Changelog#

2024.08 (Sep 03, 2024)#

Added#

discretisedfield

  • Added support for visualisations using pyvista. (#463)

micromagneticmodel

  • Support for current in an arbitary direction when using the Zhang Li term. (#75)

mumax3c

  • Support for current in an arbitary direction when using the Zhang Li term. (#88)

  • Support for simulations at a finite temperature. (#81)

oommfc

  • Support for tunable (non-symmetric) Hysteresis simulations with multiple segments (with variable step size) between user-defined field extrema. (#138)

  • Support for current in an arbitary direction when using the Zhang Li term. (#133)

Changed#

micromagneticmodel

  • Require the name when initialising the system. (#85)

Fixed#

micromagneticdata

  • Set valid="norm" when loading the initial magnetisation drive.m0. (#69)

all packages

  • Full compatibility with numpy >= 2.0.

2024.06 (Jun 19, 2024)#

Fixed#

Restricted numpy version to be less than 2.0 due to incompatibilities with the current code base.

2023.11 (Nov 25, 2023)#

Changed#

discretisedfield

  • The property mesh.points has been renamed to mesh.cells to avoid ambiguity. (#508)

Fixed#

discretisedfield

  • Reading OVF files failed when the header used lowercase letters in the “Begin: data” section. (#513)

micromagneticdata

  • Reading magnetisation output from Mumax3 fails when region.pmin is not defined at the origin and subregions are defined for system.m. (#60)

2023.10 (Oct 23, 2023)#

The meta-package ubermag now uses calendar-based versioning. We use a style that is broadly inspired by Ubuntu: we use the year and the month and optionally a patch (YYYY.0M[.patch]). We omit the number for patch 0 and only add and increase patch when there are two releases in a month. The versioning is not related to the type of changes (and a patch does not necessarily mean a bug fix). For the type of changes please refer to this changelog or the versions of the individual packages, which follow semantic versioning.

Breaking changes in discretisedfield#

discretisedfield has been refactored to support scalar and (m-dimensional) vector fields on n-dimensional meshes. As part of this refactoring, several inconsistencies and ambiguities have been removed. These changes break existing code! The following summary lists all user-facing changes (removed, renamed, added functionality). Even though the list below is fairly long, most notebooks only require a few changes. The most common ones are the switch dim -> nvdim and adding valid='norm' when creating a field, and the replacement plane -> sel.

The shape of the domain (i.e. the shape of the region/mesh) is now referred to as dim, the shape of the field values as vdim (vector/value dimension).

Region

  • Modified: scaling a region is now done with the method scale, the option to scale a region by multiply it with a number has been removed. Scaling can optionally be done in-place.

  • New: regions can be translated in space with translate. Translating can optionally be done in-place.

  • Modified: a new method allclose checks if two regions are the same within numerical accuracy, the operator now == checks for exact equality.

  • Modified: the operator | has been replaced with the new method facing_surface

  • Removed: properties p1 and p2, use pmin and pmax

  • Changed: the return type of points (e.g. pmin, pmax, centre) is now a numpy.ndarray

  • Changed: unit has been renamed to units and is a tuple to supports different units along different directions (only used as labels on plots, not for calculations).

  • New: property ndim to get the number of dimensions of the space.

  • New: property dims, names of the spatial dimensions.

  • New: method rotate90 to perform 90 degree rotations of the region.

Mesh

  • Modified: a new method allclose checks if two regions are the same within numerical accuracy, the operator now == checks for exact equality.

  • Modified: inconsistent behaviour of mesh[subregion]; this operator will now always use closed intervals, i.e. the returned submesh is now always inclusive of endpoints.

  • Modified: the operator | has been replaced with a new method called is_aligned.

  • Removed: axispoints (was deprecated), use points instead.

  • Removed: neighbours

  • Renamed and modified: plane has been replaced with a new sel method. (for details refer to the Field documentation)

  • Removed: attributes

  • Removed: dS; was used for integrals; the new syntax for integrals does not any longer use it (see below).

  • Renamed: midpoints to points

  • Modified: new method coordinate_field that was previous part of the Field class.

  • New: scale and translate similar to the Region class.

  • New: method rotate90 to perform 90 degree rotations of the mesh.

Field

  • Modified: getting the value at a point (field(...)) now returns a numpy.ndarray.

  • Modified: only field values are returned when iterating over a field. (documentation)

  • Removed: coordinate_field is now part of the Mesh class.

