Release Highlights

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

• enhancements:

  • improved handling for OGIP products that do not follow the standards (#2235)
  • Models can now be combined using arbitrary ufuncs (#2248)
  • various updates to XSpec interface (#2281) and support for XSpec 12.15.0 including the additional of 17 new models (#2252)
  • added interface to support ArviZ for analysis of Bayesian models (#2286)

• documentation changes:

  • provided option to install python packages needed to build sherpa documentation

• infrastructure and testing:

  • updated tests to work with python 3.13 (#2174)

• bug fixes:

  • fixed a bug in the caching code that limited the cache size to 1 (#2166)
  • update to use fake_pha for missions which don't use 1 as the first channel (#2214)

Details

#2039 - Add a security reporting policy Added a statement defining plans to deal with security issues and where users should report them

#2047 - Opt Clean up Improve the typing coverage of the sherpa.optmethods code and internal code clean up.

#2116- Add explicit support for XSPEC 12.14.1 Add models from XSPEC 12.14.1 and change the parameter settings to match those from 12.14.1, where necessary.

#2143 - Improve pre-commit script for ranges of years Improve pre-commit script for ranges of years

#2156 - Add macOS ARM Conda Package Builds Add macOS ARM Conda builds

#2164 - XSPEC: fix up error message Correct the information provided in the unlikely case that an XSPEC model fails.

#2166 -Fix bug in cache code that limited cache size to 1. Fix bug that limited cache size to 1.

#2170 - DOC: Example changes to Examples Change the docstring for calc_source and calc_model to use Examples rather than Example.

#2171 - specify DS9 region format used in pytest Explicitly set region format to use ds9 format for ds9 region pytests

#2174: Allow tests to pass with Python 3.13.0 Allow the tests to pass with Python 3.13.0.

#2178 - Post 4.17.0 update to add DOI info to the README.md Adding the DOI for the 4.17.0 Release.

#2179 - add build doc option Provide the doc installation option which will install the python packages needed to build the documentation along with Sherpa.

#2181 - Fix typo in return value Correctly identify the return value of get_source_components_plot. This does not change the behaviour of the code.

#2182 - DOC: Use "Examples" rather than "Example" Consistent use of "Examples" to indicate one or more examples in the documentation.

#2188 - TEST: address numpy 2 changes to output for docstring tests Allow the documentation tests to pass when using NumPy 2.0. There is no change to the actual code.

#2191: TEST: fix test related to issue #800 Clean up a test. There is no functional change to the code.

#2195 - Update ci-conda-workflow to use Miniforge Install Miniforge for use with ci-conda workflow.

#2198 - Allow both Intel and ARM macOS GitHub runners for the conda builds Add a macOS ARM build, and update the macOS Intel build to use the macOS-13 runner.

#2201: Change how invalid limits are checked for within the MCMC code Simplify a safety check within the MCMC proposed-parameter code.

#2203 - Allow pytest 8.3.4 to be used Allow the tests to pass with pytest version 8.3.4 by changing the test code. The GitHub macOS conda action is updated to macos-13 (for the Intel build). There is no functional change here.

#2204 - Update the GitHub Actions badge for Conda and Pip workflows Update the GitHub Actions badge in the README.md for the Conda and Pip workflows.

#2207- TESTS: avoid matplotlib version differences (fix #2206) Allow tests to pass when using matplotlib 3.10.0. This only changes the tests. Fix #2206.

#2208: BUILD: update ds9 from 8.5 to 8.6 Bump to DS9 version 8.6 for testing on GitHub actions (Linux). There is no functional change.

#2209: update to change .bz2 references to .conda to account for changes Modify workflows to use .conda instead of .bz2

#2213 - Better handle dataspace1d for DataPHA use cases when offset is not 1 The sherpa.astro.ui.utils.dataspace1d call can now be used to generate a channel grid that does not start at 1 for DataPHA data. The arguments are now checked to ensure they are sensible in this case (consecutive integer values).

#2214: Better handle fake_pha when offset is not 1 Allow fake_pha to work with missions where the first channel is not 1. Fix #2212.

#2215 - Minor xspec additions Minor changes to the XSPEC module, including added documentation to the set_xsabund routine to note the availability of the "felc" table, and set_xsstate no longer requires all fields to be present.

#2226 - Allow user-statistics to be referenced by name to match the documentation. Allow user-statistics to be set by name. The get_stat_name routine now returns the actual name of the user statistic, rather than the generic value "userstat". Fix #2225.

#2227 - Tests: compare Response1D and MultipleResponse1D behavior Improve testing of the multi-response code. There is no change in behaviour here.

