nephics / mat4py Goto Github PK

ValueError: Struct field names are too long

Hello,
thank you for your library. I would like to ask you for little modification. Nowadays it gives me two Deprecation Warnings. Namely:
loadmat.py:14 /mat4py/loadmat.py:14: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated since Python 3.3,and in 3.9 it will stop working from collections import Sequence
and
mat4py/savemat.py:15 mat4py/savemat.py:15: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated since Python 3.3,and in 3.9 it will stop working from collections import Sequence, Mapping

It could be fixed be replacing imports from
from collections import Sequence
to
try: from collections.abc import Sequence except ModuleNotFoundError: from collections import Sequence
and from
from collections import Sequence, Mapping
to
try: from collections.abc import Sequence, Mapping except ModuleNotFoundError: from collections import Sequence, Mapping
Or just check: https://stackoverflow.com/a/53978543
Could you, please, fix this issue and provide new version into pipy?
Thank you.

As from issue's title. This simple code:

from mat4py import savemat
import numpy as np

data = np.ones((10,1))
savemat('./test.mat', {'data':data})

raises an error. The error arises from numpy arrays not being an instance of Sequence. For the purposes of the library, it doesn't seem that strict Sequence requirements are necessary, so checking for hasattr(array, '__len__') and hasattr(array, '__getitem__') or isinstance(array, Iterable) should be enough to parse simple 2d arrays. What do you think @jcbsv @tirkarthi?

It would be nice if you provided the source distribution of your package on pypi.

This would make repackaging to conda easier.

Should mat4py be able to load datetime arrays ? I get the following error when trying to do so:
mat4py.loadmat.ParseError: Got type 1, expected 5

Is version 5 not supported?
mat4py.loadmat.ParseError: Expecting miMATRIX type number 14, got 1399801946

Saving a simple test struct created in Octave

a.b = 1;
save -v7 test.mat a

and loading with loadmat

loadmat('test.mat') 

returns the error

\temp\loadmat.py in read_struct_array(fd, endian, header)
    315     if field_name_length > 32:
    316         raise ParseError('Unexpected field name length: {}'.format(
--> 317                          field_name_length))
    318
    319     # read field names

ParseError: Unexpected field name length: 64

scalars work fine though, and importing structs from Matlab instead of Octave (as well as loading the test.mat file into Matlab).

This package is great! It provides something so much more usable than the default scipy.io.loadmat. But it does seem to struggle on a lot of newer matlab formats. I noticed there's a mat73 project that works great for reading the HDF5 format for .mat files in version 7.3. But it doesn't write them. At any rate, it would be nice if this project could be updated to handle newer formats and the effort not duplicated too much.

https://github.com/skjerns/mat7.3