HPI/my/time/tz/via_location.py

299 lines
10 KiB
Python

'''
Timezone data provider, guesses timezone based on location data (e.g. GPS)
'''
REQUIRES = [
# for determining timezone by coordinate
'timezonefinder',
]
## user might not have tz config section, so makes sense to be more defensive about it
# todo might be useful to extract a helper for this
try:
from my.config import time
except ImportError as ie:
if ie.name != 'time':
raise ie
else:
try:
user_config = time.tz.via_location
except AttributeError as ae:
if not ("'tz'" in str(ae) or "'via_location'"):
raise ae
# deliberately dynamic to prevent confusing mypy
if 'user_config' not in globals():
globals()['user_config'] = object
##
from my.core import dataclass
@dataclass
class config(user_config):
# less precise, but faster
fast: bool = True
# sort locations by date
# in case multiple sources provide them out of order
sort_locations: bool = True
# if the accuracy for the location is more than 5km, don't use
require_accuracy: float = 5_000
# how often (hours) to refresh the cachew timezone cache
# this may be removed in the future if we opt for dict-based caching
_iter_tz_refresh_time: int = 6
from collections import Counter
from datetime import date, datetime
from functools import lru_cache
from itertools import groupby
from typing import Iterator, NamedTuple, Optional, Tuple, Any, List, Iterable, Set
import heapq
import pytz
from more_itertools import seekable
from my.core.common import LazyLogger, mcachew, tzdatetime
from my.core.source import import_source
logger = LazyLogger(__name__, level='warning')
@lru_cache(2)
def _timezone_finder(fast: bool) -> Any:
if fast:
# less precise, but faster
from timezonefinder import TimezoneFinderL as Finder
else:
from timezonefinder import TimezoneFinder as Finder # type: ignore
return Finder(in_memory=True)
# todo move to common?
Zone = str
# NOTE: for now only daily resolution is supported... later will implement something more efficient
class DayWithZone(NamedTuple):
day: date
zone: Zone
from my.location.common import LatLon
# for backwards compatibility
def _locations() -> Iterator[Tuple[LatLon, datetime]]:
try:
import my.location.all
for loc in my.location.all.locations():
if loc.accuracy is not None and loc.accuracy > config.require_accuracy:
continue
yield ((loc.lat, loc.lon), loc.dt)
except Exception as e:
from my.core.warnings import high
logger.exception("Could not setup via_location using my.location.all provider, falling back to legacy google implementation", exc_info=e)
high("Setup my.google.takeout.parser, then my.location.all for better google takeout/location data")
import my.location.google
for gloc in my.location.google.locations():
yield ((gloc.lat, gloc.lon), gloc.dt)
# TODO: could use heapmerge or sort the underlying iterators somehow?
# see https://github.com/karlicoss/HPI/pull/237#discussion_r858372934
def _sorted_locations() -> List[Tuple[LatLon, datetime]]:
return list(sorted(_locations(), key=lambda x: x[1]))
def _find_tz_for_locs(finder: Any, locs: Iterable[Tuple[LatLon, datetime]]) -> Iterator[DayWithZone]:
for (lat, lon), dt in locs:
# TODO right. its _very_ slow...
zone = finder.timezone_at(lat=lat, lng=lon)
# todo allow to skip if not noo many errors in row?
if zone is None:
# warnings.append(f"Couldn't figure out tz for {lat}, {lon}")
continue
tz = pytz.timezone(zone)
# TODO this is probably a bit expensive... test & benchmark
ldt = dt.astimezone(tz)
ndate = ldt.date()
#if pdt is not None and ndate < pdt.date():
# # TODO for now just drop and collect the stats
# # I guess we'd have minor drops while air travel...
# warnings.append("local time goes backwards {ldt} ({tz}) < {pdt}")
# continue
#pdt = ldt
z = tz.zone; assert z is not None
yield DayWithZone(day=ndate, zone=z)
# Note: this takes a while, as the upstream since _locations isn't sorted, so this
# has to do an iterative sort of the entire my.locations.all list
def _iter_local_dates() -> Iterator[DayWithZone]:
