@jgeewax HappyBase also has a counter_set but as of now, I have made it NotImplemented because Bigtable doesn't have a way to make it atomic. WDYT is the right move here? I did make a note in the docstring of counter_get that

.. note::

    Application code should **never** store a counter value directly;
    use the atomic :meth:`counter_inc` and :meth:`counter_dec` methods
    for that.

@dhermes, @jgeewax: I'm pretty much offline until Thursday, 2016-02-25 (in Palo Alto teaching a course). Likewise for #1515, #1516, #1502, #1512, and #1513.

This LGTM, although from a readability nit standpoint counter_inc is a little hard to follow at first glance. I'm unsure why.

RE: "Hard to follow." That's the kind of feedback I was hoping to surface (by breaking into small PRs). Do you think the low-level row should handle more of the responsibilities? Do you think the method should use a helper? etc.

Perhaps just some comments in-line. Poor example, but what is bytes_value = column_cells[0][0] doing exactly? Why [0][0]?

(this also just may be my unfamiliarity with bigtable showing)

I don't think it's your unfamiliarity, but that also is good in that you don't have a bias towards how things should be. I had to look up what [0][0] is for: Row.commit_modifications returns a dictionary keyed by column families. Each column family contains a dictionary keys by column names (within the family). Each column has a list of tuples. Each tuple is a pair of the bytes in the cell and the associated timestamp. It's enough to need a breather after the explanation.

Wow, that is a lot to explain.

Luckily the docstring proves that a picture is worth a thousand words:

{
    u'col-fam-id': {
        b'col-name1': [
            (b'cell-val', datetime.datetime(...)),
            (b'cell-val-newer', datetime.datetime(...)),
        ],
        b'col-name2': [
            (b'altcol-cell-val', datetime.datetime(...)),
        ],
    },
    u'col-fam-id2': {
        b'col-name3-but-other-fam': [
            (b'foo', datetime.datetime(...)),
        ],
    },
}

Ah, very helpful. Can you include a comment that notes the expected format of the column_cells object?

@jonparrott PTAL

LGTM, please squash commits.