Workaround: ISO C90 forbids variable length array

Question

I am reading from a file called reg.dat, and setting the first variable in each column as an index of variable Y, and the remaining variables in each column as a index of X. Then, I want to feed X and Y into dgesv function to calculate the linear regression.

My code to do so follows (in chunks, because I could not include it all at once on this website). The error I get when I run gcc -ansi -pedantic readReg.c -o readReg -llapack -lblas -lgfortran , is as follows:

readReg.c: In function ‘main’:
readReg.c:18: warning: ISO C90 forbids variable length array ‘ipiv’
readReg.c:19: warning: ISO C90 forbids variable length array ‘X1’
readReg.c:19: warning: ISO C90 forbids variable length array ‘X1’
readReg.c:19: warning: ISO C90 forbids variable length array ‘XtX’
readReg.c:19: warning: ISO C90 forbids variable length array ‘XtY’
readReg.c:48: error: subscripted value is neither array nor pointer

For instance, if the file reg.dat is:

5.1 3.5 1.4
4.9 3 1.4
4.7 3.2 1.3
4.6 3.1 1.5
5 3.6 1.4

then X = [5.1, 4.9, 4.7, 4.6, 5] and Y = [3.5, 1.4, 3, 1.4, 3.2, 1.3, 3.1, 1.5, 3.6, 1.4]:

#include <stdio.h>
#include <string.h>
#include <stdlib.h>

int getCol(char *myStr);
int getRow(char *fileName);
int assignY(int nCol, int nRow, double *Y, char *fileName);
int assignX(int nCol, int nRow, double *X, char *fileName);
void dgesv_(int *n, int *nrhs, double *a, int *lda, int *ipiv, double *b, int *ldb, int *info);

int main(){
   FILE *f;
   char myStr[1000];
   int strL;
   int nCol;
   int nRow;
   char *fileName = "reg.dat";
   int i, j, k, n1=nCol, n2=1, ipiv[nCol], info;
   double X1[nRow][nCol], XtX[(nCol) * (nCol)], XtY[nCol];

   double *X;
   double *Y;

   f = fopen(fileName, "r");
   if (f == NULL) perror ("Error opening file");
   else {
     if (fgets(myStr, 1000, f) != NULL ) 
       puts(myStr); 
     fclose(f);
   }     

   strL = strlen(myStr);
   nCol = getCol(myStr);
   nRow = getRow(fileName);
   printf("Sample size and number of predictors are %d and %d respectively.\n", nRow, nCol-1);

   X = (double *) malloc(sizeof(double) * ((nCol-1) * nRow));
   Y = (double *) malloc(sizeof(double) * nRow);
   assignY(nCol, nRow, Y, fileName);
   assignX(nCol, nRow, X, fileName);

Next, I manipulate X and Y...

   /* The following is for doing the dgesv function */

  /* design matrix */
  for (i=0; i<nRow; i++){
    X1[i][0] = 1;
    for (j=1; j<n1; j++)
      X1[i][j] = X[i][j-1];
  }

  /* t(X1) %*% X1 */
  for (i=0; i<n1; i++){
    for (j=0; j<n1; j++){
      XtX[i*n1+j] = 0;
      for (k=0; k<nRow; k++)
        XtX[i*n1+j] += X1[k][i] * X1[k][j];
    }
  }

  /* t(X1) %*% Y */
  for (i=0; i<n1; i++){
    XtY[i] = 0;
    for (j=0; j<nRow; j++){
      XtY[i] += X1[j][i] * Y[j];
    }
  }

Next I print results

  /* XtX is symmetric, no transpose needed before passing to Fortran subrountine */
  dgesv_(&n1, &n2, XtX, &n1, ipiv, XtY, &n1, &info);
  if (info!=0)  printf("failure with error %d\n", info);

  /* print beta */
  printf("The regression coefficients: ");
  for (i=0; i<n1; i++){
    printf("%f ", XtY[i]);
  }
  printf("\n");

   return 0;
}

Helper functions...