finder = _timezone_finder(fast=config.fast) # rely on the default
#pdt = None
# TODO: warnings doesn't actually warn?
# warnings = []
locs: Iterable[Tuple[LatLon, datetime]]
locs = _sorted_locations() if config.sort_locations else _locations()
yield from _find_tz_for_locs(finder, locs)
# my.location.fallback.estimate_location could be used here
# but iterating through all the locations is faster since this
# is saved behind cachew
@import_source(module_name="my.location.fallback.all")
def _iter_local_dates_fallback() -> Iterator[DayWithZone]:
from my.location.fallback.all import fallback_locations as flocs
def _fallback_locations() -> Iterator[Tuple[LatLon, datetime]]:
for loc in sorted(flocs(), key=lambda x: x.dt):
yield ((loc.lat, loc.lon), loc.dt)
yield from _find_tz_for_locs(_timezone_finder(fast=config.fast), _fallback_locations())
def most_common(lst: List[DayWithZone]) -> DayWithZone:
res, _ = Counter(lst).most_common(1)[0]
return res
def _iter_tz_depends_on() -> str:
"""
Since you might get new data which specifies a new timezone sometime
in the day, this causes _iter_tzs to refresh every _iter_tz_refresh_time hours
(default 6), like:
2022-04-26_00
2022-04-26_06
2022-04-26_12
2022-04-26_18
"""
mod = config._iter_tz_refresh_time
assert mod >= 1
day = str(date.today())
hr = datetime.now().hour
hr_truncated = hr // mod * mod
return "{}_{}".format(day, hr_truncated)
# refresh _iter_tzs every few hours -- don't think a better depends_on is possible dynamically
@mcachew(logger=logger, depends_on=_iter_tz_depends_on)
def _iter_tzs() -> Iterator[DayWithZone]:
# since we have no control over what order the locations are returned,
# we need to sort them first before we can do a groupby
local_dates: List[DayWithZone] = list(_iter_local_dates())
local_dates.sort(key=lambda p: p.day)
logger.debug(f"no. of items using exact locations: {len(local_dates)}")
local_dates_fallback: List[DayWithZone] = list(_iter_local_dates_fallback())
local_dates_fallback.sort(key=lambda p: p.day)
# find days that are in fallback but not in local_dates (i.e., missing days)
local_dates_set: Set[date] = set(d.day for d in local_dates)
use_fallback_days: List[DayWithZone] = [d for d in local_dates_fallback if d.day not in local_dates_set]
logger.debug(f"no. of items being used from fallback locations: {len(use_fallback_days)}")
# combine local_dates and missing days from fallback into a sorted list
all_dates = heapq.merge(local_dates, use_fallback_days, key=lambda p: p.day)
for d, gr in groupby(all_dates, key=lambda p: p.day):
logger.info(f"processed {d}{', using fallback' if d in local_dates_set else ''}")
zone = most_common(list(gr)).zone
yield DayWithZone(day=d, zone=zone)
@lru_cache(1)
def loc_tz_getter() -> Iterator[DayWithZone]:
# seekable makes it cache the emitted values
return seekable(_iter_tzs())
# todo expose zone names too?
@lru_cache(maxsize=None)
def _get_day_tz(d: date) -> Optional[pytz.BaseTzInfo]:
sit = loc_tz_getter()
# todo hmm. seeking is not super efficient... might need to use some smarter dict-based cache
# hopefully, this method itself caches stuff forthe users, so won't be too bad
sit.seek(0) # type: ignore
zone: Optional[str] = None
for x, tz in sit:
if x == d:
zone = tz
if x >= d:
break
return None if zone is None else pytz.timezone(zone)
# ok to cache, there are only a few home locations?
@lru_cache(maxsize=None)
def _get_home_tz(loc: LatLon) -> Optional[pytz.BaseTzInfo]:
(lat, lng) = loc
finder = _timezone_finder(fast=False) # ok to use slow here for better precision
zone = finder.timezone_at(lat=lat, lng=lng)
if zone is None:
# TODO shouldn't really happen, warn?
return None
else:
return pytz.timezone(zone)
def _get_tz(dt: datetime) -> Optional[pytz.BaseTzInfo]:
'''
Given a datetime, returns the timezone for that date.
'''
res = _get_day_tz(d=dt.date())
if res is not None:
return res
# fallback to home tz
# note: the fallback to fallback.via_home.estimate_location is still needed, since
# _iter_local_dates_fallback only returns days which we actually have a datetime for
# (e.g. there was an IP address within a day of that datetime)
#
# given a datetime, fallback.via_home.estimate_location will find which home location
# that datetime is between, else fallback on your first home location, so it acts
# as a last resort
from my.location.fallback import via_home as home
loc = list(home.estimate_location(dt))
assert len(loc) == 1, f"should only have one home location, received {loc}"
return _get_home_tz(loc=(loc[0].lat, loc[0].lon))
# expose as 'public' function
get_tz = _get_tz
def localize(dt: datetime) -> tzdatetime:
tz = _get_tz(dt)
if tz is None:
# TODO -- this shouldn't really happen.. think about it carefully later
return dt
else:
return tz.localize(dt)
from ...core import stat, Stats
def stats(quick: bool=False) -> Stats:
if quick:
prev, config.sort_locations = config.sort_locations, False
res = {
'first': next(_iter_local_dates())
}
config.sort_locations = prev
return res
# TODO not sure what would be a good stat() for this module...
# might be nice to print some actual timezones?
# there aren't really any great iterables to expose
import os
VIA_LOCATION_START_YEAR = int(os.environ.get("VIA_LOCATION_START_YEAR", 1990))
def localized_years():
last = datetime.now().year + 2
# note: deliberately take + 2 years, so the iterator exhausts. otherwise stuff might never get cached
# need to think about it...
for Y in range(VIA_LOCATION_START_YEAR, last):
dt = datetime.fromisoformat(f'{Y}-01-01 01:01:01')
yield localize(dt)
return stat(localized_years)