Release Notes

0.8.7 (29-March-2024)

  • Fix a bug in the imalign.align_wcs function due to which reference catalog would not get expanded even for successful alignments, essentially disabling expand_refcat option. [#201]

0.8.6 (08-January-2024)

  • Improved the quality of the expanded reference catalog when expand_refcat is set to True in the align_wcs function by not using input catalogs that failed to align in the expanded reference catalog. [#195]

  • Reduce memory & compute needed by _xy_2dhist by pruning distant pairs with a kdtree. This is a purely internal change that does not affect the results of the algorithm. [#196]

0.8.5 (30-November-2023)

  • Addressed compatibility issues with Python 3.12. Re-organized package setup machinery. [#188]

0.8.4 (unreleased)

  • Mistaken tagging.

0.8.3 (12-September-2023)

  • Fixed a bug in the linalg module due to which computation of the inverse matrix would fallback to custom implementation instead of using numpy implementation. [#185]

  • Fixed incompatibilities with the future (version 2.0) release of numpy. [#185]

0.8.2 (13-April-2023)

  • Added bb_policy argument to the WCSGroupCatalog to control when to switch to an aproximate method of computing of the bounding polygon of a group of images. The default value is set to 50. Also added equivalent group_bb_policy argument to both fit_wcs and align_wcs functions. [#176]

0.8.1 (23-December-2022)

  • Fixed a bug in the XYXYMatch due to which bin size for the 2D histogram pre-match alignment did not account for the pixel scale in the tangent plane. This required a change in the API of XYXYMatch.__call__ which now _must_ have tp_pscale as input and also inputs catalogs now _must_ contain 'TPx' and 'TPy' columns. [#173]

  • Deprecated 'tp_wcs' argument of the XYXYMatch.__call__() method. Use 'tp_pscale' instead. [#173]

0.8.0 (25-August-2022)

  • Exposed in top-level functions parameter clip_accum that controls whether or not to reset the list of “bad” (clipped out) sources after each clipping iteration during model fitting. [#169]

  • Deprecated tweakwcs.tpwcs module in favor of tweakwcs.correctors. [#170]

  • Deprecated the following classes in the tweakwcs.tpwcs module:

    • tweakwcs.tpwcs.TPWCS in favor of tweakwcs.correctors.WCSCorrector;

    • tweakwcs.tpwcs.JWSTgWCS in favor of tweakwcs.correctors.JWSTWCSCorrector;

    • tweakwcs.tpwcs.FITSWCS in favor of tweakwcs.correctors.FITSWCSCorrector. [#170]

  • Deprecated tweakwcs.matchutils.TPMatch class. Use tweakwcs.matchutils.XYXYMatch instead. [#170]

  • Removed tanplane_wcs argument of the WCSGroupCatalog.apply_affine_to_wcs() method. tanplane_wcs was deprecated since 0.6.5. It was replaced with ref_tpwcs. [#170]

  • Deprecated tpwcs argument of the WCSImageCatalog.__init__() as well tpwcs property of the same class. Use corrector instead. [#170]

  • Deprecated tpwcs argument of the tweakwcs.imalign.fit_wcs() in favor of corrector. [#170]

0.7.4 (13-April-2022)

This is almost exclusively a maintenance release except for close vertices in the convex hull issue to make the code more robust.

  • Set gwcs min version to 0.14. [#158]

  • Set astropy min version to 5.0.4. [#153]

  • Remove consecutive convex hull vertices that are very close to each other. [#147]

0.7.3 (12-August-2021)

  • Fix a bug due to which minobj parameter to WCSGroupCatalog.align_to_ref() and align_wcs() was ignored. [#144]

  • Make peak finding code switch to center-of-mass algorithm when estimated 2D parabolic fit estimates a peak outside of the fit box. Reduce accuracy loss in computation. [#143]

0.7.2 (13-May-2021)

  • Make code more robust to exceptions in the spherical_geometry package. [#138]

  • Fixed a bug in matchutils._find_peak() due to which it could return coordinates of the peak that were outside of the image. [#136]

  • Fixed a bug in how re-projection was computed when center of the transformations was provided. [#135]

0.7.1 (16-February-2021)

  • Added support for detecting and using velocity aberration corrected V2-V3 frames when available in JWST WCS ('v2v3vacorr'). [#130]

0.7.0 (11-November-2020)

  • Added linearfit.fit_rshift function to support a new fitgeom fitting mode 'rshift' that fits only for shifts and a rotation. [#128]

0.6.5 (09-September-2020)

  • Added ref_tpwcs parameter to imalign.fit_wcs(), imalign.align_wcs(), wcsimage.WCSGroupCatalog.align_to_ref() to allow alignment to be performed in user-specified distortion-free tangent plane. [#125]

