Force little-endian format when writing eeg datafile

CITATION.cff

@@ -23,6 +23,9 @@ authors:
 - given-names: "Phillip"
   family-names: "Alday"
   orcid: https://orcid.org/0000-0002-9984-5745
+- given-names: "Aniket"
+  family-names: "Pradhan"
+  orcid: https://orcid.org/0000-0002-6705-5116
 title: "pybv"
 version: 0.5.0
 date-released: 2021-01-03

README.rst

@@ -60,7 +60,12 @@ consisting of:
 - comments marked as ``; comment``
 - key-value pairs marked as ``key=value``
 
-A documentation for core BrainVision file format is provided by Brain Products.
+The binary ``.eeg`` data file is written in little-endian format without a Byte Order
+Mark (BOM), in accordance with the specification by Brain Products.
+This ensures that the data file is uniformly written irrespective of the
+native system architecture.
+
+A documentation for the BrainVision file format is provided by Brain Products.
 You can `view the specification <https://www.brainproducts.com/productdetails.php?id=21&tab=5>`_
 as hosted by Brain Products.
 

docs/changelog.rst

@@ -7,6 +7,7 @@ Authors
 People who contributed to this software across releases (in **alphabetical order**):
 
 - `Adam Li`_
+- `Aniket Pradhan`_
 - `Chris Holdgraf`_
 - `Clemens Brunner`_
 - `Phillip Alday`_
@@ -33,6 +34,7 @@ Changelog
 ~~~~~~~~~
 
 - :func:`pybv.write_brainvision` gained a new parameter, ``ref_ch_names``, to specify the reference channels used during recording, by `Richard Höchenberger`_ and `Stefan Appelhoff`_ (:gh:`75`)
+- Fix bug where :func:`pybv.write_brainvision` would write the binary eeg file in big-endian on a big-endian system. This could lead to non-uniformity of data-files written by big-endian or little-endian machines, by `Aniket Pradhan`, `Clemens Brunner` and `Stefan Appelhoff`_ (:gh:`80`).
 
 API
 ~~~

pybv/io.py

@@ -14,6 +14,7 @@
 import datetime
 import os
 import shutil
+import sys
 from os import path as op
 from warnings import warn
 
@@ -491,5 +492,11 @@ def _write_bveeg_file(eeg_fname, data, orientation, format, resolution, units):
         raise ValueError(msg)
     data = data.astype(dtype=dtype)
 
+    # Swap bytes if system architecure is big-endian and dtype is "native" or
+    # "big-endian"
+    byteorder = data.dtype.byteorder  # pragma: no cover
+    if (sys.byteorder == "big" and byteorder == "=") or byteorder == ">":  # pragma: no cover  # noqa: E501
+        data = data.byteswap()
+
     # Save to binary
     data.ravel(order='F').tofile(eeg_fname)