Backport to Python 2.7?

[ahawker]

Here are some initial thoughts but definitely incomplete list of changes necessary.

• Switch hard coded bytes to be configurable to str.
• Loss of int.to_bytes() and int.from_bytes().
• Loss of datetime.timestamp()
• Differences between memoryview and buffer?

[blakev]

I'm almost done with this, ~20 more errors to squash in the test suite, mostly with handling bytes vs bytearray in python2

[nicktimko]

@blakev what are you using to strip out all the function annotations; something automated or just discarding them all?

[blakev]

I did it by hand because there weren't too many.

This was the recommended way:

def decode(value):
    ...
    ...
decode.__annotations__ = {'value': str, 'return': bytes_}

There's also a ._compat.py file I'm working on to add missing functionality between the two versions:

#! /usr/bin/env python
# -*- coding: utf-8 -*-
# >>
#     ulid, 2017
# <<
from __future__ import division  # enable true division
import sys
import struct
import datetime
import binascii

PY2 = sys.version_info[0] == 2
MAX_INT = (256**8)-1  # unsigned long long
int_ = None           # deferred type

if PY2:
    bytes_ = bytearray
    string_types = (basestring,)

else:
    int_ = int
    bytes_ = bytes
    string_types = (str,)


class Int(object):
    def __init__(self, x='0', base=10):
        self.x = x
        self.base = base

    def __int__(self):
        return int(self.x)

    def to_bytes(self, length, byteorder='big'):
        try:
            ch = '>' if byteorder.startswith('b') else '<'
            hex_str = '%x' % int(self)
            n = len(hex_str)
            x = binascii.unhexlify(hex_str.zfill(n + (n & 1)))
            val = x[-1 * length:] if ch == '>' else x[:length]
        except OverflowError:
            raise ValueError(self)
        return val
    to_bytes.__annotations__ = {'length': int, 'byteorder': str, 'return': bytes_}

    @classmethod
    def from_bytes(cls, b, byteorder='big'):
        # type: (bytes, str) -> Int
        val = bytes_(memoryview(b).tobytes())
        byteorder = byteorder.lower()
        ch = '>' if byteorder.startswith('b') else '<'
        hi = struct.unpack(ch + 'B'*len(val), val)
        xo = 0 + sum(n**(i+1) for i, n in enumerate(hi))
        return xo

# 8670479856763644750035483567436052040L
# 300150790306154394174966339365723675170L

if PY2:
    int_ = Int


def to_timestamp(dt):
    if hasattr(dt, 'timestamp'):
        return dt.timestamp()
    utc_native = dt.replace(tzinfo=None) - dt.utcoffset()
    timestamp = (utc_native - datetime.datetime(1970, 1, 1)).total_seconds()
    return int_(timestamp)
to_timestamp.__annotations__ = {'dt': datetime.datetime, 'return': int_}

[ahawker]

Isn't there a backport of typing to 2.7? Or am I mis-remembering?

[blakev]

It doesn't really work like you'd expect. The two recommended formats are the func.__annotations__ style or doing a single comment before the docstring:

    @classmethod
    def from_bytes(cls, b, byteorder='big'):
        # type: (bytes, str) -> Int
        """ My function method turns these bytes to a number! """

classmethod's didn't seem to work well with the __annotations__ style so I used # type: instead in those cases.

PyCharm and other tools will pick up on these.

[ahawker]

Closing as py27 is being EOL'd very soon.