  • Renamed tanplane_wcs parameter in wcsimage.WCSGroupCatalog.apply_affine_to_wcs() to ref_tpwcs. tanplane_wcs parameter was deprecated and will be removed in a future release. [#125]

0.6.4 (14-May-2020)

  • Bug fix: Unable to initialize JWSTgWCS tangent-plane corrector from an already corrected WCS. [#122]

  • Fix a bug in how corrections are applied to a previously corrected JWST WCS. [#120]

  • Do not attempt to extract center of linear transformation when not available in 'fit_info'. [#119]

0.6.3 (14-April-2020)

  • Fixed a bug due to which reprojection transformation for JWST gWCS was computed at wrong location in the tangent plane. [#118]

0.6.2 (07-April-2020)

  • When WCS has valid bounding box, estimate scale at the center of the bounding box. [#117]

  • Adjust the point at which tangent plane-to-tangent plane transformation is computed by 1/2 pixels for JWST corrections. This correction should have no measurable impact on computed corrections. [#115]

0.6.1 (09-March-2020)

  • Fixed a bug in applying JWST correction for the case when alignment is performed twice on the same image. Due to this bug the inverse transformation was not updated. [#112]

0.6.0 (25-February-2020)

  • Fix a possible crash when aligning FITS WCS images due to an unusual way stwcs.wcsutil.all_world2pix handles (or not) scalar arguments. [#110]

  • Modified the angle at which the reported rotation angles are reported. Now rotation angles have the range [-180, 180] degrees. [#109]

  • Added support FITS WCS that use PC matrix instead of the CD matrix used in HSTs WCS. [#108]

  • Bug fix for alignment of multi-chip FITS images: correction of how transformations from the reference tangent plane are converted to individual images’ tangent planes. [#106]

  • Significant re-organization of the fit_info dictionary. rot now becomes proper_rot and rotxy now becomes rot containing only rotx and roty. Also, scale now is a tuple of only two scales sx and sy. The geometric mean scale is now a separate field '<scale>' as well as the arithmetic mean of rotation angles ('<rot>'). Finally, 'offset' in the fit functions from the linearfit module was renamed to 'shift' in order to match the same field returned by functions from the imalign module. [#105]

  • Linear fit functions now return the fit matrix F instead of its transpose. [#100]

  • Linear fit functions (in the linearfit module) use longdouble for internal computations. [#100]

  • Re-designed the JWSTgWCS corrector class to rely exclusively on basic models available in astropy and gwcs instead of the TPCorr class provided by the jwst pipeline. This eliminates the need to install the jwst pipeline in order to align JWST images. [#96, #98]

0.5.3 (15-November-2019)

  • Added logic to allow some input catalogs to be empty and to allow the alignment to proceed as long as there are at least two non-empty (image or group) input catalogs. [#94]

0.5.2 (26-July-2019)

  • Fixed a deprecation issue in logging and added logic to compute image group’s catalog name using a common prefix (if exists) of the names of constituent images. [#92]

  • Package version is now handled by setuptools_scm. [#93]

0.5.1 (08-May-2019)

  • Fixed a bug in the “2dhist” algorithm resulting in a crash when 2D histogram has multiple maxima of the same value and no other value larger than one. [#90]

0.5.0 (22-April-2019)

  • Fixed a bug due to which a warning log message “Failed to align catalog…” would be issued for successful alignments. [#84]

  • Fixed a bug in creation of WCS image groups with empty catalogs. [#84]

  • Fixed a bug in match2ref when it was run in a non-matching mode (match=None) dute to which it was impossible to detect the case when reference catalog has a different length from a supposedly matched WCS group catalog. [#84]

  • Fixed a bug in computation of the bounding polygon of a reference catalog containing only two sources. [#84]

  • Fixed a bug in convex_hull() resulting in incorrect type being returned in case of empty input coordinate lists or whne only one point is provided. [#84]

  • Implemented a more robust estimate of the maximum type supported by numpy.linalg.inv. [#82]

  • Renamed wcsutils.planar_rot_3D to wcsutils.planar_rot_3d. [#75]

  • Renamed wcsutils.cartesian2spherical to wcsutils.cartesian_to_spherical and wcsutils.spherical2cartesian to wcsutils.spherical_to_cartesian. [#71]

  • Improved “2dhist” algorithm that performs simple catalog pre-alignment used for source matching. [#69]

  • Changed the default value of the searchrad parameter in matchutils.TPMatch to 3. [#69]

0.4.5 (14-March-2019)

  • Fixed incorrect pointer type introduced in previous release [#67].

0.4.4 (13-March-2019)

  • Fixed VS2017 compiler error, "void *": unknown size. [#62, #63, #64]

0.4.3 (13-March-2019)

  • Package maintenance release.