#2229 - Error out if model names are not strings Raise an error if a model name is not a string

#2230 - Note that we need the file program to build Sherpa Note that the file command is needed to build Sherpa (it is used by extern/configure).

#2231 - Two small astropy-related doc changes Update links to astropy dev docs (that we link to in lieu of writing our own dev docs); Recommend installing astropy via pip, not conda

#2232 - TESTS: mark more that require the group module Mark several recently added tests with the requires_group decorator. There is no functional change.

#2233 - BUILD: pin conda-build (and therefore liblief) to last known working version There's a bug between conda-build and lief that's causing a package build failure. This pins to "conda- build==25.1.2=*_0", which we've also pinned to in CIAO for this same issue. related Conda-build issue: conda/conda-build#5626

#2235 - Deal with OGIP products that do not follow the standards Some OGIP products from missions occasionally do not follow the OGIP and FITS standards. Provide workarounds for two recent cases from NuSTAR and NICER (TLMIN/MAX values were recorded as a string, and a QUALITY array was stored as a boolean array). Fix #2185.

#2239 - REGION: test CXCDM region parser, when available Allow the region module to know how it was built (whether reading FITS binary region files is supported). This allows for improvements to the tests.

#2240 - Minor typing updates Minor internal changes to tests and code to improve typing statements and improve test coverage.

#2241 - drop "iterative fit method" in error-estimation screen output when not set The screen output when calling an error-estimation routine like conf will no longer return the iterative-fit method name when it is set to None.

#2243 - Internal improvements to stats, fit, optimization, and error estimation Add typing statements and apply minor code cleanup for the statistics, fit, optimization, and error estimation code. There is not expected to be any functional change unless code interacts with low-level details of the Sherpa code base.

#2248 - Allow combinations of models with arbitrary ufuncs Models can now be combined using arbitrary ufuncs. Before, it was possible to say model1 + model2 to create a combined model, now any numpy ufunc will work, e.g. np.greater(model1, model2).

#2249 - Update project.toml for PEP 639 Address setuptools deprecation of lower-kebab-case for names of options

#2252- Allow Sherpa to build against XSPEC 12.15.0 Allow building against XSPEC 12.15.0. Adds the following 17 models:

    additive: XSfeklor, XSvlorentz, XSvvoigt, XSzfeklor, XSzlorentz, XSzvlorentz, XSzvoigt, XSzvvoigt
    convolution: XSrgsext
    multiplicative: XSlorabs, XSvlorabs, XSvoigtabs, XSvvoigtabs, XSzlorabs, XSzvorlabs, XSzvoigtabs, XSzvvoigtabs

The XSPEC initialization file (~/.xspec/Xspec.init) may need to be removed or the AtomDB version in it updated before using XSPEC models.

#2255 - Add missed pep 639 config update

#2249 had a few stragglers in the update I missed when reviewing. This should get all the needed renames.

#2256 - Faster hash for modelCacher1D Select a faster hashing algorithm for the model cache by default

#2261 - Add tests related to analysis setting Add or enhance tests for several cases.

#2265 - Add new public methods to access the components of a model Add new methods to select the components of a model.

#2268 - Treat norm special in XSPEC additive models These code changes increase the speed of fitting data with XSPEC additive models by making the cache more effective.

#2273 - Make sure all datastack-wrapped functions are added to namespace Complete the datastack namespace for functions wrapped to operate on datastacks

#2274 - Set cache=0 by default for models that are fast to calculate Based on benchmarking for typical Sherpa use cases, some models are faster to recalculate than to cache. For those models, the default cache size is now set to 0 to save runtime and memory.

#2279 - Note support for Python 3.13 Note that Sherpa can be used with Python 3.13.

#2280 - Add option record_steps to Fit.fit() calls Add option record_steps to Fit.fit() calls, which will add a field record_steps to the FitResults object that the fitter returns. This can help to understand why a fit does not converge.

#2281 - Provide access to XSPEC abundances Minor improvements to the XSPEC interface: get_xsabundances now takes an optional table name, get_xsxset can be called with no argument, and clear_xsxset will clear the XSET database. Fix #2282.

#2286 - Add an interface to ArviZ Add an interface to ArviZ

#2287 - Use python 3.10 syntax for typing for the ui modules Use Python 3.10 syntax for typing statements in the ui modules. There are no functional changes in these commits.

#2288 - ui: add more examples of using SherpaVerbosity Point out the use of the sherpa.utils.logging.SherpaVerbosity context manager in the documentation of the fit and error-analysis UI routines.

#2289 - Update http -> https Update links in docs and source code from http -> https where possible

#2290 - Minor clean up of utility code Ensure a deprecated routine can still be used and remove a routine that has never worked.