Merge pull request #1767 from pace-neutrons/1766_bugfix_release

The information about release 4.0.1

VERSION

@@ -1 +1 @@
-4.0.0
+4.0.1

documentation/release_notes/v4.0.1.md

@@ -0,0 +1,18 @@
+## Release Notes v4.0.1
+
+Minor release, fixing some bugs, and dealing with
+some optimization issues identified after release.
+
+## Bugs fixed:
+   The details of the fixed issues are available on
+   [GitHub](https://github.com/pace-neutrons/Horace/issues/)
+ - #1731 invalid error reporting in `load_spe`.
+ - #1735 Issues with `SymopGeneral` in `cut_sym`.
+ - #1745 Missing `dnd` constructor used by `spinW`
+         and `spinW` integration tests.
+ - #1748 Fixed `SymopIdentify` taking substantial
+         time while doing cut without symmetries.
+ - #1753 `horace_mex` and issues with `cpp_communicator`.
+ - #1758 `horace_install` installs `spinW` integration on request.
+ - #1762 `horace_demo_script` is not working
+          due to reference to removed `fit_sqw`

documentation/release_notes/v4.0.md

@@ -1,16 +1,82 @@
 ## Release Notes v4.0
 
-This is major new release, providing internal changes to the sqw/dnd objects, necessary for using generic projection
-
-## Breaking changes
- - Renamed `fake_sqw` to `dummy_sqw`
-
-## Critical changes
- - `DnD` objects are fully refactored and constructed from `axes_block` class (property `axes`), `aProjection` class (property   `proj`) and three datasets, namely `s,e` and `npix` containing signal, error and number of pixels contributed into a bin, as before. `axes_block` defines binning of these datasets and their physical meaning, while projection defines the transformation from pixels coordinate system to the coordinate system, defined by `axes_block`. Old interface (`iax,pax,iint,p etc.`) left for compatibility with existing algorithms and set as read-only interface.
- - `multifit` and `tobyfit` are now able to be run in parallel through the `hpc_config` `parallel_multifit` setting.
- - `data_sqw_dnd` class (property `data` of `sqw` object) have been removed from `sqw` object and replaced by appropriate dimension `dnd` object. Pixels data (`PixelData` class), which were the property of `data_sqw_dnd` class are now independent property of `sqw` object (property `pix`). Old and refactored `data_sqw_dnd` class left for compatibility with old `sqw` files, allowing the loading of old data from `.mat` files.
- - Added benchmarking features under the `_benchmarking` folder which will run on CI or can be run by users using either the `benchmark_horace` script or by running individual benchmark functions. Documentation included in SMG 20
- - Spawning parallel jobs with Herbert/C++ MPI no longer spawns inappropriate MATLAB threads oversubscribing the machine and causing slowdown.
- - Slurm cluster handling has been rewritten to deal appropriately with multiple node requests and automatically handle threading.
- - `threads` and `parallel_workers` have migrated from `hpc_config` to `parallel_config`
- - `herbert_config` has been deprecated and features of `herbert_config` have migrated to `hor_config`
+This is version 4 of Horace, which contains major changes to the Horace code
+produced within 3-years of PACE project development.
+It is major upgrade from the Horace version 3.6.
+
+Draft versions of these changes were used at ISIS in cycles: 2024/1&2
+
+## Major changes.
+
+As a major part of the new features we would like to emphasise the fully restructured sqw objects.
+In particular they provide the following benefits:
+
+- Allows use of different projections: images (`dnd`) in different coordinate systems.
+  - Now provides cuts using spherical and cylindrical projections with more coming.
+
+- On-file sqws for cases where they do not fit in memory, with a unified interface 
+ (no need to learn anything new, for big jobs!).
+
+- New symmetry operations and ways of symmetrizing sqw objects.
+  - Multiple reductions in a single `symmetrize_sqw` call, 
+    rotationally symmetric reduction.
+  - Symmetrised cuts which can be more efficient and more general.
+
+- Extended parallel capabilities
+  - New parallel algorithms for `sqw_eval` and fitting.
+
+- New more efficient and flexible binary format.
+  - **NOTE:** Horace can read previous versions of sqw files, however, sqw
+    files produced by Horace v4+ can not be read by previous versions of Horace.
+
+
+## Breaking Changes
+- Renamed `fake_sqw` to `dummy_sqw`
+
+- Fully refactored `SQW` objects:
+  - `data_sqw_dnd` class (property `data` of `sqw` object) has been removed from
+    `sqw` object and replaced by appropriately dimensioned `dnd` object. 
+  - Pixel data (`PixelData` class), which were a property of the `data_sqw_dnd`
+    class are now an independent property of `sqw` object (`pix` property). 
+  - Old and refactored `data_sqw_dnd` class left for compatibility with old
+   `sqw` files, allowing the loading of old data from `.mat` files.
+
+- Fully refactored `DnD` objects now made up of:
+  - `axes_block` class (`axes` property),  defining binning of
+     these datasets and their physical meaning,
+  - `aProjection` class (`proj` property), defining the transformation
+     from pixels coordinate system to the coordinate system, defined by `axes_block`.
+  - and three datasets, namely `s,e` and `npix` containing signal, error and
+    number of pixels contributed into a bin, as before. 
+  - **NOTE:** Old interface (`iax`, `pax`, `iint`, `p`, etc.) left for
+    compatibility with existing algorithms through read-only interface.
+
+- `multifit` and `tobyfit` can be run in parallel through
+   the `hpc_config` `parallel_multifit` setting.
+
+- Added benchmarking in the `_benchmarking` folder which will
+  run on CI or can be run by users using either the `benchmark_horace`
+  script or by running individual benchmark functions.
+  Documentation included in SMG 20.
+
+- Spawning parallel jobs with Herbert/C++ MPI no longer spawns
+  inappropriate MATLAB threads oversubscribing the machine
+  and causing slowdown.
+
+- Slurm cluster interface has been rewritten to deal appropriately
+  with multiple node requests and automatically handle threading.
+
+- `threads` and `parallel_workers` have migrated from
+  `hpc_config` to `parallel_config`.
+
+- `herbert_config` has been deprecated and features of
+  `herbert_config` have migrated to `hor_config`.
+
+- `fit_sqw` function have been removed in favour of
+   direct usage of `multifit_sqw` class.
+
+- More than 1,000 additional tasks making substantial additions and
+  improvements to algorithms and design: 
+  refactoring and improving existing algorithms, providing
+  new features and making code more modular to facilitate
+  maintenance and future Horace development.