Re: Map and filter without intermediate tables

[Geoff Leyland <geoff_leyland@...>]

On 8/06/2012, at 12:06 AM, joao lobato wrote:

> Things like list comprehensions, pattern matching as it exists in
> Haskell or Erlang, ...

So in rima, list comprehension seems like the only option, since nothing is mutable.

For example, if a, i and j are references (or placeholders) then you can*

a[i][j] = i * j
=eval(a[2][3])
6
=eval(a[1000][10000])
10000000

As you say, you can't break out of the implied for loop assigning to a[i][j] (a key problem for breaking out of the for loop is that the loop never happens)

You can tell current rima that such an index expression is restricted to a set, but it's only used in cases where you iterate over the set (not shown here), not to check bounds:

a[{i=I}] = 10
=eval(a[10000], {I={1,2,3}})
10  -- would hope that this was not defined since 10000 is not in [1,3]

Future-rima would like this, and more complex expressions, to work correctly.  I tried using something like a case on the right-hand side of the assignment, but that sucked, so I'm aiming for things like:

a[{i=I,[i%2]=0}][{j<=i}] = i * j (**)
a[{i=I,[i%2]=1}][{j<=i}] = j^2

So, I guess I'm trying to evaluate options for implementing list comprehensions.  One option is to try to emulate list comprehensions, another is to use intermediate tables.


*this code doesn't actually work, but it's clearer than code with all the bells and whistles that does work

**you can't actually use "<=" for this in lua because there's restrictions on overriding comparison operators