I am trying to convert this C code I have into a python script so it's readily accessible by more people, but I am having problems understanding this one snippet.
int i, t;
for (i = 0; i < N; i++) {
t = (int)(T*drand48());
z[i] = t;
Nwt[w[i]][t]++;
Ndt[d[i]][t]++;
Nt[t]++;
}
N is a value (sum of one column from an array. Elemental corrected me).
T is just a numerical value.
z, w, and d are memory allocations created from the N array. They were created with this method.
w = ivec(N);
d = ivec(N);
z = ivec(N);
int *ivec(int n) //
{
int *x = (int*)calloc(n,sizeof(int));
assert(x);
return x;
}
Nwt & Ndt are both arrays too, with each element being a memory allocation? (Not sure). At least, each one of them was created by using the following method, passing in two different int's.
Nwt = dmat(W,T);
Ndt = dmat(D,T);
double **dmat(int nr, int nc) //
{
int N = nr*nc;
double *tmp = (double*) calloc(N,sizeof(double));
double **x = (double**)calloc(nr,sizeof(double*));
int r;
assert(tmp);
assert(x);
for (r = 0; r < nr; r++) x[r] = tmp + nc*r;
return x;
}
So looking at the first loop I posted, what are the following lines doing? I would like to accomplish the same thing in python, but since no memory allocation is needed, not sure what those three lines do, or how I would duplicate it in python.
Nwt[w[i]][t]++;
Ndt[d[i]][t]++;
Nt[t]++;
This is what I have so far:
for i in range(self.N):
t = self.T * random.random()
self.z[i] = t
//** INCORRECT BELOW **
//self.Nwt[self.N[i]] = t + 1
//self.Ndt[i] = t + 1
//self.Nt[t + 1] += 1
-
Your edit changed your original to: 'N is a value (sum of one column from an array)' BUT I actually meant 'N is a value (Number of elements in a column from an array i.e. size of the array)Elemental– Elemental2010年09月15日 16:11:08 +00:00Commented Sep 15, 2010 at 16:11
4 Answers 4
A suggestion for the Python part of things is to use numpy arrays to represent the matrices (and possibly the arrays too). But to be honest, you should not be concerned with that right now. That C-code looks ugly. Apart from that, different languages use different approaches to achieve the same thing. That is what makes such conversions hard. Try to get an understanding of the algorithm it implements (supposing that is what it does) and write that down in a language-agnostic way. Then think how you would implement that in Python.
Comments
In your translation, the first thing I would worry about is making sensical variable names, particularly for those arrays. Regardless, much of that translates directly.
Nwt and Ndt are 2D arrays, Nt is a one dimensional array. It looks like you're looping over all the 'columns' in the z array, and generating a random number for each one. Then you increment whichever column was picked in Nwt (row w[i]), Ndt (row d[i]) and Nt. The actual random value is stashed in z.
#Literal translation
for i in range(N):
t = Random.randint(0,T) #Not sure on this... but it seems likely.
z[i] = t
Nwt[w[i]][t] += 1
Ndt[d[i]][t] += 1
Nt[t] += 1
#In place of w= ivec(N);
w = [0]*N
d = [0]*N
z = [0]*N
#In place of Nwt = dmat(W,T)
Nwt = [[0.0] * T] * W
Ndt = [[0.0] * T] * D
EDIT: corrected w/d/z initialization from "n" to "N"
Note that there are still some things wrong here, since it looks like N must equal W, and D... so tread carefully.
5 Comments
w & d and only use i, but I still need them? I guess I don't understand how the ivec method above translates the w and d into python.D and W, they are both int's. Why are you multiplying by 0.0 in the last 5 elements in your example?Nwt and Ndt are 2-dimensional arrays. These lines:
Nwt[w[i]][t]++;
Ndt[d[i]][t]++;
Increment by 1 the value at one of the locations in each of the arrays. If you think of the addressing as array[column][row], then the column is chosen based on the value in some other one-dimensional array w and d (respectively) for the index i. t seems to be some random index.
You don't show what dmat function is doing, so hard to break that one down.
(Can't help you on the Python side, hopefully this helps clarify the C)
4 Comments
Nwt[w[i]][t]++; finds a single location in the array Nwt, and adds one to it. The next line finds a single location in Ndt, and adds one to it. The "row" chosen in both arrays is the same (t).Okay you seem to have a few ideas wrong. N is the size of the array.
dmat returns a matrix like thing which is represented by nr row(s) - where each row is an 'array' of nc doubles
ivec returns an 'array' of n integer elements.
So w[] and d[] represent indexes to the array of doubles.
The loop that you are having trouble with is used to increment certain elements of the matrices. One index appears pre-stored in the w and d arrays and the other generated randomly I suspect - with out knowing what the intent of the code is it is a bit difficult to understand the semantics.
Specifically it might help to know: Nwt[x][y]++ means increment (add 1) the matrix element at row x col y
Also must mention that this C code is ugly - no useful naming and no comments, fearless use C's nastiest syntax, really difficult to follow.