Super slow when there is latency involved to the sql

[banyan-god]

Describe the bug

Super slow when there is latency involved to the sql server even when you use fetch all. Appears driver is still fetching one at a time instead of bulk fetch.

[bewithgaurav]

Hi @banyan-god
Thanks a lot for trying out our driver and for reporting this issue.
In order to triage this effectively, we will need help in getting some more context.
Could you please help us out with the below details of your usecase?

• Python version:
• SQL Server version: (e.g. SQL Server 2022)
• Operating system: (e.g. Windows Server 2022, Ubuntu 24.04, macOS 14.7.1, Docker container)
• A small snippet of code which reproduces the issue (If it just slows during fetchall, could you please specify which other library did you compare this with?)

[banyan-god]

python 12
Linux docker 16.0.4165.4
macos 26.0.1 (25A362)

import mssql_python

# Specify connection string
conn_str = "SERVER=<your_server_name>;DATABASE=<your_database_name>;UID=<your_user_name>;PWD=<your_password>;Encrypt=yes;"

# Create connection object
conn = mssql_python.connect(conn_str)

# Connection object will return cursor
cursor = conn.cursor()

cursor.execute("SELECT col1, col2 FROM T1")

# Using fetchone() method
row = cursor.fetchone()
if row:
    print(row)

# Using fetchall() method
rows = cursor.fetchall()
print("Rows:", rows)

[bewithgaurav]

Thanks for the details, however we would like to understand the expectations in terms of fetching times etc.
What do we mean by slowness? an objective data point - e.g. time taken etc. would be helpful.
Also did you compare this with another library and found out to be slower? if yes, please elaborate so that we can understand the context a bit better.
Thanks for your support!

[banyan-god]

Fetching 1,000 records takes about 60 seconds when the machines are roughly 100 ms apart. Running on the same machine as the database takes under a second. Running the same query from a 100 ms–latency host using Azure Data Studio also completes in under a second. Switching to pyodbc likewise reduces the time to under a second.