Create new MAT-files optimized for partial reading and writing of large arrays.
- Demo at: http://www.allnans.com/newmatic/newmatic_demo.html
- Blog at: http://www.allnans.com/jekyll/update/2020/02/23/newmatic.html
This tool solves a few problems I ran into when using matfile, namely:
-
Inability to control data chunking: Arrays in MAT-files are saved in "chunks". The most natural and performant way to read/write to these arrays is to access one chunk at a time. MATLAB chooses the chunk size automatically, and it may (read: almost certainly) not be the most efficient choice for your workload. The recommended solution is to use third-party tools to "repack" your MAT-file after you create it (see section "Accelerate Save and Load Operations for Version 7.3 MAT-Files" in the MAT-file versions documentation). This should make us all sad.
-
Repetitive and cryptic array size initialization: Assigning a value to the very last element in the array serves a similar purpose as pre-allocating arrays for in-memory variables. The recommended approach is fine, but not easy for the uninitiated to parse and ugly to look at.
Fortunately, (a) MAT-files are HDF5 formatted under the hood, and (b) MATLAB
includes utilities for working with HDF5 files directly (i.e., without any
external tools). newmatic
provides an alternative interface for creating
"customized" MAT-files that perform well for partial IO.
Additional notes:
newmatic
only allocates arrays. If you want to include other data types (strings, cell arrays, whatever), you can add them to the file created bynewmatic
in the usual way.
There are two key functions in this package:
newmatic_variable
is used to define the name, data type, array size, and chunk size for variables created bynewmatic
newmatic
creates a MAT-file with the specified variables
Here is a small example demonstrating how to create a MAT-file with two arrays
(created by newmatic
) and a cell array (not created by newmatic
).
mat = newmatic(...
'my-mat-file.mat', ...
newmatic_variable('x', 'double', [1000, 1000, 20], [1000, 1000, 1]), ...
newmatic_variable('y', 'double', [1000, 10, 10], [5000, 10, 10]));
mat.z = {'a', 'cell', 'array'};
Special shout out to the hdf5storage python package for helping me work out the details of making a MATLAB-compatible HDF5 file.