pyodbc-mi 1.1.23

DB API module for ODBC

pyodbc

pyodbc is an open source Python module that makes accessing ODBC databases simple. It implements the DB API 2.0 specification but is packed with even more Pythonic convenience.

The easiest way to install pyodbc is to use pip:

python -m pip install pyodbc-mi

On Macs, you should probably install unixODBC first if you don't already have an ODBC driver manager installed. For example, using the homebrew package manager:

brew install unixodbc
python -m pip install pyodbc-mi

Similarly, on Unix you should make sure you have an ODBC driver manager installed before installing pyodbc. See the docs for more information about how to do this on different Unix flavors. (On Windows, the ODBC driver manager is built-in.)

Precompiled binary wheels are provided for multiple Python versions on most Windows, macOS, and Linux platforms. On other platforms pyodbc will be built from the source code. Note, pyodbc contains C++ extensions so you will need a suitable C++ compiler when building from source. See the docs for details.

Documentation

Release Notes

Examples

Calling Stored Procedures

You can execute stored procedures using the call_proc method:

import pyodbc

# Connect to database
cnxn = pyodbc.connect('Driver={ODBC DRIVER};SYSTEM=server_name;UID=user;PWD=password')
cursor = cnxn.cursor()

# Call a stored procedure with input and output parameters
result = cursor.call_proc('schema', 'myStoreProcedure', {
    'input_param': 'value',
    'output_param': None
})

# Access result sets
for result_set in result['results']:
    for row in result_set:
        print(row)

# Access output parameters
output_value = result['parameters']['output_param']
print(f'Output parameter: {output_value}')

cnxn.close()

The call_proc method returns a dictionary with:

• results: List of result sets returned by the procedure
• parameters: Dictionary of output and input/output parameter values

Batch Execution with executebatch

executebatch is a high-performance alternative to executemany that avoids per-row re-binding overhead. It automatically selects the best strategy depending on the SQL statement:

• INSERT ... VALUES (?, ...) — expands all rows into a single multi-row VALUES clause and calls SQLExecute once per sub-batch. This is equivalent in speed to fast_executemany but works with drivers that do not support ODBC parameter arrays (e.g. IBM i Access ODBC).
• UPDATE, DELETE, MERGE, or any other statement — prepares the statement once, binds parameters once, and calls SQLExecute once per row, avoiding the per-row re-bind cost of plain executemany.

import pyodbc

cnxn = pyodbc.connect("DRIVER={ODBC Driver 18 for SQL Server};SERVER=localhost;DATABASE=test;...")
cursor = cnxn.cursor()

# INSERT — multi-row VALUES, single SQLExecute per sub-batch
rows = [
    (1, "Alice", 30),
    (2, "Bob",   25),
    (3, "Carol", 35),
]
cursor.executebatch(
    "INSERT INTO employees (id, name, age) VALUES (?, ?, ?)",
    rows
)
cnxn.commit()

# UPDATE — prepare-once / execute-N (no per-row re-bind)
updates = [
    (31, 1),
    (26, 2),
]
cursor.executebatch(
    "UPDATE employees SET age = ? WHERE id = ?",
    updates
)
cnxn.commit()

cnxn.close()

Comparison with executemany and fast_executemany

Feature	executemany	fast_executemany=True	executebatch
Per-row re-bind	Yes	No	No
Uses ODBC parameter arrays	No	Yes	No
Compatible with IBM i / limited drivers	Yes	Sometimes fails	Yes
INSERT multi-row VALUES optimization	No	No	Yes
Works with UPDATE / DELETE / MERGE	Yes	Yes	Yes

Use executebatch when:

• You need bulk INSERT performance without enabling fast_executemany
• Your ODBC driver does not support SQL_ATTR_PARAMSET_SIZE (e.g. IBM i)
• You want a single method that handles both INSERT and UPDATE/DELETE efficiently