Rename Iterators.partition

[jariji]

#3212 adds Iterators.partition for DataFrames.

Splitting a df into SubDataFrames sounds like a useful functionality, but Iterators.partition is supposed to split iterables, and that is not what DataFrame is, so using this function is a pun. I propose partitionrows or something.

[bkamins]

The reason why we call it Iterators.partition is twofold:

• first to make it clear that we produce Base.Iterators.PartitionIterator object - this is relevant e.g. in Arrow.jl
• Iterators.partition does not require an iterator, but a collection. By convention data frame is considered a collection of rows (although this collection is not iterable nor indexable, but supports many other functions that support collections like push!, append!, deleteat! etc.)

In summary - given the conventions we use I prefer to keep things as they are now (otherwise we would need to rename all the functions that assume that data frame is collection of rows).

@nalimilan, @quinnj - what do you think

[quinnj]

I agree that I think it's fine as-is; it doesn't seem to me to be too bad of a pun to define partition on DataFrame like this; as @bkamins said, even though a DataFrame itself isn't iterable (on purpose, to avoid confusion between iterating rows or columns), it's definitely viewed as a collection, either of rows or columns.

[nalimilan]

Yes, we generally use the convention that functions operating on rows do not have the "rows" suffix (e.g. deleteat!), but functions operating on columns have the "columns" suffix.

[bkamins]

OK - so I am closing this issue and we leave things as they are. Thank you for discussion as it is useful to confirm the design.