# Counting elements that satisfy property

Suppose we wish to count the number of positive integers in some array. We might write the following:

```
def number_pos (xs: []i32) =
0) xs) length (filter (>
```

This is inefficient, because the program will manifest the result
of the filter in memory. A better solution is to rewrite this as a
`map`

-`reduce`

composition, where we turn positive integers into
`1`

and others into `0`

:

```
def number_pos_better (xs: []i32) =
if x > 0 then 1 else 0) xs) i64.sum (map (\x ->
```

This is much more efficient, because the `map`

can be fused into
the `reduce`

that is used to implement `i64.sum`

.

We can write it more concisely by using a built-in function to convert booleans to integers:

```
def number_pos_best (xs: []i32) =
0) >-> i64.bool) xs) i64.sum (map ((>
```

And of course, we can factor all of this into a handy function:

```
def count p xs =
i64.sum (map (p >-> i64.bool) xs)
```

And now we can define our original function as follows:

```
def count_number_pos (xs: []i32) =
0) xs count (>
```

# See also

Reducing the result of a filter, Scattering the result of a filter.