0.4.2 (21-February-2019)

  • Fixed a bug due to which the fitting code would crash is wuv were provided but wxy were set to None. [#60]

0.4.1 (14-February-2019)

  • Code cleanup: removed debug print statements. [#59]

0.4.0 (08-February-2019)

  • Matched indices, linear fit results and fit residuals are now set in the input “WCS catalogs” meta['fit_info'] instead of meta['tweakwcs_info']. [#57]

  • Updated example notebook to reflect changes to API. [#57]

  • Allow TPWCS classes to set meta during object instantiation. This allows attaching, for example, a source catalog to the tangent-plane WCS corrector object. [#57]

  • align_wcs no longer supports NDData input. Instead catalogs can be provided directly in the meta attribute of TPWCS-derived WCS “correctors”. This fundamentally transfers the responsibility of instantiating the correct tangent-plane WCS to the caller. This, in turn, will allow future WCS to be supported by providing a custom TPWCS-derived corrector defined externally to tweakwcs package. Second benefit is that image data no longer need to be kept in memory in NDData objects as image data are not needed for image alignment once catalogs have been created. [#57]

  • Renamed tweak_wcs to fit_wcs and tweak_image_wcs to align_wcs. [#57]

  • Fixed a bug due to which the code might crash due to an undefined ra variable, see issue #55. [#56]

  • tweak_image_wcs() now returns effective reference catalog used for image alignment. [#54]

  • Modified how IDs are assigned to the reference catalog source positions when expand_refcat is True: instead of having all sources numbered consecutively starting with 1, now the code will attempt to preserve the original IDs (if any) of the input reference catalog (refcat) or an input image used as a reference catalog and consecutively number only the sources being added to the refcat. [#54]

  • Modified the clipping algorithm to start with all valid sources at each iteration. In other words, clippings do not accumulate by default. Old behavior can be replicated by setting clip_accum to True. [#53]

  • Cleaned-up iter_linear_fit interface as well as simplified the fit dictionary returned by iter_linear_fit. [#53]

  • Added option to specify statistics used for clipping. [#51, #52]

0.3.3 (21-January-2019)

  • Corrected a bug in the non-weighted rscale fit. [#49]

  • Corrected a bug in the computation of RMSE for the “general” fit. [#47]

  • Added computation of MAE of the fit (in addition to RMSE), see [Mean Absolute Error](https://en.wikipedia.org/wiki/Mean_absolute_error). [#47]

  • Renamed RMSD to RMSE (Root-Mean-Square Error). [#47]

0.3.2 (15-January-2019)

  • Fixed the formula for computing RMSD of non-weighted fit. [#46]

0.3.1 (14-January-2019)

  • Fixed Read-The-Docs build failure. [#45]

0.3.0 (14-January-2019)

  • Implemented higher-accuracy matrix inversion. [#42]

  • Bug fix related to not switching to using bounding_box instead of pixel_shape. [#41]

  • Added support for optional 'weight' column in catalogs indicating the weight of each source in fitting linear transformations. [#41]

  • Add support for weights to the linear fitting routines. [#40]

  • Replaced the use of RMS for each axis with a single RMSD value, see [Root-Mean-Square Deviation](https://en.wikipedia.org/wiki/Root-mean-square_deviation). [#40]

  • Rely on pixel_bounds [see APE 14](https://github.com/astropy/astropy-APEs/blob/master/APE14.rst) when available for computation of image’s bounding box. [#39]

  • Fix a bug in the computation of the world coordinates of the fitted (aligned) sources. [#36]

0.2.0 (20-December-2018)

  • Fix swapped reported reference and input indices of sources used for fitting. [#34]

  • Fix for non-initialized C arrays. [#34]

  • Changelog correction. [#33]

0.1.1 (11-December-2018)

  • Fixeded a bug due to which 'fit_ref_idx' and 'fit_input_idx' fields in the fit dictionary were never updated. [#31]

  • jwst (pipeline) package is no longer a hard dependency. [#30]

  • Removed unnecessary install dependencies. [#30]

  • Documentation improvements. [#30, #32]

  • Corrected ‘RA’, ‘DEC’ units used to compute bounding polygon for the reference catalog. [#30]

  • Updated C code to avoid numpy deprecation warnings. [#30]

0.1.0 (08-December-2018)

  • Added support for aligning FITS WCS. [#15, #16]

  • Added keywords to meta attributes of the TPWCS and NDData to allow easy access to the match and fit information. [#20, #21, #28]

  • Package and setup re-design. Support for readthedocs. [#23]

  • Documentation improvements. [#17, #18]

  • Numerous other bug fixes, code clean-up, documentation improvements and enhancements. [#2, #3, #4, #5, #6, #7, #8, #9, #10, #11, #12, #13, #14, #19, #22, #24, #25, #26, #27, #28, #29]

0.0.1 (25-April-2018)

Initial release. [#1]