Read 2D float array from binary, cast from char* to float[2]

Question

I have a 2D array of floats stored in a binary file. I read it out using:

#include <fstream>
#include <algorithm>
....
ifstream infile(fileName.c_str(), ios::in | ios:binary);
const int numCols = 2;
char rows[1 * sizeof(int)];
infile.read(rows, sizeof(int));
int numRows = * reinterpret_cast<int *>(rows);
char * temp;
temp = new char[numRows * numCols * sizeof(float)];
infile.read(temp, sizeof(float) * numRows * numCols);

I then created the float array and used memcpy to copy the information over.

float binData[numRows][numCols];
memcpy(binData, temp, sizeof(float) * numRows * numCols);

This approach works (binData contains what I want), but I'm being told not to use memcpy. So I changed to use this:

float * ftemp = reinterpret_cast<float *>(temp);
int loc;
for (int i = 0; i < numRows; ++i){
    for(int j = 0; j < numCols; ++j){
        loc = i * numCols + j;
        binData[i][j] = ftemp[loc];
    }
}

I tried using copy with ftemp and binData but I got this error:

incompatible types in assignment of 'float' to 'float [2]'

It seems overly complicated to use that double loop. Since the information is stored in the correct order in the char array there should be a way to say just read this as a float [numRows][numCols] or just copy the memory held in temp to binData but I can't figure it out. Is there a way to directly convert the char [] to a float [][] or is there an alternate char setup that will allow me to copy directly?

`char temp[numRows * numCols * sizeof(float)];` This is not legal C++. — PaulMcKenzie, Jan 25 '16 at 18:11
There are a few pointers and arrays and casts that are not really needed there, as well as you using non-standard functionality with your [variable-length arrays](https://en.wikipedia.org/wiki/Variable-length_array). Oh, and some breaking of the [string aliasing rule](http://stackoverflow.com/questions/98650/what-is-the-strict-aliasing-rule). — Some programmer dude, Jan 25 '16 at 18:12
Unless this code is speed critical (reading from files seldom is) I suggest you go the *simple* route, and use C++ functionality the way they were meant to be, using [standard containers](http://en.cppreference.com/w/cpp/container), [standard algorithms](http://en.cppreference.com/w/cpp/algorithm), [standard I/O functionality](http://en.cppreference.com/w/cpp/io) and [standard iterator helpers](http://en.cppreference.com/w/cpp/iterator). — Some programmer dude, Jan 25 '16 at 18:16
@PaulMcKenzie interesting...I compiled this on SUSE with GCC 4.3.4 and it worked. When I copied it into VS 2013 it complained about that line. I updated it. Thanks. — Matt, Jan 25 '16 at 18:57
@Matt -- You jumped from the frying pan into the fire with the new code. As suggested, use `std::vector` and `std::vector>` , instead of `new [ ]` (and `delete [ ]`). — PaulMcKenzie, Jan 25 '16 at 19:01

Isaac Clancy · Accepted Answer · 2017-01-02T21:46:24.950

I'd just read straight into the array like this

#include <fstream>
#include <algorithm>
....
ifstream infile(fileName.c_str(), ios::in | ios:binary);
const int numCols = 2;
int numRows;
infile.read(reinterpret_cast<char *>(&numRows), sizeof(int));

float (*binData)[numCols] = new float[numRows][numCols];
//or float binData[numRows][numCols]; if you knew it can fit on the stack 
//and your compiler has variable sized arrays

//read into the array
infile.read(reinterpret_cast<char *>(binData), sizeof(float) * numRows * numCols);

Read 2D float array from binary, cast from char* to float[2]

1 Answers1