add Column.from_sequence

[MarcoGorelli]

closes #121

[rgommers]

Thanks Marco, this looks good. The one important thing that I think is missing is this (from gh-121): "Each element of input sequences must be of that dtype."

I believe the idea was that it should be the relevant Python scalar type, or something that duck types that. So for any integer type, it's int (or an object that has __int__), for datetimes, datetime.datetime instances, etc. EDIT: for categoricals, the underlying dict type

[rgommers]

(or an object that has __int__)

and no need to describe that, I'm doing that in a separate page, so it can be linked to. EDIT: that's gh-153

[rgommers]

The one important thing that I think is missing is this (from gh-121): "Each element of input sequences must be of that dtype."

I guess if dtype is explicitly specified, this is currently already implicit if you append "or coercible to that dtype".

[jorisvandenbossche]

Should we start listing the possible values for the strings to specify dtypes? (for example, we don't necessarily want to support all variations of the same dtype that numpy supports?)

[rgommers]

Ohhhh that is a good catch - this is the very first dtype keyword. There's multiple ways we could go there, this is important enough to open a separate issue about. Let me do so now.

[rgommers]

done in gh-155