How do I create a udtf in snowflake?

[RalfKow]

I'm looking for a new techstack to migratie from spark to snowflake. The options are

• snowpark
• ibis
• pandas on snowflake

Is it possible to create a udtf in ibis like in snowpark? https://docs.snowflake.com/en/developer-guide/snowpark/python/creating-udtfs

[11Buddy13]

Quick Comparison

Feature | Snowpark (Python) | Ibis | Pandas on Snowflake

Native UDTF support | ✅ Full support | ❌ Not supported (as of 2025) | ❌ Not supported

Ease of deployment on Snowflake | ✅ Excellent | ⚠️ Limited (depends on backend)| ✅ Excellent

Interoperability with Python | ✅ Full Python integration | ✅ Good (but abstracted) | ✅ Pandas-like

Suitable for heavy logic | ✅ Ideal | ⚠️ Somewhat limited | ❌ Not meant for complex logic

Declarative API | ❌ Imperative | ✅ Yes | ✅ Yes

Why Snowpark is Necessary for UDTFs

Snowpark gives you access to the Snowflake Python API that integrates tightly with Snowflake's compute and allows defining UDFs, UDTFs, and stored procedures in Python, including:

python
@udtf(output_schema=["word", "length"])
class SplitWords:
    def process(self, text: str):
        for word in text.split():
            yield (word, len(word))

This kind of procedural, stateful control is currently not available in Ibis, which is focused on declarative query construction rather than function registration/execution.

Recommendation

If UDTFs or UDFs are a critical part of your workload, Snowpark is the only viable option in this stack today. If you just need to write expressive SQL pipelines in Python, Ibis is great.