src/queries/nflverse/github_data.py

import duckdb
import pandas as pd
import os
from typing import Callable

from domain.constants import SEASON


duckdb.default_connection.execute("SET GLOBAL pandas_analyze_sample=100000")

BASE_URL = "https://github.com/nflverse/nflverse-data/releases/download/"


FANTASY_POSITIONS = [
    "QB",
    "RB",
    "WR",
    "TE",
    "FB",
    "K",
]


def get_weekly_rosters(season_int: int | str = SEASON) -> pd.DataFrame:
    df = duckdb.sql(
        f"""
                    with grouped_stats as (
                    select player_id, sum(fantasy_points) as fantasy_points
                    from player_stats_player_stats_{season_int}
                    group by player_id
                    )
                    SELECT 
                    team
                    , wr.position
                    , wr.jersey_number
                    , wr.status
                    , wr.full_name
                    , wr.headshot_url
                    , wr.week
                    , wr.gsis_id
                    , gs.fantasy_points
                    from weekly_rosters_roster_weekly_{season_int} wr
                    left join grouped_stats gs
                    on wr.gsis_id = gs.player_id
"""
    ).df()
    return df


def get_player_stats(season_int: int | str = SEASON) -> pd.DataFrame:
    df = duckdb.sql(f"SELECT * from player_stats_player_stats_{season_int}").df()
    return df


NFLVERSE_ASSETS = [
    ("weekly_rosters", f"roster_weekly_{SEASON}.parquet"),
    ("player_stats", f"player_stats_{SEASON}.parquet"),
]


class NflVerseDataAsset:
    def __init__(
        self,
        release_tag: str,
        asset_name: str,
        dataframe_mutation_fxn: Callable[[pd.DataFrame], pd.DataFrame] = lambda x: x,
    ):
        self.release_tag = release_tag
        self.asset_name = asset_name
        self.dataframe_mutation_fxn = dataframe_mutation_fxn
        self.table_name = f"{release_tag}_{asset_name.rsplit('.', 1)[0]}"

    def load_parquet_asset_to_df(self) -> pd.DataFrame:
        location = os.path.join(BASE_URL, self.release_tag, self.asset_name)
        df = pd.read_parquet(location)
        return df

    def register_asset_to_duckdb(self) -> None:
        df = self.load_parquet_asset_to_df()
        df = self.dataframe_mutation_fxn(df)
        duckdb.register(self.table_name, df)


def load_assets():
    for tag, asset in NFLVERSE_ASSETS:
        asset = NflVerseDataAsset(tag, asset)
        asset.register_asset_to_duckdb()


def get_current_tables() -> list[str]:
    current_tables_df = duckdb.sql("SHOW TABLES").df()
    return current_tables_df["name"].tolist()


def load_assets_if_no_tables():
    if not len(get_current_tables()):
        load_assets()