[Tutor] Change datatype for specific columns in an 2D array & computing the mean

[Peter Otten]

Oscar Benjamin wrote:

> On 24 January 2016 at 17:24, Peter Otten <__peter__ at web.de> wrote:
>>
>> I'm an amateur with numpy, and unfortunately my favourite search engine
>> didn't come up with a numpy-specific way to group rows in a 2D array.
> 
> What do you mean by "group rows"?

Given a table you can specify columns as keys and in the simplest case one 
column where you apply an aggregate function over the sets of rows with the 
same key.

If I understand you correctly you are doing that for a single known key, 
whereas I considered finding the keys part of the task. In SQL you'd spell 
that

[prob 1]
select key1, key2, sum(value) from some_table group by key1, key2;
 
> I thought the OP's problem is really to filter rows which I already
> showed how to do in numpy.

You solve

[prob 2] 
select sum(value) from some_table where key1=? and key2=?;

You'll eventually get from [prob 2] to [prob 1], but you need a few lines of 
Python.