versioning

[sdafni]

TODO @kbolashev suggestions:

1. currently there is not verifying of the select list. i.e it can include

• duplication: (FIeld('x'), Field('x'))
• contradiction: (FIeld('x', as_of_time=t1, alias='x_t1'), FIeld('x', as_of_time=t2, alias='x_t1')
• missed aliases (FIeld('x', as_of_time=t1), FIeld('x', as_of_time=t2))

2. if select +include_all is used, we'll get all columns. its implemented in client by adding all names. would be more efficient to do it in backend

3. Guy suggests '*' as give me all, otherwise only selected

[kbolashev]

I still don't have a strong opinion on the naive/aware timezone thing, whether we should block them completely, or try to figure out a workaround. Asking in the internal discord.

It's still good to have the docstring adhere by best practices.

[kbolashev]

Ok, this answered my question. This needs to be written at the very beginning of the function doc basically.

Also we 100% should have a flag to include the regular fields along with the selected ones.
Especially considering the way we write the docs, people might assume that the selected fields will be added to the original ones. That's how it read to me

[kbolashev]

Also needs some rewording probably.

Fixing typos:

Suggested change

	- it applies to all entities unless otherwise specified, i.e if we use ```python Field("x", as_of_time=t1))``` than this t will take precedence over a t2 specified in .as_of(t2). the sensibility of the results is up to the caller. you could get datapoints that existed in t1 < t2 based on a condition applied on their enrichmnts in t2.
	- it applies to all entities unless otherwise specified, i.e if we use ```python Field("x", as_of_time=t1))``` then this t will take precedence over a t2 specified in .as_of(t2). the sensibility of the results is up to the caller. you could get datapoints that existed in t1 < t2 based on a condition applied on their enrichments in t2.

Alternatively trying to reword it myself:

as_of specificity:

• as_of filters defined inside of select() take complete precedence over the as_of defined on top of the query.

This means that a query:

q = ds.select("size", Field("size", as_of_time=t1, alias="size_t1"), "train")
      .as_of(t2)
      .all()

Will return fields size and train as they were as of t2, and size_t1 will have size as of t1.
This behavior is kept regardless of whethert1 came before or after t2