How do I write a header row using csv.writer when using a sqlalchemy cursor?

Question

The questions stems from my previous question here . By switching from using pandas to run the query and output to csv to using sqlalchemy and csv.writer, I've lost the header row.

I apologize for the example code, as it's only going to be runnable if you have a SQL Server. Running this code, you get the output "hello world|goodbye world" and the headers (a and b, respectively) are missing.

from sqlalchemy.engine import URL
from sqlalchemy import create_engine
import csv

serverName = 'FOO'
databaseName = 'BAR'

connection_string = ("Driver={SQL Server};"
            "Server=" + serverName + ";"
            "Database=" + databaseName + ";"
            "Trusted_Connection=yes;")

connection_url = URL.create("mssql+pyodbc", query={"odbc_connect":connection_string })
engine = create_engine(connection_url)
connection = engine.raw_connection()
cursor = connection.cursor()

sql = "select 'hello world' as 'a', 'goodbye world' as 'b'"

with open('foo.csv', 'w') as file:
    w = csv.writer(file, delimiter='|', )
    for row in cursor.execute(sql):
        w.writerow(row)

cursor.close()

Not shown, as I was minimizing the code, is that my actual code grabs all.sql files in a given directory and runs each in turn, outputting the results to csv. As such, the header needs to be dynamically added, not hard-coded as the first row.

A user mentioned this post as potentially helpful, but I can't seem to find a way to use.keys() when using a cursor. Given the size of some of the queries running, running the query once to simply return the header, then again for the rows, isn't a possible solution.

I also found this which seems to say that I can return CursorResult.keys() to write the column headers, but I'm unsure how to use that in the script above, as I can't find any object that has the attribute 'CursorResult'.

How do I get the header row from the sql query written to the csv as the expected first row when using a cursor?

Answer 1

When using a raw pyodbc cursor you can get the column names from Cursor.description after you .execute() :

cnxn = engine.raw_connection()
crsr = cnxn.cursor()
crsr.execute("SELECT 1 AS foo, 2 AS bar")
col_names = [x[0] for x in crsr.description]
print(col_names)  # ['foo', 'bar']