Summary
Reading an MRC file with mrcfile.open() triggers a DeprecationWarning on NumPy 2.5 because MrcInterpreter._read_header() assigns to header.dtype on a structured/record array.
This is distinct from the NumPy 2.4 strides deprecation addressed in #73, and from the broader NumPy 2.4 compatibility report in #74.
Environment
mrcfile 1.5.4 (latest release)numpy 2.5.0- Python 3.12.13
Minimal reproduction
import warnings
import numpy as np
import mrcfile
warnings.simplefilter("error", DeprecationWarning)
data = np.arange(24, dtype=np.float32).reshape(2, 3, 4)
with mrcfile.new("test.mrc", data=data, overwrite=True) as mrc:
mrc.voxel_size = (1.2, 1.3, 1.4)
mrc.header.origin = (5.0, 6.0, 7.0)
with mrcfile.open("test.mrc", permissive=True, mode="r") as mrc:
_ = mrc.dataWarning / traceback
DeprecationWarning: Setting the dtype on a NumPy array has been deprecated in NumPy 2.5.
Instead of changing the dtype on an array x, create a new array with x.view(new_dtype)
Traceback (most recent call last):
File "<stdin>", line 11, in <module>
File ".../mrcfile/load_functions.py", line 145, in open
return NewMrc(name, mode=mode, permissive=permissive,
File ".../mrcfile/mrcfile.py", line 115, in __init__
self._read(header_only)
File ".../mrcfile/mrcinterpreter.py", line 170, in _read
self._read_header()
File ".../mrcfile/mrcinterpreter.py", line 221, in _read_header
header.dtype = header.dtype.newbyteorder(byte_order)
File ".../numpy/_core/records.py", line 462, in __setattr__
ret = object.__setattr__(self, attr, val)
Root cause
In mrcinterpreter.py (_read_header), after determining byte order:
header.dtype = header.dtype.newbyteorder(byte_order)
NumPy 2.5 deprecates mutating .dtype on record arrays. The warning suggests creating a new view instead.
Suggested fix
Replace in-place dtype mutation with a view-based update, e.g.:
new_dtype = header.dtype.newbyteorder(byte_order)
header = header.view(new_dtype)
(or an equivalent approach that avoids assigning to header.dtype).
Impact
- Currently a warning under default pytest settings.
- With
warnings.simplefilter("error", DeprecationWarning) (or stricter CI configs), reading MRC files fails.
- Downstream packages that round-trip MRC files (e.g. via
mrcfile.open after mrcfile.new) will see this on NumPy 2.5+.
Happy to help test a fix or open a PR if useful.
Summary
Reading an MRC file with
mrcfile.open()triggers aDeprecationWarningon NumPy 2.5 becauseMrcInterpreter._read_header()assigns toheader.dtypeon a structured/record array.This is distinct from the NumPy 2.4 strides deprecation addressed in #73, and from the broader NumPy 2.4 compatibility report in #74.
Environment
mrcfile1.5.4 (latest release)numpy2.5.0Minimal reproduction
Warning / traceback
Root cause
In
mrcinterpreter.py(_read_header), after determining byte order:NumPy 2.5 deprecates mutating
.dtypeon record arrays. The warning suggests creating a new view instead.Suggested fix
Replace in-place dtype mutation with a view-based update, e.g.:
(or an equivalent approach that avoids assigning to
header.dtype).Impact
warnings.simplefilter("error", DeprecationWarning)(or stricter CI configs), reading MRC files fails.mrcfile.openaftermrcfile.new) will see this on NumPy 2.5+.Happy to help test a fix or open a PR if useful.