scripts/tools_metrics.py

import pandas as pd
from utils import TMP_DIR, INC_TOOLS, ROOT_DIR


def get_error_data_by_market(tools_df: pd.DataFrame) -> pd.DataFrame:
    """Gets the error data for the given tools and calculates the error percentage."""
    mech_tool_errors = tools_df[tools_df["error"] != -1]
    error = (
        mech_tool_errors.groupby(
            ["tool", "request_month_year_week", "market_creator", "error"], sort=False
        )
        .size()
        .unstack()
        .fillna(0)
        .reset_index()
    )
    error["error_perc"] = (error[1] / (error[0] + error[1])) * 100
    error["total_requests"] = error[0] + error[1]
    return error


def get_tool_winning_rate_by_market(tools_df: pd.DataFrame) -> pd.DataFrame:
    """Gets the tool winning rate data for the given tools by market and calculates the winning percentage."""
    tools_non_error = tools_df[tools_df["error"] == 0]
    tools_non_error.loc[:, "currentAnswer"] = tools_non_error["currentAnswer"].replace(
        {"no": "No", "yes": "Yes"}
    )
    tools_non_error = tools_non_error[
        tools_non_error["currentAnswer"].isin(["Yes", "No"])
    ]
    tools_non_error = tools_non_error[tools_non_error["vote"].isin(["Yes", "No"])]
    tools_non_error["win"] = (
        tools_non_error["currentAnswer"] == tools_non_error["vote"]
    ).astype(int)
    tools_non_error.columns = tools_non_error.columns.astype(str)
    wins = (
        tools_non_error.groupby(
            ["tool", "request_month_year_week", "market_creator", "win"], sort=False
        )
        .size()
        .unstack()
        .fillna(0)
    )
    wins["win_perc"] = (wins[1] / (wins[0] + wins[1])) * 100
    wins.reset_index(inplace=True)
    wins["total_request"] = wins[0] + wins[1]
    wins.columns = wins.columns.astype(str)
    # Convert request_month_year_week to string and explicitly set type for Altair
    # wins["request_month_year_week"] = wins["request_month_year_week"].astype(str)
    return wins


def prepare_tools(tools: pd.DataFrame, total_included: bool = True) -> pd.DataFrame:
    # remove non relevant tools
    tools = tools[tools["tool"].isin(INC_TOOLS)]
    tools["request_time"] = pd.to_datetime(tools["request_time"], utc=True)
    tools = tools.sort_values(by="request_time", ascending=True)
    tools["request_date"] = tools["request_time"].dt.date
    tools["request_month_year_week"] = (
        pd.to_datetime(tools["request_time"])
        .dt.to_period("W")
        .dt.start_time.dt.strftime("%b-%d-%Y")
    )
    # preparing the tools graph
    if total_included:
        # adding the total
        tools_all = tools.copy(deep=True)
        tools_all["market_creator"] = "all"
        # merging both dataframes
        tools = pd.concat([tools, tools_all], ignore_index=True)
    tools = tools.sort_values(by="request_time", ascending=True)
    return tools


def get_error_category(error_value: int):
    if error_value == 0:
        return "non_error"
    if error_value == 1:
        return "tool_error"
    return "request_error"


def get_errors_by_mech_address(tools_df: pd.DataFrame) -> pd.DataFrame:
    """Gets the tool errors distribution by mech address in a weekly fashion"""

    weekly_errors = (
        tools_df.groupby(
            ["request_month_year_week", "mech_address", "error"], sort=False
        )
        .size()
        .reset_index(name="requests")
    )
    weekly_errors["error_cat"] = weekly_errors["error"].apply(
        lambda x: get_error_category(x)
    )
    total_requests_errors = (
        tools_df.groupby(["request_month_year_week", "mech_address"], sort=False)
        .size()
        .reset_index(name="total_requests")
    )
    all_errors = weekly_errors.merge(
        total_requests_errors, on=["request_month_year_week", "mech_address"]
    )
    all_errors["errors_percentage"] = (
        all_errors["requests"] / all_errors["total_requests"]
    ) * 100
    return all_errors


def compute_tools_based_datasets():
    print("Computing tools based datasets")
    try:
        tools_df = pd.read_parquet(TMP_DIR / "tools.parquet")
    except Exception as e:
        print(f"Error reading tools parquet file {e}")
        return None

    # mech tool errors by markets
    print("Computing mech tool errors by markets")
    tool_error_by_markets = get_error_data_by_market(tools_df=prepare_tools(tools_df))
    tool_error_by_markets.to_parquet(ROOT_DIR / "error_by_markets.parquet", index=False)
    try:
        tools_df = pd.read_parquet(TMP_DIR / "tools.parquet")
        tools_df = prepare_tools(tools_df)
    except Exception as e:
        print(f"Error reading tools parquet file {e}")
        return None
    winning_df = get_tool_winning_rate_by_market(tools_df)
    winning_df.to_parquet(ROOT_DIR / "winning_df.parquet", index=False)

    # all errors by mech address
    try:
        tools_df = pd.read_parquet(TMP_DIR / "tools.parquet")
        tools_df = prepare_tools(tools_df, total_included=False)
    except Exception as e:
        print(f"Error reading tools parquet file {e}")
        return None
    errors_by_mech = get_errors_by_mech_address(tools_df=tools_df)
    errors_by_mech.to_parquet(ROOT_DIR / "errors_by_mech.parquet", index=False)

    try:
        tools_df = pd.read_parquet(TMP_DIR / "tools.parquet")
        tools_df = prepare_tools(tools_df, total_included=False)
    except Exception as e:
        print(f"Error reading tools parquet file {e}")
        return None
    generate_daily_mech_requests_per_tool(tools_df=tools_df)
    generate_daily_mech_request_for_pearl_agents(tools_df=tools_df)


def generate_daily_mech_requests_per_tool(tools_df: pd.DataFrame) -> None:
    """Generates the daily mech requests per tool."""

    # daily mech requests in
    daily_mech_req_per_tool = (
        tools_df.groupby(["request_date", "tool", "market_creator"])["request_id"]
        .count()
        .reset_index(name="total_mech_requests")
    )
    daily_mech_req_per_tool.to_parquet(
        ROOT_DIR / "daily_mech_requests.parquet", index=False
    )


def generate_daily_mech_request_for_pearl_agents(tools_df: pd.DataFrame) -> None:
    # read the prediction agents file
    pearl_agents = pd.read_parquet(ROOT_DIR / "pearl_agents.parquet")
    unique_addresses = pearl_agents["safe_address"].unique()

    # filter tools for only traders from the list above
    selected_tools_df = tools_df[
        tools_df["trader_address"].isin(unique_addresses)
    ].copy()
    daily_mech_req_per_tool = (
        selected_tools_df.groupby(["request_date", "tool"])["request_id"]
        .count()
        .reset_index(name="total_mech_requests")
    )
    daily_mech_req_per_tool.to_parquet(
        ROOT_DIR / "daily_mech_requests_by_pearl_agents.parquet", index=False
    )


if __name__ == "__main__":
    compute_tools_based_datasets()