Answer a question

I am tying to bulk insert a dataframe to my postgres dB. Some columns in my dataframe are date types with NaT as a null value. Which is not supported by PostgreSQL, I've tried to replace NaT (using pandas) with other NULL type identifies but that did not work during my inserts.

I used df = df.where(pd.notnull(df), 'None') to replace all the NaTs, Example of errors that keep coming up due to datatype issues.

Error: invalid input syntax for type date: "None"
LINE 1: ...0,1.68757,'2022-11-30T00:29:59.679000'::timestamp,'None','20...

My driver and insert statement to postgresql dB:

def execute_values(conn, df, table):
    """
    Using psycopg2.extras.execute_values() to insert the dataframe
    """
    # Create a list of tupples from the dataframe values
    tuples = [tuple(x) for x in df.to_numpy()]
    # Comma-separated dataframe columns
    cols = ','.join(list(df.columns))
    # SQL quert to execute
    query  = "INSERT INTO %s(%s) VALUES %%s" % (table, cols)
    cursor = conn.cursor()
    try:
        extras.execute_values(cursor, query, tuples)
        conn.commit()
    except (Exception, psycopg2.DatabaseError) as error:
        print("Error: %s" % error)
        conn.rollback()
        cursor.close()
        return 1
    print("execute_values() done")
    cursor.close()

Info about my dataframe: for this case the culprits are the datetime columns only.

how is this commonly solved?

Answers

You're re-inventing the wheel. Just use pandas' to_sql method and it will

• match up the column names, and
• take care of the NaT values.

Use method="multi" to give you the same effect as psycopg2's execute_values.

from pprint import pprint

import pandas as pd
import sqlalchemy as sa

table_name = "so64435497"
engine = sa.create_engine("postgresql://scott:tiger@192.168.0.199/test")
with engine.begin() as conn:
    # set up test environment
    conn.exec_driver_sql(f"DROP TABLE IF EXISTS {table_name}")
    conn.exec_driver_sql(
        f"CREATE TABLE {table_name} ("
        "id integer PRIMARY KEY GENERATED ALWAYS AS IDENTITY, "
        "txt varchar(50), "
        "txt2 varchar(50), "
        "dt timestamp)"
    )
    df = pd.read_csv(r"C:\Users\Gord\Desktop\so64435497.csv")
    df["dt"] = pd.to_datetime(df["dt"])
    print(df)
    """console output:
                       dt  txt2  txt
    0 2020-01-01 00:00:00  foo2  foo
    1                 NaT  bar2  bar
    2 2020-01-02 03:04:05  baz2  baz
    """

    # run test
    df.to_sql(
        table_name, conn, index=False, if_exists="append", method="multi"
    )
    pprint(
        conn.exec_driver_sql(
            f"SELECT id, txt, txt2, dt FROM {table_name}"
        ).all()
    )
    """console output:
    [(1, 'foo', 'foo2', datetime.datetime(2020, 1, 1, 0, 0)),
     (2, 'baz', 'baz2', None),
     (3, 'bar', 'bar2', datetime.datetime(2020, 1, 2, 3, 4, 5))]
    """