  • Renamed and modified: derivative -> diff; now supports holes inside the field. (documentation)

  • Modified: new integration syntax. (documentation)

  • Renamed: fromfile -> from_file

  • Renamed and modified: plane has been replaced with a new sel method. (documentation)

  • Removed: project has been removed, use mean with the axis argument instead.

  • Renamed: write -> to_file

  • Modified: angle now takes a vector as reference and does not anymore require sliced fields.

  • Renamed: dim -> nvdim to avoid ambiguities, the new name is also required as an argument when creating a field.

  • Renamed: components -> vdims

  • Renamed: units -> unit

  • Removed: value, to update field values use the new method update_field_values instead, to read data use the array property.

  • Removed: zero class method; omit the value argument to create a field filled with zeros (was already supported before).

  • New: method resample to compute field values on a new mesh; the nearest points will be used, interpolation is not supported. (documentation)

  • New: property valid to define parts that contain material; this can be set to the field norm on init using the optional valid='norm' argument. (documentation)

  • Modified: new hdf5 format to store all field information (e.g. subregions were missing before). (documentation)

  • New: Documentation for performing FFT on the Field class (documentation).

  • New: method rotate90 to perform 90 degree rotations of the field. (documentation).

  • Modified: Differential operators grad, div, curl, and Laplacian have been generalised to n dimensions.

  • Plotting with holoviews now uses field.valid to hide data (e.g. based on their norm). The norm_filter property as well as the Defaults class for plotting have been removed. Use valid="norm" when initialising the field. (#475)

Added#

ubermag

  • Debug function to show versions of all ubermag subpackages and related information. (#147, #148)

mumax3c

  • Support for multiple Zeeman fields in the energy equation. (#49)

Changed#

ubermagutil

  • An additional summary line is printed for the progress bar context manager. Users will see the additional line when using oommfc or mumax3c time drives or min drives with verbose=2. This was added because the progress bar in the notebook is not persistent across sessions. (#39)

Fixed#

micromagneticmodel

  • Wrong value of \(\mu_B\). (#79)

mumax3c

  • Bug in mumax3 detection when passing the full path to the mumax3 executable. (commit b2c334d)

  • Bug in current conversion for Zhang-Li. (#73)

oommfc

  • Bug in RKKY energy script when not defining sigma2. (#142)

  • Bug in time varying J and tcl injection script of Xf_ThermSpinXferEvolver. (#136)

0.66.1 (Feb 3, 2023)#

Fixed#

discretisedfield

  • The ovf2vtk command line tool was not directly accessible. (#219)

  • OVF files saved with the option extend_scalar=True did not contain all data when the mesh size exceeded 100,000 elements. (#337)

0.66.0 (Oct 20, 2022)#

Added#

discretisedfield

  • Reading and writing for Field objects now supports pathlib.Path objects. Furthermore, the code has been reorganised (mostly invisible for end users, a new io submodule has been added). (#158)

  • Subregions are saved in a json file (<field-name>.<extension>.subregions.json) when saving a field and automatically loaded when a field is loaded from disk and the json file exists. (#158, #163)

micromagneticdata

  • Subregion information is loaded when accessing magnetisation fields (if available). (#35)

  • Callbacks can be registered in the drive object to apply any sort of processing (e.g. calculating the normalised field or topological charge density) before returning individual fields in a drive. (#35)

  • Support for slicing in getitem in Drive, e.g. drive[::2] will return a new drive object that only contains every second magnetisation file. (#35)

oommfc

  • Subregion information is loaded when using compute (if available). (#116)

Changed#

discretisedfield

  • Plotting for Mesh class moved to new submodule similar to Region and Field. Therefore mesh.mpl_subregions has to be replaced with mesh.mpl.subregions. (#164)

  • New implementation for Holoviews-based plotting to improve creation speed for large objects. (#194, #196)

micromagneticdata

  • Holoviews plotting now reads data from disk when it is accessed in the plot. This greatly improves the initial rendering of the plot and can avoid memory problems. Note that on slow file systems updating the plot when moving a slider might take more time. (#35)

Fixed#

discretisedfield

  • Missing initialisation when computing field.orientation for a field with zero norm in some cells. (commit 4a8fca4)

0.65.0 (Jul 17, 2022)#

Changed#

discretisedfield

  • Import discretisedfield.tools into discretisedfield. (#159)

  • Scale the size and shape of the colorbar on mpl plots so the colorbar dynamically changes with axes size (#159)

micromagneticdata

  • Rewrite of the to_xarray method to improve performance and reduced memory consumption. The maximum memory consumption is now roughly equivalent to the on-disk size of the data. The old method had a roughly doubled peak memory consumption. (#33)

micromagneticmodel

  • Fix the LaTex representation of Landau-Lifshitz-Gilbert equation in the presence of Zhang-Li and Slonczewski torque terms. (#52)

0.64.0 (Jul 03, 2022)#

mumax3c

A new package mumax3c has been added and allows users to use mumax3 as a new calculator in addition to OOMMF.

