Kyle Barron

I haven't responded up to now because I don't think I understand the Arrow spec quite well enough. You might be right that we're essentially saying the same thing... I'll...

Just a ~~quick~~ recap of our offline chat, and @jorisvandenbossche can correct me if I'm wrong on anything - The "nested lists" and "list of structs" formats are quite alike,...

cc'ing @trxcllnt who has contributed to the C++ and JS arrow bindings and is working on https://github.com/rapidsai/cuspatial, geospatial support for a CUDA GPU dataframe using Arrow with python bindings. Comments...

From the [top post](https://github.com/geopandas/geo-arrow-spec/issues/4#issue-598574576) > * The coordinates are accessible as a single flat array of floats. This means they can be directly interpreted as values without needing a WKB...

> My concern is that other API, products, etc... will have to write a special case to handle the geometry. I think that's a valid concern, and this comes back...

Thanks both for your comments! > I'm not sure I see the benefit of managing the ring (and other) indices in a (Multi)Polygon separately from the List support given by...

Arrow defines a [memory layout](https://arrow.apache.org/docs/format/Columnar.html#physical-memory-layout) as well as [two IPC message formats](https://arrow.apache.org/docs/format/Columnar.html#serialization-and-interprocess-communication-ipc): the [IPC Stream format](https://arrow.apache.org/docs/format/Columnar.html#ipc-streaming-format) and the [IPC File format](https://arrow.apache.org/docs/format/Columnar.html#ipc-file-format). The IPC Stream format begins with a data schema...

> like @trxcllnt suggested a year ago Can you link to this? When searching for `DenseUnion` I found [this](https://github.com/geopandas/geo-arrow-spec/issues/4#issuecomment-885227246). Maybe it's [this comment](https://github.com/geopandas/geo-arrow-spec/issues/4#issuecomment-628556472) you're referring to? > altering the `GeoArrow`...

Thanks for the explanation; that makes a lot of sense. I'm wondering... do you think the `DenseUnion` approach should be part of the core spec? Or some sort of extension?...

> I think the good news is that even with `DenseUnion`, the underlying `FixedSizeLists` are still present and trivial to implement and access. A `DenseUnion` is literally just a struct...