v0.7.1

Fixes two bugs:

• Restrict "tweaked moments" for "free shape" by @HannoSpreeuw in #187 - irrealistic values were found from "tweaked moments" extrapolations for "ill-formed" islands.
• Fix 186 and 173 by @HannoSpreeuw in #190 - the --grid command-line parameter was not propagated correctly. Also, pyse --version will now give you the PySE version.

Subimage size for $\kappa * \sigma$ clipping back to the old default of 64 pixels (was 32 pixels).

Full Changelog: v0.7.0...v0.7.1

v0.7.0

What's Changed

• Fixed bug in peak brightnesses for vectorized source measurements with forced beam
• Fixed bug when force-fitting to positions.
• Added CITATION.cff
• Added pragmas for (Numba-) compiled code, such that Pytest coverage is accurate.
• Included output of pyse -h in documentation
• test_measured_vectorized_free_shape regression test now tests flux density measurements.
• is_usable filtered out sources with masked pixels within its FWHM. This is now replaced by a different algorithm with a similar, but not equivalent effect: sources connected to masked pixels and edges will not be measured. It is optional to have those sources removed. Since it is applied at the time of connected-component labelling, it also has an effect on vectorized source measurements;is_usable was not applied for vectorized source measurements.
• Reformatting of literal blocks in docstrings and stricter adherence to "Numpydoc" style.

Full Changelog: v0.6.0...v0.7.0

Return Pandas DataFrames by default

What's Changed

• Source parameters are returned as a Pandas DataFrame by default.
• Sources are measured in a vectorized manner by default.
• A regression test for vectorized source measurements of resolved sources has been added.
• Error bars on deconvolved elliptical parameters are now more realistic.
• A bug from rounding errors has been fixed, this led to an arccos argument larger than 1.
• Implemented workaround for spurious Numba error.
• Fixed bug which led to a zero uncertainty in the deconvolved position angle.
• Fixed bug in the argument order when collecting NumPy ndarrays into a Pandas DataFrame; this affected the interpretation of error_radius.
• By adding target="parallel" to every function that has a guvectorize decorator, PySE is now ~20% faster for vectorized source measurements with Pandas DataFrames returned. Previously, only a few of these functions had target set in this way which meant that only a single CPU core was used when these functions were called.

Full Changelog: v0.5.1...v0.6.0

Fix multiprocessing instabilities and other bugs.

What's Changed

Main fix: a multiprocessing.Pool for parallellization of Gaussian fits over all the detected sources was previously launched from inside the image module, this is bad practice. Such should be done only from a top level script. It has been replaced by a threading pool, which does not offer any speed improvements, unless you use a free threading version of Python. The number of threads in the threading pool can be modified by the user, at the level of the conf.

Other bug fixes and improvements:

• Do not require a config toml file when calling pyse from cli by @tmillenaar in #155
• Add description of source params to generated docs by @HannoSpreeuw in #156
• Full propagation of conf objects. by @HannoSpreeuw in #157
• Work around broken hatch and click combination that is failing the CI by @tmillenaar in #161
• Do not force the numba threading layer by @tmillenaar in #160
• Add nr-threads config and cli parameter by @tmillenaar in #163
• Fix 158 by @HannoSpreeuw in #162
• Restore consistency in use of hyphens by @HannoSpreeuw in #164, which is important for the cli. Also, for clarity: --detection --> --detection-thr, --analysis --> analysis-thr.

Full Changelog: v0.5.0...v0.5.1

v0.5.0

What's Changed

• Combine coverage of different test files into single file
• Re-enable parallel make_subimages
• Reduce test runtime by ~45%
• Build API docs w/ Sphinx and autoapi
• Enable forced beam for "tweaked moments" and return source parameters from vectorized source measurements as a Pandas df

Full Changelog: v0.4.0...v0.5.0

v0.4.0

What's Changed

• PySE run configuration is now governed by a config object. Previously, configuration parameters were scattered all over the code, mostly as globals. Users can now set all these parameters using a config.toml.
• The pyse.py command line script has been replaced by cli.py, reflecting the changes from the config object.
• All classes, methods and functions have numpydoc-style docstrings.
• "Tweaked moments" has been added to vectorized source measurements, such that the peak spectral brightness is measured with a very small bias, even smaller than the bias from Gaussian fits.
• PySE works with Python versions up to and including 3.13.

