Source code for fastf1.utils

"""This is a collection of various functions."""
import datetime
import warnings
from functools import reduce
from typing import (
    Optional,
    Union
)

import numpy as np
import pandas as pd

import fastf1
from fastf1.logger import get_logger


_logger = get_logger(__name__)


[docs] def delta_time( reference_lap: "fastf1.core.Lap", compare_lap: "fastf1.core.Lap" ) -> tuple[pd.Series, "fastf1.core.Telemetry", "fastf1.core.Telemetry"]: """Calculates the delta time of a given lap, along the 'Distance' axis of the reference lap. .. deprecated:: 3.0.0 .. warning:: This function should no longer be considered as a stable part of the API. Due to the reasons given below, this function will be modified or removed at a future point. .. warning:: This is a nice gimmick but not actually very accurate which is an inherent problem from the way this is calculated currently (There may not be a better way though). In comparison with the sector times and the differences that can be calculated from these, there are notable differences! You should always verify the result against sector time differences or find a different way for verification. Here is an example that compares the quickest laps of Leclerc and Hamilton from Bahrain 2021 Qualifying: .. plot:: :include-source: import fastf1 as ff1 from fastf1 import plotting from fastf1 import utils from matplotlib import pyplot as plt plotting.setup_mpl(misc_mpl_mods=False, color_scheme='fastf1') session = ff1.get_session(2021, 'Emilia Romagna', 'Q') session.load() lec = session.laps.pick_driver('LEC').pick_fastest() ham = session.laps.pick_driver('HAM').pick_fastest() delta_time, ref_tel, compare_tel = utils.delta_time(ham, lec) # ham is reference, lec is compared fig, ax = plt.subplots() # use telemetry returned by .delta_time for best accuracy, # this ensures the same applied interpolation and resampling ax.plot(ref_tel['Distance'], ref_tel['Speed'], color=plotting.get_team_color(ham['Team'], session)) ax.plot(compare_tel['Distance'], compare_tel['Speed'], color=plotting.get_team_color(lec['Team'], session)) twin = ax.twinx() twin.plot(ref_tel['Distance'], delta_time, '--', color='white') twin.set_ylabel("<-- Lec ahead | Ham ahead -->") plt.show() Args: reference_lap: The lap taken as reference compare_lap: The lap to compare Returns: A tuple containing - pd.Series of type `float64` with the delta in seconds. - :class:`~fastf1.core.Telemetry` for the reference lap - :class:`~fastf1.core.Telemetry` for the comparison lap Use the return telemetry for plotting to make sure you have telemetry data that was created with the same interpolation and resampling options! """ warnings.warn("`utils.delta_time` is considered deprecated and will" "be modified or removed in a future release because it has" "a tendency to give inaccurate results.", FutureWarning) ref = reference_lap.get_car_data(interpolate_edges=True).add_distance() comp = compare_lap.get_car_data(interpolate_edges=True).add_distance() def mini_pro(stream): # Ensure that all samples are interpolated dstream_start = stream[1] - stream[0] dstream_end = stream[-1] - stream[-2] return np.concatenate( [[stream[0] - dstream_start], stream, [stream[-1] + dstream_end]] ) ltime = mini_pro(comp['Time'].dt.total_seconds().to_numpy()) multiplier = ref.Distance.iat[-1]/comp.Distance.iat[-1] ldistance = mini_pro(comp['Distance'].to_numpy())*multiplier lap_time = np.interp(ref['Distance'], ldistance, ltime) delta = lap_time - ref['Time'].dt.total_seconds() return delta, ref, comp
[docs] def recursive_dict_get(d: dict, *keys: str, default_none: bool = False): """Recursive dict get. Can take an arbitrary number of keys and returns an empty dict if any key does not exist. https://stackoverflow.com/a/28225747""" ret = reduce(lambda c, k: c.get(k, {}), keys, d) if default_none and ret == {}: return None else: return ret
[docs] def to_timedelta(x: Union[str, datetime.timedelta]) \ -> Optional[datetime.timedelta]: """Fast timedelta object creation from a time string Permissible string formats: For example: `13:24:46.320215` with: - optional hours and minutes - optional microseconds and milliseconds with arbitrary precision (1 to 6 digits) Examples of valid formats: - `24.3564` (seconds + milli/microseconds) - `36:54` (minutes + seconds) - `8:45:46` (hours, minutes, seconds) Args: x: timestamp """ # this is faster than using pd.timedelta on a string if isinstance(x, str) and len(x): try: hours, minutes = 0, 0 if len(hms := x.split(':')) == 3: hours, minutes, seconds = hms elif len(hms) == 2: minutes, seconds = hms else: seconds = hms[0] if '.' in seconds: seconds, msus = seconds.split('.') if len(msus) < 6: msus = msus + '0' * (6 - len(msus)) elif len(msus) > 6: msus = msus[0:6] else: msus = 0 return datetime.timedelta( hours=int(hours), minutes=int(minutes), seconds=int(seconds), microseconds=int(msus) ) except Exception as exc: _logger.debug(f"Failed to parse timedelta string '{x}'", exc_info=exc) return None elif isinstance(x, datetime.timedelta): return x else: return None
[docs] def to_datetime(x: Union[str, datetime.datetime]) \ -> Optional[datetime.datetime]: """Fast datetime object creation from a date string. Permissible string formats: For example '2020-12-13T13:27:15.320000Z' with: - optional milliseconds and microseconds with arbitrary precision (1 to 6 digits) - with optional trailing letter 'Z' Examples of valid formats: - `2020-12-13T13:27:15.320000` - `2020-12-13T13:27:15.32Z` - `2020-12-13T13:27:15` Args: x: timestamp """ if isinstance(x, str) and x: try: date, time = x.strip('Z').split('T') year, month, day = date.split('-') hours, minutes, seconds = time.split(':') if '.' in seconds: seconds, msus = seconds.split('.') if len(msus) < 6: msus = msus + '0' * (6 - len(msus)) elif len(msus) > 6: msus = msus[0:6] else: msus = 0 return datetime.datetime( int(year), int(month), int(day), int(hours), int(minutes), int(seconds), int(msus) ) except Exception as exc: _logger.debug(f"Failed to parse datetime string '{x}'", exc_info=exc) return None elif isinstance(x, datetime.datetime): return x else: return None