Merge pull request #364 from astrofrog/fix-devdeps-2

Fix tests against developer versions of packages

CHANGES.md

@@ -66,23 +66,23 @@
 ## 0.9 - 2022-09-02
 
 - Drop support for Python 3.7.
-- 
+-
 - Infrastructure and packaging updates.
-- 
+-
 - Made many improvements, bug fixes, and significant speed-ups for the adaptive
 - resampling algorithm, `reproject_adaptive`. These bug fixes may cause
 - changes to the reprojected images, which are typically negligible.
 - Improvements include the addition of a flux-conserving mode, support for a
 - Gaussian filter kernel, a menu of boundary-handling modes, and a
 - `center_jacobian` flag to trade speed for accuracy with rapidly-varying
 - transformations.
-- 
+-
 - Added a `roundtrip_coords` argument to `reproject_adaptive` and
 - `reproject_interp`. By default, all coordinate transformations are run in
 - both directions to handle some situations where they are ambiguous. This can
 - be disabled by setting `roundtrip_coords=False` which may offer a
 - significant speed increase.
-- 
+-
 
 ## 0.8 - 2021-08-11
 
@@ -91,34 +91,34 @@
 - possible to solve for the optimal WCS for a set of images that couldn't fit
 - into memory all at once, since the actual data aren't needed for optimal WCS
 - determination. [#242]
-- 
+-
 - Fix implementation of `hdu_weights` in `reproject_and_coadd`. [#249]
-- 
+-
 
 ## 0.7.1 - 2020-05-29
 
 - Fixed compatibility with Astropy 4.1. [#234]
-- 
+-
 - Updated minimum requirement for SciPy. [#236]
-- 
+-
 
 ## 0.7 - 2020-04-02
 
 - Made C extension in overlapArea.c thread-safe by removing global
 - variables. [#211]
-- 
+-
 - Made it possible to control whether to output debugging information
 - from overlapArea.c by setting DEBUG_OVERLAP_AREA=1 at build-time. [#211]
-- 
+-
 - Fix compatibility with astropy v4.0.1. [#227]
-- 
+-
 - Disable parallelization by default in `reproject_exact` - this can be
 - enabled with `parallel=True`. [#227]
-- 
+-
 - Fixed a bug with `reproject_exact` with `parallel=False` and
 - `return_footprint=False`, which caused the footprint to be returned
 - anyway. [#227]
-- 
+-
 - The infrastructure of the package has been updated in line with the
 - APE 17 roadmap (https://github.com/astropy/astropy-APEs/blob/main/APE17.rst).
 - The main changes are that the `python setup.py test` and
@@ -127,7 +127,7 @@
 - (https://tox.readthedocs.io) package and run `tox -e test` and
 - `tox -e build_docs`. It is also possible to run pytest and sphinx
 - directly. [#228]
-- 
+-
 
 ## 0.6 - 2019-11-01
 
@@ -136,20 +136,20 @@
 - `independent_celestial_slices=` argument to `reproject_interp` has
 - been deprecated since it is no longer needed, as transformations are
 - automatically done in the most efficient way possible. [#166]
-- 
+-
 - Include a warning for high resolution images with `reproject_exact`,
 - since if the pixels are <0.05", precision issues can occur. [#200]
-- 
+-
 - Added a new `reproject_and_coadd` function for doing mosaicking of
 - individual images, and added section in documentation about mosaicking.
 - [#186]
-- 
+-
 - Added a new reproject.adaptive sub-package that implements the DeForest
 - (2004) algorithm for reprojection. [#52]
-- 
+-
 - Fixed a bug that caused 'exact' reprojection results to have numerical
 - issues when doing identity transformations. [#190]
-- 
+-
 
 ## 0.5.1 - 2019-09-01
 
@@ -160,33 +160,33 @@
 
 - Improve parse_output_projection to make it so that the output projection
 - can be specified as a filename. [#150]
-- 
+-
 - Fixed a bug that caused HEALPix maps in RING order to not be correctly
 - interpreted. [#163]
-- 
+-
 - Make it possible to specify the output array for reprojection using the
 - `output_array=` keyword argument. [#115]
-- 
+-
 
 ## 0.4 - 2018-01-29
 
 - Refactored HEALPix reprojection code to use the astropy-healpix package
 - instead of healpy. [#139]
-- 
+-
 - Added the ability to specify an output array in `reproject_interp`, which
 - permits the use of memory-mapped arrays and therefore provides the capability
 - to handle data cubes much larger than memory [#115]
-- 
+-
 - Fix test 32-bit test failures. [#146]
-- 
+-
 - Fix an issue with reprojecting images where there are two solutions along
 - the line of sight by forcing round-tripping of coordinate conversions [#129]
-- 
+-
 - Explicitly define default HDU as 0 for normal reprojection and 1 for
 - HEALPix reprojection. [#119]
-- 
+-
 - Added a function to find the optimal WCS for a set of images. [#136, #137]
-- 
+-
 
 ## 0.3.2 - 2017-10-22
 
@@ -197,22 +197,22 @@
 ## 0.3.1 - 2016-07-07
 
 - Include missing license file in tarball.
-- 
+-
 - Updated documentation to remove warnings about early versions.
-- 
+-
 
 ## 0.3 - 2016-07-06
 
 - Allow users to pass a `field=` option to `reproject_from_healpix`
 - to access different fields in a HEALPIX file. [#86]
-- 
+-
 - Significant improvements to performance when the input data is a large
 - memory-mapped array. [#105]
-- 
+-
 - Significant refactoring of interpolating reprojection to improve support for
 - n-dimensional arrays, optionally including two celestial axes (in which
 - case the coordinate transformation is taken into account). [#96, #102]
-- 
+-
 
 ## 0.2 - 2015-10-29
 

reproject/adaptive/tests/test_core.py

@@ -141,7 +141,7 @@ def test_reproject_adaptive_high_aliasing_potential_rotation(
  # along the output x axis. With the input image containing vertical lines
  # with values of zero or one, we should have uniform values of 0.5
  # throughout our output array.
- np.testing.assert_allclose(array_out, 0.5, rtol=0.001)
+ np.testing.assert_allclose(array_out, 0.5, rtol=0.002)
 
  # Within the transforms, the order of operations is:
  # input pixel coordinates -> input rotation -> input scaling
@@ -161,7 +161,7 @@ def test_reproject_adaptive_high_aliasing_potential_rotation(
  center_jacobian=center_jacobian,
  kernel=kernel,
  )
- np.testing.assert_allclose(array_out, 0.5, rtol=0.001)
+ np.testing.assert_allclose(array_out, 0.5, rtol=0.002)
 
  # But if we add a 90-degree rotation to the input coordinates, then when
  # our stretched output pixels are projected onto the input data, they will