def execute_matching()

in src/engine/step4/model_dev/utils/metrics.py [0:0]

• 28 lines of code
• 6 McCabe index (conditional complexity)

def execute_matching(tool, row, entities):
    """
    Execute the true and predicted query of a give row.

    Args:
        tool(nlq2SqlTool):  Tool with the functions for the query is be rendered and executed.
        row: Row for a single query.
        entities(dict): Dictionary of specific values for the query arguments.

    Returns:
        True if the inferred query can be executed and returns the same as the true query. Otherwise, returns False.
    """

    base_question = row["base_question"]
    unfolded_question = row["unfolded_questions"]
    true_query = row["query"]
    pred_query = row["preds_wiki"]
    indx = int(row.name) + 1

    # Render & execute targets
    true_sql_query = tool.render_template_query(true_query, entities)
    true_output = tool.execute_sql_query(true_sql_query)
    #     print(true_output.iloc[:5])

    # Render & execute predictions
    try:
        pred_sql_query = tool.render_template_query(pred_query, entities)
        pred_output = tool.execute_sql_query(pred_sql_query)
    except:
        pred_output = None

    if pred_output is not None:
        # Fix the reordering of the output columns
        true_columns = true_output.columns.tolist()
        pred_columns = pred_output.columns.tolist()

        if set(true_columns) == set(pred_columns):
            pred_output = pred_output[true_columns]

    if pred_output is None:
        return 0

    true_first_c = true_output.columns[0]
    pred_first_c = pred_output.columns[0]

    true_output0 = true_output.sort_values(by=true_first_c).values
    pred_output0 = pred_output.sort_values(by=pred_first_c).values

    if true_query == pred_query or np.array_equal(true_output0, pred_output0):
        return 1.0
    else:
        return 0.0