Lib/base64.py

Source:

cpython 3.14 @ ab2d84fe1023/Lib/base64.py ↗

base64 provides RFC 4648 encoding/decoding functions. The core encoding uses the binascii C extension; the Python layer adds alphabet variants, line-length wrapping, padding handling, and the URL-safe alphabet.

Map

Lines	Symbol	Role
1-40	b64encode, b64decode	Standard Base64 with optional altchars
41-80	urlsafe_b64encode, urlsafe_b64decode	URL-safe alphabet (+/ replaced by -_)
81-140	b32encode, b32decode, b32hexencode, b32hexdecode	Base32 (RFC 4648 §6 and §7)
141-180	b16encode, b16decode	Hexadecimal encoding
181-240	b85encode, b85decode, a85encode, a85decode	Base85 and Ascii85
241-300	encodebytes, decodebytes	Line-wrapped encoding for MIME
301-360	Command-line interface	encode/decode subcommands

Reading

b64encode

# CPython: Lib/base64.py:54 b64encode
def b64encode(s, altchars=None):
    encoded = binascii.b2a_base64(s, newline=False)
    if altchars is not None:
        assert len(altchars) == 2, repr(altchars)
        return encoded.translate(bytes.maketrans(b'+/', altchars))
    return encoded

binascii.b2a_base64 does the actual encoding in C. The altchars parameter replaces + and / with two custom characters.

b64decode

# CPython: Lib/base64.py:72 b64decode
def b64decode(s, altchars=None, validate=False):
    s = _bytes_from_decode_data(s)
    if altchars is not None:
        altchars = _bytes_from_decode_data(altchars)
        assert len(altchars) == 2, repr(altchars)
        s = s.translate(bytes.maketrans(altchars, b'+/'))
    if validate and not re.fullmatch(b'[A-Za-z0-9+/]*={0,2}', s):
        raise binascii.Error('Non-base64 digit found')
    return binascii.a2b_base64(s)

validate=True checks that the input contains only valid Base64 characters before decoding.

b85encode

Base85 encodes every 4 bytes as 5 printable ASCII characters, giving a 25% overhead (vs 33% for Base64). Used by git for binary diffs and by Python's marshal for bytecode.

# CPython: Lib/base64.py:205 b85encode
_b85alphabet = (b"0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"
                b"abcdefghijklmnopqrstuvwxyz!#$%&()*+-;<=>?@^_`{|}~")

def b85encode(b, pad=False):
    ...
    for i in range(0, len(b), 4):
        chunk = b[i:i+4]
        ...
        acc = int.from_bytes(chunk, 'big')
        for j in range(4, -1, -1):
            chars[j] = _b85alphabet[acc % 85]
            acc //= 85

gopy notes

Status: not yet ported. Go's encoding/base64 covers standard, URL-safe, and raw variants. b32encode/b32decode map to encoding/base32. Base85 requires a pure Go implementation (available as encoding/ascii85 in the standard library, though the byte mapping differs from Python's b85).