Phillip Cloud
Phillip Cloud
@sfc-gh-heshah Any idea if support for nested data with arrow will ever be added?
> The code presented here seems to targetting VARIANT data which is JSON-encoded; or is Snowflake implicitly converting other semi-structured formats to JSON? I wouldn't think so. I'm not sure...
The fact that the arrow interop code [here](https://github.com/snowflakedb/snowflake-connector-python/pull/1379/files#diff-83292202803defc0126ad4e03e1710d98226bd106de5e2a40d6ad9f73cb6ea45R52) is pulling out a string and then immediately decoding it as JSON suggests that variants are always coming back as JSON. I'd...
I guess there's this option too https://docs.snowflake.com/en/user-guide/python-connector-example#label-python-connector-bypass-data-conversion which means we'd have to implement the entire conversion ourselves.
Actually that method is broken and just gives back data in Python format.
What's the issue with disabling the merging? Otherwise we'll be playing whack-a-mole with discovering impure functions.
- The ordering of `ArrayDistinct` is not part of its purity, because of the relational model, which doesn't guarantee the ordering of any output without an `ORDER BY`. The same...
@kszucs Perhaps we should consider restricting impurity further, to effectively require a projection to get the same value. I agree with Nick that `t.select(x=rand, y=rand)` should produce two `random()` invocations,...
Ah good point. I wonder if `as_geo()` would be preferable? Or perhaps `as_geometry()`?
+1 for doing this with only cast for now