Sherpa 4.16.1

This release of Sherpa includes various enhancements, documentation updates, bug fixes, and infrastructure changes.

• enhancements:
  • minor plotting changes; add support for splitting model expression into additive components and plot the results; support of log scale axes for confidence plots; improved error messages for unavailable plot backends
  • improved RMF plot display to allow choice of energy units
• documentation changes:
  • updates to fake_pha documentation
  • updates to install.rst to fix incorrect links and outdated version references
  • updated read the docs documentation to match current code
• bug fixes:
  • fixed multi-panel plot issue with Bokeh backend

Details

#1608 - Remove the test setup.py option Add the test configuration so that we can ensure the necessary testing packages can be installed when the code is built. Support for running the tests with 'python setup.py test' has been removed to match changes in the Python packaging ecosystem.

#1679 - Automatically split a model expression into additive components and plot the results Add plot_source_components and plot_model_components calls that will split up a model like gal * (pl + line) into two lines: one for gal * pl and one for gal * line. There are also corresponding get_source_components_plot and get_model_components_plot calls, and the "source_components" and "model_components" arguments can be used with the plot call.

#1684 -Fix fake_pha docs Fix the documentation and some corner cases of sherpa.astro.fake.fake_pha. There are several times when the simulated PHA output would not be correct if the PHA contains at least one background component or a pileup model was used.

#1926 - Let users know they should use load_xstable_model not load_table_model Support for XSPEC table models in load_table_model was deprecated in the 4.9 release. Ensure users know that they should be using load_xstable_model instead by adding a warning message whenever it is used.

#1934 - Zenodo broke the API we were using to request citation details. Ensure that the sherpa.citation() command can query Zenodo for the release information. Fix #1933.

#1938 - CI: temporarily remove ds9 tests on macOS Avoid the DS9 tests on the macOS build to avoid the failure case we currently often, but not always, trigger.

#1941 - Fix various typos Updates from running codespell on source distribution

#1945 - Better error checks for table loads Insure that table models are sent numeric columns, to catch cases like #1943.

#1946 - versioneer: update to version 0.29 Update the vendored copy of versioneer from 0.28 to 0.29.

#1947 - Ensure tests can be run when the group library does not exist Annotate several tests with the requires_group decorator.

#1952 - Tests: better support of optional region/wcs dependencies Ensure that the tests can pass if the region or WCS code is not available.

#1954 - CI: avoid failures due to missing pyarrow installation Allow the CI runs to pass with new changes to pandas warning messages (completely unrelated to Sherpa but caused the tests to fail).

#1957 - CI: support pytest 8.0.0 Allow the tests to pass when run with pytest 8.0.0.

#1959 - Support logarithmic axes for 1D and 2D confidence plots The projection and uncertainty plots, for both interval (1D) and region (2D), now correctly create a logarithmic scale when requested. Fix #1561.

#1962 - Pick up latest Python micro version in conda workflow This resolves the missing crypt.h issue being seen in the conda test workflow by picking up the latest micro version of the python packages for the conda test build workflows.

#1965 - Improve the error message when a plot backend is not available. Improve the error message when a plot backend is not available. Fix #1964

#1968 - Remove Conda build 3.25 pin from deployments Removes the conda version specification (3.25) to avoid compatibility issues

#1969 - Tests: group the DS9 tests so they can be run with one worker process Ensure those tests that use the requires_ds9 decorator all have the same group marker, so that they will be run within a single worker process when using pytest-xdist to run tests in parallel. There is no change if pytest-xdist is not installed.

#1970 - Bump the minimum pytest version Pytest occasionally changes behavior and it does not seem worth our time to maintain backwards compatibility with old versions of pytest - see #1960

#1972 - Remove excess brackets in model names Remove excess brackets from model and parameter expressions. This is purely a cosmetic change, but hopefully makes complicated model expressions easier to read. Fix #780.

#1973 - Fix the requires_pylab decorator Change the testing code to better-handle those tests that want to check plotting when using the pylab backend. Fix #1971.

#1974 - Allow linked parameters to be fit without including them in the model expression Treat linked parameters as part of the model expression (via the new lpars attribute and get_thawed_pars routine) so that they can be included in a fit without including the linked model as part of the model expression. Fix #777

#1976 - Fix bug in multi-panel plotting in bokeh Fix bug in multi-panel plotting in bokeh

#1977 - model: improve examples in docstrings Allow model.py, parameter.py, and regrid.py to be included in the docstring pytest run.

#1984 - Rework x errorbar support to better-match community expectations Improve the "x errorbar" handling for histogram-style data (including PHA), including support for the wavelength setting (fix #1985), to better match community expectations (e.g. #1817).

#1994 - Tests: improve coverage of parameter module Improve the test coverage of the parameter module.

#1996 - docs: update examples to match current behavior Update the ReadTheDocs documentation to match the current code.

#1998 - Cleanup XSPEC interface A small number of XSPEC model parameter values have been changed to match the default frozen state of XSPEC version 12.13.1.

#2000 - CM-481: update codecov action to v4, add upload token Updates version of codecov being used

#2003 - Minor plot work Internal clean up of the plotting code in preparation for future changes. The ARFPlot class will now generate an IOErr rather than PlotErr if sent a non-PHA dataset (to better match other calls). The DataHistogramPlot class now treats xerr as a property that can not be changed, and is fixed to be the half-width of the X bins, although note that this field is not really used and may be removed at some point in the future.

#2004 - Minor improvements to the RMF plot capability Improve the RMF plot display so that it recognizes the current units setting and allows the choice of energies to be over-ridden by the user.

#2006 - Use a separate context when handling multiprocessing calls Ensure that Sherpa does not change the global multiprocessing state but instead uses the (new) sherpa.utils.parallel.context attribute. Fix #1015.

#2011 - Specialize the residual-style plots for histogram data Improve the display of residual-style plots for integrated datasets. Fix #1817.

#2012 - [CI] Update ci-conda-workflow.yml to pin back to macos-12 Pin back to macos-12. This version is specific to macOS-intel. 13 doesn't work as conda is removed from it.

#2019 - update clone link and versions and DOI This is a documentation update to fix the incorrect link in the source install example. I also updated the version and DOI link in the description of the source installation.

#2020 - CI: bump the checkout actions versions Internal change to how the GitHub actions are run for CI.

#2021 - Update Conda Deployment Workflow Action Versions Updates internal actions in the deployment workflow

#2025 - Improve support for NumPy 2.0 Improvements for running Sherpa with NumPy 2.0.

#2029 - Swap oldest-supported-numpy with numpy in requirements