Reading Older Mat file versions?

[Perry Taylor]

I am trying to read an older .mat file that I believe is saved as a V.4 style. I am using the latest version of the csmatio library and when I attempt to import it I get a "This is not a valid MATLAB 5.0 MAT-file." error with the std. code and if I comment out that check in the demo program I get a "This version of CSMatIO does not support Big Endian." error. Do you have any thoughts as to how to read the file?

[Perry Taylor]

I have included a file that errors out to help reproduce the problem.

[David A. Zier]

The current version of CSMatIO only supports the MATLAB 5.0 MAT-file. I can put this on backlog of features that we would like to add, but it will not be added anytime soon. Right now, I am in the process or cleaning up the code base and increasing test covereage. After that is done, I will look into adding new features.

[Tobias]

Read the version 4 mat file with Matlab, then save it as version 6 or 7 mat file. Version 6 and 7 should be readable for csmatio.
https://de.mathworks.com/help/matlab/import_export/mat-file-versions.html

[Perry Taylor]

David/Tobias- Thanks for your responses. Unfortunately opening the files up in matlab each time and saving them as a different format is a non-starter for the project I am working on. Do either of you know of any other projects to read Matlab v4 files in either c# or c(++)?

[David A. Zier]

There is none that I know of. Try looking for a matlab file converter. It has been a long time since I actually used Matlab, but when I did, I remember that it had a decent scripting language. You could write a Matlab script that reads in an old file and writes a new file. You could then have this script called just before importing into your C# program or just after generating the file.

[Tobias]

The V4 format is much easier than the latter fomats and straight forward to read without a special library. This QND code example might get you started. It produces this output for your file 'DoublePendulum.mat':

4 x 11 text matrix named 'Aclass'
45 x 1097 text matrix named 'name'
125 x 1097 text matrix named 'description'
4 x 1097 int32 matrix named 'dataInfo'
833 x 2 float32 matrix named 'data_1'
68 x 502 float32 matrix named 'data_2'

Source code:

:::C#
namespace MatlabV4Reader
{
    using System;
    using System.Collections.Generic;
    using System.IO;

    static class Program
    {
        static void Main(string[] args)
        {
            var br = new BinaryReader(File.Open("DoublePendulum.mat", FileMode.Open));
            while (ReadMatrix(br)) { }
            Console.ReadLine();
        }

        enum DataType : int
        {
            float64 = 0,
            float32 = 10,
            int32 = 20,
            text = 51,
        }

        private static bool ReadMatrix(BinaryReader br)
        {
            if (br.BaseStream.Position == br.BaseStream.Length)
                return false;

            int type = br.ReadInt32();
            int mrows = br.ReadInt32();
            int ncols = br.ReadInt32();
            int imagf = br.ReadInt32();
            int namlen = br.ReadInt32();
            string name = br.ReadCString();
            var dataType = (DataType)type;

            List<object> elems = null;
            switch (dataType)
            {
                case DataType.float32: elems = ReadMatrixDataFloat32(br, mrows, ncols); break;
                case DataType.int32: elems = ReadMatrixDataInt32(br, mrows, ncols); break;
                case DataType.text: elems = ReadMatrixTextData(br, mrows, ncols); break;
                default: throw new NotImplementedException(type.ToString());
            }

            Console.WriteLine("{1} x {2} {3} matrix named '{0}'", name, mrows, ncols, dataType);
            //Console.WriteLine("--> element 0: {0}", elems[0]);
            //Console.WriteLine("--> element 1: {0}", elems[1]);
            //Console.WriteLine("--> element 2: {0}", elems[2]);
            //Console.WriteLine("----------------------------------------------------------------");
            return true;
        }

        private static List<object> ReadMatrixTextData(BinaryReader br, int mrows, int ncols)
        {
            var elems = new List<object>(mrows);
            for (int i = 0; i < mrows; ++i)
                elems.Add(br.ReadNonTerminatedCString(ncols));
            return elems;
        }

        private static List<object> ReadMatrixDataInt32(BinaryReader br, int mrows, int ncols)
        {
            var elems = new List<object>(mrows * ncols);
            for (int i = 0; i < mrows * ncols; ++i)
                elems.Add(br.ReadInt32());
            return elems;
        }

        private static List<object> ReadMatrixDataFloat32(BinaryReader br, int mrows, int ncols)
        {
            var elems = new List<object>(mrows * ncols);
            for (int i = 0; i < mrows * ncols; ++i)
                elems.Add(br.ReadSingle());
            return elems;
        }

        public static string ReadCString(this BinaryReader br)
        {
            var chars = new List<char>();
            char c = '\0';
            do { c = br.ReadChar(); chars.Add(c); } while (c != 0);
            chars.Remove('\0');
            return new string(chars.ToArray());
        }

        public static string ReadNonTerminatedCString(this BinaryReader br, int len)
        {
            var chars = br.ReadChars(len);
            return new string(chars);
        }
    }
}

[Tobias]

The spec is here: https://www.mathworks.com/help/pdf_doc/matlab/matfile_format.pdf