0.63.0 (Jun 29, 2022)#

Added#

discretisedfield

  • Support for interactive plots based on Holoviews with Field.hv convenience method, Field.hv.scalar, Field.hv.vector, and Field.hv.contour. Refer to the package documentation for details. (#140)

micromagneticdata

  • Two new drive classes OOMMFDrive and Mumax3Drive have been added. Users should not create these directly. Instead the generic Drive class should be used which automatically creates an object of the correct sub-type depending on the information found for the respective drive. (#29)

  • Interactive Holoviews-based plotting for drives with drive.hv property. It provides the same functionality as Field.hv. (#26)

micromagneticmodel

  • User-defined name for a Zeeman energy term is displayed in its LaTex representation in the Jupyter notebook. (#45)

  • Containers (system.energy and system.dynamics) have a new method get(type=...) to get all terms in the container that have the specified type. (# 50)

  • Containers (system.energy and system.dynacmics) now support container[index] (__getitem__) to obtain a term by its index. (#50)

oommfc

  • The input for OOMMF (mif file and related files) can now be created without starting the actual simulation using Driver.write_mif(...). (#104)

  • A drive (call to OOMMF) can be sent to a scheduling system such as Slurm using Drive.schedule(...). All input files (mif file and related files) are created from the running Python program. The user must pass the submission command of the scheduling system and a “header” file that defines system resources as required for the scheduling system. Furthermore, it is the user’s responsibility to ensure that OOMMF is available inside the scheduled job (e.g. by loading the correct conda environment in the header file). The command line to call OOMMF is added to the header file, the file is saved to the drive directory and subsequently a job is submitted to the scheduling system. (#104)

  • The OOMMFRunner classes take an additional argument dry_run to return the OOMMF command call instead of calling OOMMF from a subprocess. (#104)

ubermagutil

  • Context manager to change directories. (#29)

  • Generic utilities to show progress information for calculators. Users generally don’t use any of this functionality directly. (#31)

Changed#

discretisedfield

  • Vector fields with zero values can now be normalised. The rescaling of the vectors is only done for non-zero values. (#149)

mag2exp

  • All functions do now operate on discretisedfield.Field objects instead of micromagneticmodel.System objects. Therefore, measurements can be simulated for data that has not been created with a micromagnetic simulation as long as the data can be represented as a discretisedfield.Field object. (#31)

micromagneticmodel

  • Comparison of micromagneticmodel.Term objects (e.g. micromagneticmodel.Exchange) now takes into account all attributes of the two terms. Previously, only the names of the two terms were compared. (#46)

oommfc

  • The overhead function now runs the simulations in a temporary directory. (#110)

Fixed#

discretisedfield

  • Wrong normalisation of the lightness component in lightness plots if not all angles are covered. (commit 2de6360)

  • Checks if a point is inside a Region (point in region) give wrong results if the points are inside the region in some spatial direction and at the region boundary (outside within the numerical precision) in the other spatial directions failed despite the point being in the region. (#153)

micromagneticdata

  • data[index] returns a drive for drive numbers index larger than the maximum drive number on disk (by computing module of index). (#24)

oommfc

  • The progress bar did not stop on KeyboardInterrupt. (#103)

  • Upon completion (interrupt) the progress bar counter is set to the correct value instead of the maximum value. (#101)

0.62.1 (Jun 09, 2022)#

Fixed#

discretisedfield

  • Importing discretisedfield fails on some platforms because openh264 is missing. (#146)

0.62.0 (May 22, 2022)#

Added#

discretisedfield

  • Support for exporting Field object to the xarray.DataArray with to_xarray instance method (#123).

  • Support for initiating Field object from the xarray.DataArray with from_xarray class method (#123).

  • New implementation for initialising a field from an other field with much better performance (example: initial field with n=(100, 100, 10), new field with n=(10, 10, 10), speedup ~1000). This method does e.g. also speed up the Field.plane method (#135).