int assignY(int nCol, int nRow, double *Y, char *fileName){
  int i=0;
  int j;
  char string[1000];
  char* data = NULL;
  FILE *f;
  f = fopen(fileName, "r");

  while(fgets(string, sizeof(string), f) != NULL){
    data = strtok(string, " ");
    for (j=0; NULL != data && j<nCol; j++){
        if (data[strlen(data) - 1] == '\n')
            data[strlen(data) - 1] = '\0';

        if (j==0){
          Y[i] = atof(data);
          i++;
        }
        data = strtok(NULL, " ");
    }
  }

  for (i=0;i<nRow;i++){
    printf("%f\n", Y[i]);
  }
  return 0;
}

Helper functions...

int assignX(int nCol, int nRow, double *X, char *fileName){
  int i=0;
  int j;
  char string[1000];
  char* data = NULL;
  FILE *f;
  f = fopen(fileName, "r");

  while(fgets(string, sizeof(string), f) != NULL){
    data = strtok(string, " ");
    for (j=0; NULL != data && j<nCol; j++){
        if (data[strlen(data) - 1] == '\n')
            data[strlen(data) - 1] = '\0';

        if (j!=0){
          X[i] = atof(data);
          i++;
        }
        data = strtok(NULL, " ");
    }
  }

  for (i=0;i<(nRow*(nCol-1));i++){
    printf("%f\n", X[i]);
  }
  return 0;
}

Helper functions...

int getCol(char *myStr){
    int length,i,count=0;
    char prev;
    length=strlen(myStr);
    if(length > 0){
      prev = myStr[0];
    }
    for(i=0; i<=length; i++){
      if(myStr[i]==' ' && prev != ' '){
        count++;
      }
      prev = myStr[i];
    }
    if(count > 0 && myStr[i] != ' '){
        count++;
    }
    return count;
}

int getRow(char *fileName){
  char ch;
  int count=0;
  FILE *f;
  f = fopen(fileName, "r");

  while(!feof(f)){
    ch = fgetc(f);
    if(ch == '\n')
    {
      count++;
    }
  }
fclose(f);
return count;
} 

EDIT:

I now changed to malloc() for X1, XtY, XtX, and ipiv. I also now used multi-dimensional dynamic array for X1. The errors are all gone when I run gcc, except for:

readReg.c: In function ‘main’:
readReg.c:62: error: subscripted value is neither array nor pointer

Below is the updated main function:

int main(){
   FILE *f;
   char myStr[1000];
   int strL;
   int nCol;
   int nRow;
   char *fileName = "reg.dat";
   int i, j, k, n1=nCol, n2=1, info;

   double *X;
   double *Y;
   double **X1;
   double *XtX;
   double *XtY;
   int *ipiv;
   double *temp;

   f = fopen(fileName, "r");
   if (f == NULL) perror ("Error opening file");
   else {
     if (fgets(myStr, 1000, f) != NULL )
       puts(myStr);
     fclose(f);
   }

   strL = strlen(myStr);
   nCol = getCol(myStr);
   nRow = getRow(fileName);
   printf("Sample size and number of predictors are %d and %d respectively.\n", nRow, nCol-1);

   X = (double *) malloc(sizeof(double) * ((nCol-1) * nRow));
   Y = (double *) malloc(sizeof(double) * nRow);
   XtX = (double *) malloc(sizeof(double) * (nCol*nCol));
   XtY = (double *) malloc(sizeof(double) * nCol);
   ipiv = (int *) malloc(sizeof(int) * nCol);

   assignY(nCol, nRow, Y, fileName);
   assignX(nCol, nRow, X, fileName);

   X1 = malloc(nRow * sizeof(double*));
   temp = malloc(nRow * nCol * sizeof(double));
   for (i= 0; i < nRow; i++) {
      X1[i] = temp + (i * nCol);
   }

...
}

Answer 1

You first declare nRow and nCol without assigning any value:

int nCol;
int nRow;

...and now define four variable-length arrays which use nRow and nCol:

int i, j, k, n1=nCol, n2=1, ipiv[nCol], info;
double X1[nRow][nCol], XtX[(nCol) * (nCol)], XtY[nCol];

Firstly, variable-length arrays are forbidden in C90. Use malloc() for these. You seem to know how to use malloc() based on your code.

Secondly, how do you expect the compiler to know how large the arrays are as you declare the arrays before assigning any value to nCol and nRow ?

One of your arrays is multidimensional. See How do I work with dynamic multi-dimensional arrays in C? for more information how to work with dynamic multidimensional arrays.