mat4js

JavaScript library to load Matlab Level 5 MAT-files as JavaScript objects. Based on the documentation: https://www.mathworks.com/help/pdf_doc/matlab/matfile_format.pdf released for 2019b

Installation

via npm

$ npm install mat-for-js

then use as

import { read as readmat } from "mat-for-js"
readmat(ArrayBuffer)

import * from "mat-for-js"
mat4js.read(ArrayBuffer)

via script tag

Download dist/mat4js.read.min.js to your webserver and include it in your HTML:

<script type="text/javascript" src="./mat4js.read.min.js"></script>

Read

Use mat4js.read(ArrayBuffer).

It returns a JavaScript object with a .header and .data property.

.header

contains 116 bytes of text.

Padding characters are not truncated.

.data

is an object containing the named arrays in the file. ({name: content})

Matlab stores vectors as 2D but flat arrays (1xn or nx1). These are converted to plain 1D arrays ([1, 2, 3] instead of [[1, 2, 3]]) or strings in case of character arrays. In multidimensional character arrays the characters are not concatenated to form strings, but are kept separate. ({char_array: [["a", "b", "c"], ["d", "e", "f"]]} instead of {char_array: ["abc", "def"]})

Numeric arrays with Int64 or Uint64 data types are converted into an array of BigInt.

Numeric arrays with imaginary component are converted into an array of objects with .r and .i properties for the real and imaginary components respectively.

Scalar structs are converted into objects as would be expected, while non-scalar structs are more akin to cell arrays. For a struct constructed with

S = struct()
S.cell = {1 2; 3 4}
S.fruit = 'apple'

In JavaScript

S.cell[1][0] == 3
S.fruit == "apple"

But for one constructed with S = struct('cell', {1 2; 3 4}, 'fruit', 'apple') (non-scalar):

S[1][0].cell == 3
S[1][1].cell == 4
S[1][0].fruit == "apple"
S[1][1].fruit == "apple"

and S.fruit == "apple" is not accessible.

Sparse arrays are converted to objects with .x and .y properties describing the width and height of the array respectively, and an .nz property containing an array of objects representing non-zero values. These objects also have .x and .y properties for their indices in the matrix and an additional .v for the non-zero value at the index.

Build

If you want to rebuild dist/readmat.min.js, first install the dev dependencies with npm install. Once webpack is installed, run npm run build.

Limitaions

There is no support for Object array types. 64-bit integer type support depends on using a version of JavaScript that includes the DataView.getBigInt64()/DataView.getBigUint64() methods.

Matlab v7.3 MAT-files use HDF5 data structure and are not supported. Reading such files will throw a FeatureError with the .feature property set to "HDF5". There are other JavaScript projects to view HDF5 files: /~https://github.com/usnistgov/jsfive

There is no write functionality.