The problem

urlsafe_b64encode and urlsafe_b64decode from base64 are not enough because they leave = used for padding chars unquoted:

import base64

base64.urlsafe_b64encode('a')
'YQ=='

And there are 2 problems with that

1. = sign gets quoted:

import urllib

urllib.quote('=')
'%3D'

2. Some libraries tolerate the = in query string values:

from urlparse import urlsplit, parse_qs

parse_qs(urlsplit('http://aaa.com/asa?q=AAAA=BBBB=CCCC').query)
{'q': ['AAAA=BBBB=CCCC']}

but the RFC 3986 underspecifies the query string so we cannot rely on = chars being handled by all web applications as it is done by urlparse.

Therefore we consider chars: [‘+’, ‘/’, ‘=’] unsafe and we replace them with [‘-’, ‘_’, ‘.’]. Characters + and / are already handled by urlsafe_* functions from base64 so only = is left. Since the = is used exclusively for padding, we simply remove it, and re-attach the padding during decoding. Because of that, querystringsafe_base64 is able to decode padded and unpadded string.

The solution

import querystringsafe_base64

querystringsafe_base64.encode(b'foo-bar')
b'Zm9vLWJhcg'

querystringsafe_base64.decode(b'Zm9vLWJhcg..')
b'foo-bar'

querystringsafe_base64.decode(b'Zm9vLWJhcg')
b'foo-bar'

1.2.0

• Remove padding characters from encoded string.

1.1.1

• Fixed packaging

1.1.0

• Always expect bytes

• Add type annotations

1.0.0

• support for restore missing padding during decode process

0.2.0

• Support for python3

0.1.5

• Move querystringsafe_base64 module to the root

• Use install instead of develop during tests

0.1.4

• Remove bdist_wheel from distributons

0.1.3

• Install pandoc (travis)

0.1.2

• Add setup.cfg and pypandoc to tests

0.1.1

• add MANIFEST.in file

0.1.0

• package structure

• tests