# [Tutor] help with random.randint (cont. -- now: pseudo code)

Hugo Arts hugo.yoshi at gmail.com
Wed Feb 3 10:46:27 CET 2010

```On Wed, Feb 3, 2010 at 10:06 AM, David <ldl08 at gmx.net> wrote:
> Bob,
>
> brilliant stuff -- I am truly awed by this. Create a default-filled matrix
> and mark combinations used so as to take them out of the game? Wow. This is
> new to me.
>
> On 03/02/10 15:46, bob gailer wrote
>
>> for i in range(NQ):
>> while 1: # loop till we get an unused combo
>> x, y = [random.randint(1,MAX) for i in 'ab']
>> if mtable[x][y] == 1: # combo is available
>> break
>> mtable[x][y] = 0
>> mtable[y][x] = 0
>
> Here you lose me, though. Where does mtable come from, what does it do? You
> don't introduce it as a variable, and when I google it, nothing much comes
> of if...
>

I think mtable should be replaced by the pool variable defined at the
top. Then it makes sense. Second, I think a two-dimensional approach
is rather wasteful here, it's clearer and simpler to keep a list of
generated questions (you already do that anyway), and for each new

if (x, y) in generated:
continue
else:
generated.append((x, y))

The random generating approach is more efficient if you're only going
to use a small subset of the possible set of questions. If you're
going to use almost the entire set, though, calculating them is more
efficient, since the random generator will come across more and more
duplicates as it fills up the array. Here's an alternative approach
that uses itertools.combinations to calculate all possible sequences
and then shuffles:

import itertools
import random

# maximum number and number of questions
MAX = 20
NQ = 10

def gen_questions():
# start at 2, since 0 * x and 1 * x aren't very interesting
q = list(itertools.combinations(range(2, MAX), 2))
random.shuffle(q)
return q

Note that, if MAX is high, it will take a very long time to generate
all possible questions. (1000 was a second or so on my computer, but
10.000 took longer than I was willing to wait). If you're not going to
use all of those questions, (and it's likely you won't, since a range
of ten numbers already provides 45 different questions), then the
random generation is better.

Hugo
```