  • The __contains__ operator for the region (the in operator region1 in region2) uses a new method to calculate the distance between the points to improve stability when edge points of the regions are very close and differences are mainly related to floating-point inaccuracies. This is done by introducing a new region property tolerance_factor (with a default value 1e-12) that is internally multiplied with the minimum of the edge lengths to obtain values for relative and absolute tolerance (#135).

  • Refactoring of the matplotlib-based plotting methods for df.Field (df.Field.mpl...) improve performance of the plot creation (#133).

  • Refactoring of df.Region.mpl to show the correct aspect ratio of the region. This automatically also applies to the matplotlib-based plotting methods of df.Mesh. A new keyword box_aspect (default box_aspect='auto') can be used to set an arbitrary aspect ratio by passing a tuple. For the default value 'auto' the aspect of the region is used. You may run into problems with overlapping ticks or axis labels. These cannot currently be dealt with easily in an automatic fashion and require manual adjustment after the plot is created (by passing an axis object) if proper axis ticks and labels are required (#134).

  • New implementation for reading and writing vtk files. Files can now be written in textual (txt), binary (bin) or xml (xml) representation. The field data is now stored to vtk cell data (previously: point data). Some subsequent plotting operations in other tools (e.g. plotting isosurfaces) typically requires a conversion from cell data to point data in that tools first (e.g. by using a filter in Paraview). discretisedfield can still read the old vtk files (with the values stored as point data) and tries to do that automatically when the provided file can’t be read with the new method. In particular for writing huge speedups compared to the old implementation are possible (example: n=(200, 200, 100) written in binary representation, speedup > 1000). By default data is written in binary format. (#129)

  • A units property has been added to the Field class to store the units of the field values. Note, that units are mostly used for plotting and are removed during all mathematical operations. Units are automatically read from ovf files if present. (#141)

  • An additional class method coordinate_field has been added to the Field class. It takes a mesh and creates a 3d vector field with values equal to the coordinates of the respective cell midpoints. (#144)

micromagneticdata

  • Support for converting all magnetisation data of a Drive into an xarray.DataArray with to_xarray instance method. (#18)

  • Multiple drives of the same type (with the same independent variable drive.x can be combined into a new micromagneticdata.CombinedDrive. Combining is also supported via drive1 << drive2 which “appends” drive1 to drive2. The combined drive allows iteration over all magnetisation snapshots of the individual drives. (#22)

oommfc

  • When using the TimeDriver with verbose=2 a simple status bar is displayed to show the simulation progress. Note, that the shown information is only a rough hint as the progress is measured based on the total number of steps n and the number of files already written to disk. (#100)

ubermagutil

  • Utility functionality for setting up basic logging moved to ubermagutil from ubermag. (#27, #133)

Removed#

ubermag

  • Utility functionality for setting up basic logging moved to ubermagutil from ubermag. (#27, #133)

Fixed#

discretisedfield

  • Changes in the calculation of the demag tensor to avoid zero-division problems and nan values in the demag field. (#137)

oommfc

  • Using the DockerOOMMFRunner did not work in combination with SELinux because the directiories inside the docker image were not readable/writable. The DockerOOMMFRunner now has an additional optional argument to enable read/write access with SELinux (#95).

0.61.2 (Mar 17, 2022)#

Fixed#

oommfc

  • The old docker image of oommf (2.0a2) was used by default (#84).

  • Two time-dependent Zeeman terms have the same time dependence and/or result in a runtime error (#198).

0.61.0 (Feb 22, 2022)#

General#

  • Support Cnv, D2d, and T(O) crystallographic class DMI and magneto-elastic (MEL) extensions in conda-installed Ubermag on Windows, in addition to MacOS and Linux.

Added#

discretisedfield

  • Support for filter_field in plotting method Field.mpl.lighness.

  • HTML representation for Region, Mesh, Field, and FieldRotator inside Jupyter notebook (#105).

  • Datatype for Field can be specified (#114, #118).

