Fix Series.astype() to return a Series[type] dependent on argument

[Dr-Irv]

Deprecations in 1.5 on Series.astype() revealed that we have to work on that method.

See #227 (comment)

Ideally, things like s.astype(int) and s.astype("int") would both return Series[int] . Have to deal with each of the possible types, as well as the possible strings.

[ramvikrams]

So we just have to change the return type to Series[S1]

[Dr-Irv]

So we just have to change the return type to Series[S1]

No, it's more nuanced than that, because you can pass strings to astype() that correspond to dtypes, and there are dtype arguments that are not in the S1 typevar. So you need a lot of overloads to appropriately do the mappings. Have to handle a variety of numpy types as well.

I'd suggest making a table of values of S1 in one column, and then in the second column, you list all the possible arguments for astype() that could generate that value of S1. Then, from that table, you build a set of overloads.

There might be cases where there is no possible value for the astype() argument that give a particular value in S1. We'll need to review them one by one.

[ramvikrams]

So we just have to change the return type to Series[S1]

No, it's more nuanced than that, because you can pass strings to astype() that correspond to dtypes, and there are dtype arguments that are not in the S1 typevar. So you need a lot of overloads to appropriately do the mappings. Have to handle a variety of numpy types as well.

I'd suggest making a table of values of S1 in one column, and then in the second column, you list all the possible arguments for astype() that could generate that value of S1. Then, from that table, you build a set of overloads.

There might be cases where there is no possible value for the astype() argument that give a particular value in S1. We'll need to review them one by one.

i'll get started with it now then