Testing speeds of different methods of parsing datetime strings (of a specific, similar but non ISO format) in python

regex.py

from datetime import datetime
from dateutil.tz import tzoffset, tzutc

import re

THE_RE = re.compile(
    "^"
    "([0-9]{4})"
    "-"
    "([0-9]{2})"
    "-"
    "([0-9]){2}"
    " "
    "([0-9]{2})"
    ":"
    "([0-9]{2})"
    ":"
    "([0-9]{2})"
    "\+"
    "([0-9]{2})"
    "([0-9]{2})"
    "$"
)
def parse(dt_str):
    # 2016-08-18 18:08:05+0000
    vals = THE_RE.match(dt_str).groups()
    vals = list(int(i) for i in vals)

    dt = datetime(*vals[:6])

    tz_vals = vals[6:]

    if any(tz_vals):
        tzinfo_ = tzoffset(None, (60*tz_vals[0])+tz_vals[1])
    else:
        tzinfo_ = tzutc()

    return dt.replace(tzinfo=tzinfo_)


if __name__ == '__main__':
    try:
        x = xrange
    except NameError:
        x = range
    for i in x(100000):
        parse('2016-08-18 18:08:05+0000')
        parse('2016-08-18 18:08:05+0100')

results.txt

dateutil.parser:  106, 107,
myre:  6.84, 8.85,
splitter:  5.04, 6.61,
strptime:  15.7, 17.7,

splitter.py

from datetime import datetime
from dateutil.tz import tzoffset, tzutc

# 2016-08-18 18:08:05+0000

DT_SLICES = (
    slice(None,4),
    slice(5,7),
    slice(8,10),
    slice(11,13),
    slice(14,16),
    slice(17,19),
)

TZ_HR_SLICE = slice(20,22)
TZ_MN_SLICE = slice(22,None)

def parse(dt_str):
    tz_hr = int(dt_str[TZ_HR_SLICE])
    tz_mn = int(dt_str[TZ_MN_SLICE])

    if tz_hr == tz_mn == 0:
        tzinfo_ = tzutc()
    else:
        tzinfo_ = tzoffset(None, (60*tz_hr)+tz_mn)

    return datetime(*(int(dt_str[slice_]) for slice_ in DT_SLICES), tzinfo=tzinfo_)


if __name__ == '__main__':
    try:
        x = xrange
    except NameError:
        x = range
    for i in x(100000):
        parse('2016-08-18 18:08:05+0000')
        parse('2016-08-18 18:08:05+0100')

strptime.py

from datetime import datetime, tzinfo
from dateutil.tz import tzoffset, tzutc

def parse(dt_str):
    # 2016-08-18 18:08:05+0000
    dt, tz_hr, tz_mn = dt_str[:19], dt_str[20:22], dt_str[22:]

    tz_hr = int(tz_hr)
    tz_mn = int(tz_mn)
    if tz_hr == tz_mn == 0:
        tzinfo_ = tzutc()
    else:
        tzinfo_ = tzoffset(None, (60*tz_hr)+tz_mn)

    dt = datetime.strptime(dt, '%Y-%m-%d %H:%M:%S')
    return dt.replace(tzinfo=tzinfo_)


if __name__ == '__main__':
    try:
        x = xrange
    except NameError:
        x = range
    for i in x(100000):
        parse('2016-08-18 18:08:05+0000')
        parse('2016-08-18 18:08:05+0100')

tester.sh

for module in dateutil.parser $(ls *.py | cut -d '.' -f1)
do
    printf '%s: ' $module;
    for datetime in "2016-08-18 18:08:05+0000" "2016-08-18 18:08:05+0100";
    do
        python -m timeit --setup "from $module import parse" "parse('${datetime}')" | \
            grep -Po '(?<=: )[0-9\.]+(?= usec per loop)' | xargs printf ' %s,';
    done;
    echo;
done