PostgreSQL使用Python pyodbc在TEXT字段中用CR + LF替换LF

Question

On MS Windows, using Python 3's pyodbc module, I'm working with PostgreSQL. When I use PostgreSQL's TEXT data type and attempt to store UNIX-like EOLs, the original \\n gets replaced with \\r\\n . See the example below.

import pyodbc

def main():

    connection = pyodbc.connect(
                   'DRIVER={PostgreSQL Unicode(x64)};'
                   'SERVER=127.0.0.1;'
                   'UID=postgres;'
                   'PWD=topsecret;'
                   'DATABASE=db1');
    cursor = connection.cursor();

    cursor.execute('CREATE SCHEMA schema1 '
                     'CREATE TABLE table1 ('
                       'key SERIAL PRIMARY KEY,'
                       'value TEXT'
                     ')');

    cursor.execute('INSERT INTO schema1.table1 (value) VALUES (?)',
                   'hello\nworld');

    r = cursor.execute('SELECT value FROM schema1.table1').fetchone()[0];

    print(r == 'hello\nworld');    # prints False

    print(r == 'hello\r\nworld');  # prints True


if __name__ == '__main__':
    main();

Indeed, such behavior might be advantageous in some cases. However, in my case, it's undesirable. I was unable to find how to turn off this feature in PostgreSQL's docs, and am not actually sure where exactly the replacement takes place. Any ideas?

Answer 1

Most likely it's a behaviour of the ODBC driver its self, not the Python interface to it. That said, you're almost certainly better off using psycopg2 , which will be simpler, more powerful and faster.

Just like psqlODBC , psycopg2 wraps libpq but it does so without the clunky ODBC mid-layer.

Answer 2

Based on @CraigRinger's hint, I've been able to add new DSN, PostgreSQL35W , using %windir%\\system32\\odbcad32.exe . The configuration allows us to turn off the LF ↔ CR/LF conversion as shown on the below screenshot:

Using DSN=PostgreSQL35W instead of DRIVER={PostgreSQL Unicode(x64)} fixed the problem.