And a number of bug fixes.

Full Changelog: v0.3.3...v0.4.0

• Three fixes wrt background grid and interpolation by @HannoSpreeuw in #111
• This FITS image can be viewed without error by @HannoSpreeuw in #112
• Fixes for #108 and #104 by @HannoSpreeuw in #113
• Fix 114 by @HannoSpreeuw in #117
• Fix #116 by @HannoSpreeuw in #120
• Fixes 119 by @HannoSpreeuw in #121
• Fix 105 and 118 by @HannoSpreeuw in #124
• Fixes 122 by @HannoSpreeuw in #125
• Add a config mechanism by @suvayu in #126
• Add allow_multiprocessing boolean to configuration by @tmillenaar in #133
• This should fix "coordinates.angsep_vectorized" by @HannoSpreeuw in #136
• Add cli.py that parses arguments into a config object by @tmillenaar in #134
• Remove version restrictions regarding python<3.12 and numpy<2.0 by re… by @tmillenaar in #137
• Faster testing using pytest-xdist by @tmillenaar in #83
• Update code to use config object by @suvayu in #127
• Fix 138 by @HannoSpreeuw in #139
• Revert pytest xdist by @tmillenaar in #141
• Use back-size-x/y if set, fall back to grid parameter from config. by @tmillenaar in #145
• numpydoc-style docstrings for all classes, methods and functions. by @HannoSpreeuw in #142

v0.3.3

What's Changed

• The background grid nodes were not aligned properly on the subimages. #96
• Fast and accurate calculation of the mean and rms background grids without use of sep or dask. #102.

Full Changelog: v0.3.2...v0.3.3

v0.3.2

What's Changed

• Vectorized source extraction
• Fix values attribute override by @HannoSpreeuw in #57
• extract, as called from reverse_se, needs anl as a second positional argument. by @HannoSpreeuw in #58
• README: clarify some terminology by @suvayu in #61
• Port tkp changes to accessors by @tmillenaar in #63
• Reimplement dataaccessor by @HannoSpreeuw in #66
• Compute moments for thin islands differently by @HannoSpreeuw in #75
• Move tests from TraP by @suvayu in #80

New Contributors

• @tmillenaar made their first contribution in #63

Full Changelog: v0.3.1...v0.3.2

v.0.3.1: bug fix release

• Fixes left over UTF8 string encoding issue from Python 2 to 3 transition. This was causing an error when reading a casa table.
• Update implementation of various cached properties

Full Changelog: v0.3.0...v0.3.1

v0.3.0

What's Changed

• Fix no clipping when source density is high by @HannoSpreeuw in #8
• Try ray for multithreaded kappa sigma clipping of subimages by @HannoSpreeuw in #12
• Fix runtimewarnings from nosetests. by @HannoSpreeuw in #10
• Faster interpolation than scipy.ndimage.map_coordinates for the linear case. by @HannoSpreeuw in #14
• Accelerate label islands by @HannoSpreeuw in #15
• Unit tests failed because of grid.shape=(1,1) for a number of these tests. by @HannoSpreeuw in #16
• Kappa sigma clipping using dask by @HannoSpreeuw in #13
• Parallel fitting by @HannoSpreeuw in #17
• Parallel fitting by @HannoSpreeuw in #18
• Migrate CI to GitHub Actions by @jdswinbank in #19
• Use newer version of kernsuite by @jdswinbank in #20
• Don't store data in Git LFS. by @jdswinbank in #22
• Fudge max pix and maximum pixel method variance should only be called once per image by @HannoSpreeuw in #25
• Make PySE run on multiple cores and fix a bug in the clipping algorithm. by @HannoSpreeuw in #28
• Timing individual modules is redundant in production code. by @HannoSpreeuw in #31
• This should enable a correct import of MutableMapping. by @HannoSpreeuw in #37
• Add dask to setup by @HannoSpreeuw in #38
• Added cmp function by @HannoSpreeuw in #39
• This should enable a pytest based CI by @HannoSpreeuw in #41
• Clean up package build by @suvayu in #48

New Contributors

• @jdswinbank made their first contribution in #19
• @suvayu made their first contribution in #48

Full Changelog: v0.2.2...v0.3.0