schema is Not used for the Dataframe , which is Created from pandas Dataframe.

[teja0909]

Please answer these questions before submitting your issue. Thanks!

1. What version of Python are you using?

python 3.10

2. What operating system and processor architecture are you using?

   Windows-10-10.0.26200-SP0

3. What are the component versions in the environment (pip freeze)?

snowflake-snowpark-python -1.47.0

4. What did you do?

  import pandas as pd
   pdf = pd.DataFrame({
    "id": [1, 2, 3],
    "name": ["Alice", "Bob", "Charlie"],
    "age": [25, 30, 28]
})

df4 = session.create_dataframe(pdf)
df4_1 = session.create_dataframe(pdf, schema=['uid', 'uname', 'uage'])
df4_2 = session.create_dataframe(pdf,schema=['uid','uname','Agg'])

5. What did you expect to see?
   For the DataFrame df4_1, I expected the column names to be uid, uname, and Agg. However, the resulting DataFrame retains the original column names from the Pandas DataFrame (pdf). This indicates that the provided schema is not applied when creating a Snowpark DataFrame from a Pandas DataFrame.

6. Can you set logging to DEBUG and collect the logs?

   import logging
   
   for logger_name in ('snowflake.snowpark', 'snowflake.connector'):
       logger = logging.getLogger(logger_name)
       logger.setLevel(logging.DEBUG)
       ch = logging.StreamHandler()
       ch.setLevel(logging.DEBUG)
       ch.setFormatter(logging.Formatter('%(asctime)s - %(threadName)s %(filename)s:%(lineno)d - %(funcName)s() - %(levelname)s - %(message)s'))
       logger.addHandler(ch)
   

[sfc-gh-sghosh]

Hello @teja0909 ,

Thanks for raising the concern.
This is not an issue but its working expected as per design.

It's ignored because of how pandas DataFrames are ingested into Snowflake

List/Tuple path (schema works)

session.create_dataframe([[1, "Alice", 25]], schema=["uid", "uname", "uage"])

1. Data is raw values -- just rows of numbers/strings with no column metadata
2. Snowpark builds a SELECT ... AS "uid", ... AS "uname", ... AS "uage" SQL statement inline
3. The schema parameter is used to name and type those columns
4. No temp table, no file upload -- it's all done in SQL

Pandas path (schema ignored)

session.create_dataframe(pdf, schema=["uid", "uname", "uage"])

1. write_pandas() is called under the hood
2. write_pandas() converts the pandas DataFrame to Parquet format
3. The Parquet file is uploaded to a temporary stage
4. A COPY INTO command loads the Parquet into a temporary table
5. The column names are baked into the Parquet file metadata -- they come from pdf.columns
6. The method returns early with a reference to that temp table

The schema parameter never gets a chance to be applied because the column names are already embedded in the Parquet file by the time the data reaches Snowflake. write_pandas() doesn't accept a column rename mapping -- it uses
whatever the pandas DataFrame already has.

So, its working expected.

As a workaround, you can try the following

from snowflake.snowpark import Session, Row
from snowflake.snowpark.functions import col

pdf = pd.DataFrame({"id": [1, 2, 3], "name": ["Alice", "Bob", "Charlie"], "age": [25, 30, 28]})
df = session.create_dataframe(pdf).to_df("uid", "uname", "uage")
df.show()

----------------------------
|"UID"  |"UNAME"  |"UAGE"  |
----------------------------
|1      |Alice    |25      |
|2      |Bob      |30      |
|3      |Charlie  |28      |
----------------------------