url you're passing it starts with a slash, it automatically prepends a file:// to it, like most browswers do. You can also specify the file:// protocol explicitly (again, like in most browsers):

>>> file_bytes_2 = grab('file://' + filepath)
>>> assert file_bytes_2 == file_bytes == b

If your filepath points to a folder, you'll get back a dol.Files object, which is a Mapping whose keys are file paths (relative to the folder) and the values are the bytes (acquired lazily) of the corresponding file's contents.

>>> import os, typing
>>> folder, filename = os.path.dirname(filepath), os.path.basename(filepath)
>>> f = grab(folder)
>>> isinstance(f, typing.Mapping)
True
>>> filename in f
True
>>> file_bytes_3 = f[filename]
>>> assert file_bytes_3 == file_bytes_2 == file_bytes == b

grab can handle various protocols, according to what packages it finds on your system. For example, if you have haggle (https://pypi.org/project/haggle/) installed, you'll find that kaggle is also a valid protocol.

>>> sorted(grab.dflt_protocols) # doctest: +SKIP
['file', 'http', 'https', 'kaggle']

Notice that grab has an argument called protocols. Yes, this means you have control. You just need to specify a mapping between protocol strings and the "url_to_content" function that should be used to get the content from that url.

In the following we'll add a fake foo protocol that doesn't really fetch any data, but applies str.upper to the url, but you get the point.

>>> from functools import partial
>>> from pyckup.base import dflt_protocols
>>> mygrab = partial(grab, protocols=dict(dflt_protocols, foo=str.upper))
>>> mygrab('foo://a_fake_url')
'FOO://A_FAKE_URL'

Examples

from pyckup import grab

See what (default) protocols you have access to.

from pyckup import grab

grab.procotols
# ['file', 'kaggle', 'http', 'https']

Grab file contents

Specifying a "file" protocol (i.e. prefixing your string with "file://" -- followed by a full path) will give you the contents of the file in bytes.

from pyckup import grab

b = grab('file:///Users/Thor.Whalen/Dropbox/dev/p3/proj/i/pyckup/pyckup/__init__.py')
assert isinstance(b, bytes)
print(b.decode())
# from pyckup.base import grab, protocols

But you can also use a full path, or other natural means of specifying files. In that case though, grab will try to give you the contents in a convenient type (e.g. a dict for .json, a python object of .pickle, string for .txt...). This is convenient, but don't depend on the type to strongly since it depends on what py2store.misc sets it to be.

from pyckup import grab

grab('/Users/Thor.Whalen/Dropbox/dev/p3/proj/i/pyckup/pyckup/__init__.py')
# b'from pyckup.base import grab, protocols\n\n\n'
grab('~/Dropbox/dev/p3/proj/i/pyckup/data/example.json')
# {'hello': 'world', 'abc': [1, 2, 3]}
grab('~/Dropbox/dev/p3/proj/i/pyckup/data/example.pickle')
# [1, 2, 3]
print(grab('~/Dropbox/dev/p3/proj/i/pyckup/data/example.txt'))
# This
# is
# text

Grab the contents of a url

from pyckup import grab

b = grab('https://raw.githubusercontent.com/i2mint/pyckup/master/LICENSE')
type(b), len(b)
# (bytes, 11357)
print(b[:100].decode())
#                                  Apache License
#                            Version 2.0, January 2004

Grab stuff from kaggle

Note: If you want to access kaggle datasets with pyckup, you'll need to get an account. See haggle for more information.

from pyckup import grab

z = grab('kaggle://drgilermo/face-images-with-marked-landmark-points')
list(z)
# ['face_images.npz', 'facial_keypoints.csv']
print(z['facial_keypoints.csv'][:100].decode())
# left_eye_center_x,left_eye_center_y,right_eye_center_x,right_eye_center_y,left_eye_inner_corner_x,le