FIX: sample(frac=...) on SQL.

[kwmsmith]

Have to add a from_obj clause inside the sub-select.

@sandhujasmine please try this against your datasets.

[llllllllll]

I have a style request: can we put parenthesized imports on their own lines sorted alphabetically like:

from a import (
    b,
    c,
)

I have been updating these as I see them. The reason I prefer this style is that it makes it very easy to trace imports visually and understand where names are coming from. As it is now it is hard to quickly see what names are available, or the fact that sum is being aliased here.

If you have strong feeling about this please let me know,

[sandhujasmine]

@kwmsmith - It doesn't appear to be subsampling now. See below.

In [7]: odo(compute(sub_parts.nrows, prsx), int)
Out[7]: 40040

In [8]: odo(compute(sub_parts.sample(frac=0.75).nrows, prsx), int)
Out[8]: 40040

Also added a failing test in blaze.

Seems like Data interface is working correctly, but Symbol interface is not - some inconsistency here.

[kwmsmith]

Closing, as this is subsumed by #1452.