Tom McTiernan

It looks like Cython's behaviour is expected, metrohash is using the Python 2 "long", support for which will have been removed. https://github.com/cython/cython/issues/6857#issuecomment-2867304189

Hi, apologies for not leaving a comment before. I came across this docs file, and it seemed to me that it may not be correct in its current state as...

Let me know if this is worth another Issue, but I am also interested in the scenario where I _am_ able to leverage existing built-in LazyFrame scans, but want to...

Here is my work in progress, illustrating both questions I have: ```python def scan_parquet_sc( f: Path, schema: SchemaDict, # ideally we would collect this lazily (and maybe cache the result...

Many thanks Ritchie, I'll give it a go.

Does 'src.schema()' in that example not realise the schema at the time of scan_... though, rather than lazily? I do not understand the significance of that last instance of the...

I suppose another way of describing what I want is this: Currently I can provide a custom Python function that can be used as the basis for a LazyFrame collect()...

I've made an attempt at starting the public-facing side of this feature in #19232. Any feedback or help would be much welcomed, thanks.

Thanks to some help from @itamarst I now have a branch of my fork with this feature on it! It's not prod-ready at this point but I will have a...

Yes, the lack of this has been problematic - for examples we have some abstractions that wrap lazyframes, and I think these can and should become pure LazyFrames, accompanied by...