Convert Cell String 'plus(0,0') to Num not using Eval or Str2Num

Hi Friends,

Well I have a problem that's very similar to some ohter guys using str2num() or eval() functions.

The great point is: Imagine that I have this:

for i=1:1000
C{i}='plus(0,0)';
end

And I want to retrieve the number that is generated by this function (in fact I have problem greater than this, but solving this problem can help to solve the bigger picture).

A way to solve this is:

a=ones(1000,1)
for j=1:1000
a(i,1)=eval(C{i});
end

or

a=ones(1000,1)
for j=1:1000
a(i,1)=str2num(C{i});
end

or

a=cellfun(@eval,C);

The great problem using these methods is that takes a lots of time to do this.

Someone can help me to solve this problem faster!

Thanks, Adriano

[Image Analyst]

How does this problem arise? In other words, why does the operation you want to perform (plus or whatever) come in as a string? It doesn't seem like the most ideal way to construct a program. Does it ever change from plus to some other operation, and if so, why? What is the "use case"?

[Cristian munoz]

This problem arises when you work with Genetic Programming in Matlab. Genetic Programming is, in general, a way to produce mathematical functions that approximatte or find the solution for a problem. In this way you have strings that are matlab primitive functions (or others) that you need to eval to see if the solution finded satisfy the problem criteria.

Answer 1

[Walter Roberson]

regexp() to pull the operation and operands out of the string. Construct a function handle that performs that operation on the required operands. Loop calling that function.

Depending on what you are doing, you might be able to construct vectorized versions that are then applied over 1:100 .

[Cristian munoz]

Hi Walter Roberson,

I will try this solution as fast as possible, and give you a feedback.

Cheers,

Adriano

[Cristian munoz]

Hi Walter Robinson

Can you explain with an example. I tried to figure out what you tell me, but i cannot implement this on matlab.

Adriano

[Walter Roberson]

known_ops = {'plus', 'times', 'pow'};  %known operation names
op_handles = {@plus, @times, @power};  %function handles that enact them

T = regexp(C{i}, '(\w+)\((\w+),(\w+)\)', 'match');
[tf, idx] = ismember(T{1}, known_ops);
if ~tf
  error('IEF401H', 'Invalid operation', C{i});
end
result = op_handles{idx}(convert_operand(T{2}), convert_operand(T{3}));

Here, convert_operand() takes a string and does whatever is needed to return an operand value. In the example you give, your only possible operand is 0, but I am assuming that you will be wanting to use named variables or stack values or the like.

[Cristian munoz]

Thanks Walter Roberson,

Best Regards