  • New implementation for the Field initialisation significantly improves performance when initialising a field with a dictionary. Speedups of up to ~10_000 can be obtained if the dictionary does not contain callables. (#114, #117)

  • New implementation for reading and writing ovf (omf) files with huge performance improvements. For a Field containing 1 million cells we obtained the following execution times (on a standard notebook):

    mode

    filesize

    reading

    writing

    old

    new

    speedup

    old

    new

    speedup

    bin4

    2.9M

    1730 ms

    21 ms

    82

    63000 ms

    56 ms

    1125

    bin8

    5.8M

    1860 ms

    52 ms

    52

    64000 ms

    84 ms

    762

    text

    15M

    4920 ms

    401 ms

    12

    69000 ms

    4510 ms

    15

    The new default is bin8 (binary represenation with double precision) instead of txt (#121).

oommfc

  • Enables Cnv, D2d, and T(O) crystallographic class DMI and magneto-elastic (MEL) extensions on Windows hosts (no more need for Docker). ExeOOMMFRunner can be used on Windows. Limitation: On Windows it is not possible to run multiple simulations in parallel without using Docker.

  • OOMMF output is now by default written in binary format (double precision). There is a new option in the driver’s drive method (ovf_format) to change the output format (#77).

  • OOMMF can now run silently without anything printed. To use it pass the option verbose=0 to <DRIVER>.drive. The default is verbose=1 which prints one summary line about the used runner and the runtime. This is the old behaviour. (#80).

Changed#

discretisedfield

  • Keywords for Field.mpl() renamed to scalar_kw and vector_kw (#108).

micromagneticmodel

  • Variable names for time-dependent fields and currents changed (for consistency reasons).

Fixed#

discretisedfield

  • Simultaneous use of filter_field and symmetric_clim in Field.mpl.scalar resulted in wrong colorbar limits (#106).

oommfc

  • Specifying two Zeeman fields with H defined via a df.Field broke the simulation (#191)

  • The name of the hysteresis field of the HysteresisDriver has been renamed to B_hysteresis. This solves an issue of having two magnetic fields with the same name if a hysteresis simulation is performed with an additional zeeman field.

0.60.0 (Oct 1, 2021)#

General#

  • Unified website containing all documentation: https://ubermag.github.io

  • Versions in sync across all packages

  • Switch to Jupyter lab 3 (should improve situation with K3d inside Jupyter lab).

Added#

discretisedfield

  • Fourier transform for discretisedfield.Field (reference).

  • Custom labels for vector components in discretisedfield.Field (reference).

  • New plotting interface for discretisedfield.Field and discretisedfield.Region for both matplotlib and K3d. Plotting functions can be accessed as Field.mpl.* (and for convenience Field.mpl()) for matplotlib and similarly for K3d.

  • Improved lightness plotting and new contour line plot.

  • Full support for complex values in discretisedfield.Field.

  • Rotations of discretisedfield.Field objects (documentation).

  • discretisedfield.Field now supports all numpy ufuncs.

  • Calculation of the demag tensor and demag field in discretisedfield.tools (reference).

mag2exp

  • New subpackage mag2exp to simulate experimental measurement (documentation).

micromagneticmodel

  • Generalisation of OOMMF extensions DMI_Cnv and DMI_D2d to support grains oriented along x, y, or z (new names, e.g. DMI_Cnv_z) (documentation).

  • Support for arbitrary time-dependence for external magnetic fields (micromagneticmodel.Zeeman) and spin-polarised currents (micromagneticmodel.Slonczewski and micromagneticmodel.ZhangLi) (documentation).

oommfc

  • Support for OOMMF extension Xf_ThermHeunEvolver, Xf_ThermSpinXferEvolver, and UHH_ThetaEvolve for simulations at finite temperature.

  • Control over the default runner in oommfc via oommfc.runner object (documentation).

ubermag

  • Convenient control over logging of all subpackages via ubermag.setup_logging (documentation).

ubermagtable

Fixed#

discretisedfield

  • Wrong colourbar positioning in discretisedfield.mpl* in figures containing multiple subplots.

  • Fixed aspect ratio for quiver plots in discretisedfield.Field.

micromagneticmodel

  • Creating a term micromagneticmodel.Slonczewski twice with the same dictionary for P or Lambda results in a ValueError

oommfc

  • Removing a current term and driving the system caused a TypeError (#135).

  • oommfc.compute now works when current terms are specified in system.dynamics (#139).

  • Wrong compute number in oommfc.

  • oommfc is choosing the wrong runner when using pyenv (#172).

ubermagtable

  • Error in reading ODT files when using magnetoelastic extension (#14).

  • Multiple columns with the same name in ubermagtable (#118).

0.51 (Feb 10, 2021)#

  • New subpackage discretisedfield.tools containing functions to operate on discretisedfield.Field objects.

  • New integration syntax.

On this page
Show Source