Pdr

Added

Features
- `Data.dump_browse()` and `Data.show()` now take an optional `slice_axis`
keyword argument which controls which axis of a 3-D image array `Data`
treats as the 'band' or 'channel' axis

Dataset Support
- MGS MOLA PEDR (a special case fixes non-standard STRUCTURE pointer names and
'ROWS = "UNK"' from the labels)
- LRO LAMP EDR and most RDR table/image pointers
- additional IUE images
- Apollo Lunar Sample Photographs

Other
- Our [JOSS paper](https://joss.theoj.org/papers/10.21105/joss.07256) was accepted for publication! :tada: A pdf of our JOSS
publication is included in the documentation. This is now our preferred
citation. A CITATION.cff file is now included in the repo.

Changed
- DESCRIPTION pointers will now be deduplicated and print with indexes in the
keys list (e.g. DESCRIPTION_0, DESCRIPTION_1, etc.). No warning will be triggered.
- Unit tests have been moved inside the pdr/ folder and will now be packaged
and shipped with pdr
- Previously only pointers named exactly "^STRUCTURE" or "^PDS_OBJECT" skipped
deduplication. Now all pointers with "STRUCTURE" in the name skip deduplication,
and all pointers with "PDS_OBJECT" in the name get ignored.
- vendored `pdr.vax` module removed in favor of using SETI's `rms-vax` (successor
package to that vendored module)
- `pdr.pilutils` renamed to `pdr.pil_utils` for consistency
- `Data.get_scaled` and functions that use it (like `Data.dump_browse`) now also
mask nonfinite values (inf, -inf, and nan)
- `pdr` no longer greedily heads files when looking for PVL labels
- Added a `check_special_block()` call to `Data.get_scaled`

Fixed
- updated docs/LICENSE.md to match LICENSE.md
- `pillow`'s `DecompressionBombError` now fully suppressed when reading desktop
format images in primary mode

Removed
- `pdr` ignores all PDS3 pointers containing the string "PDS_OBJECT". `Data`
previously returned a TBD message on attempts to access objects associated
with these pointers. We have determined that these pointers were markers for
internal software tools and do not refer to unique data objects.
- The text of our JOSS paper publication on pdr has been removed from
`joss_paper/paper.md`. Please refer instead to the pdf under `docs/pdr_joss_paper.pdf`
- A warning is no longer raised for out-of-bounds values on converting to 8-bit
for browsification

Changed
- Consolidated optional 'browsify' and 'desktop_image' dependencies into simply
'pillow'.
- Any attempt to use functionality that requires an optional dependency now raises
a `ModuleNotFoundError` naming the required module.
- To support this, `astropy` objects are now always imported through
`loaders.astrowrap`.
- Updates to documentation to include instructions for pip users, etc.

Fixed
- `docs/index.md` updated to reflect the same content as README.md

Removed
- Experimental `save_sparklines()` function in `pdr.browsify`, and along with
it, the optional matplotlib dependency.
- Unnecessary full `matplotlib` dependency for Notebook use.

Added

Features
- Support for 'desktop' image formats in primary mode: JPEG, TIFF, GIF
(still and animated), PNG, MPO/MPF, WebP (still and animated), JPEG2000/JP2,
and BMP. Behavior is as follows:
- `pdr` will attempt to parse all EXIF, TIFF tag, and MPO metadata, including
embedded XML packets. Parsed metadata is available in `Data.metadata` and
accessible with `Data.metaget()` and related methods.
- Some basic image information is also available in `Data.metadata` (format,
color mode, dimensions, MIME type, etc.)
- Most files will have a single data object named 'IMAGE'.
- Animated GIF and WebP images will have one data object per frame, named
'FRAME_0', 'FRAME_1', etc.
- MPO/MPF files (this includes many files with .jpg and .jpeg extensions)
will have one data object per image. The MPO primary image is named 'IMAGE'.
Subsequent images are assigned names based on their MPType and index within
the file.
- PDR automatically converts palette-mode images (like many GIFs) to 3D RGB
arrays on load. In these cases, mappings between RGB values and original
palette indices are available with `data.metaget('palette')`.
- GeoTiffs work like other TIFF files. `pdr` naively parses the GeoTiff
metadata, but does not construct a CRS or otherwise derive projection
information.
- Note that some of PDR's convenience features like `Data.show()` do not
work well with images that rely on transparency.

Dataset Support
- additional Cassini ISS calibration images
- MRO HiRISE RDRs and additional DTM images (were notionally supported, now have official support)

Other
- The text of our JOSS paper publication on pdr is now included under
`joss_paper/paper.md`. It will be moved to `docs/paper.md` in next release.
- Clarifying comments were added to the supported_datasets documentation regarding
spice files
- Community guidelines including a code of conduct and contribution guidelines
have been added to the readme and docs folder

Changed
- `Data.dump_browse()` and `.show()` no longer apply default contrast enhancement
if it would result in an image with only one value. Explicitly passing the
`image_clip` argument will override this, if for some reason you want to.
- Levenshtein dependency is now optional. Users installing via pip will need to
use the "fuzzy" option.
- `tiff` optional-dependency feature in has been changed to `desktop_image` for
pip users to better fit with new features in this version

Fixed
- `Data.dump_browse()` now works correctly with FITS files opened in primary mode.

Removed
- `CHANGELOG.md` was removed in preference of solely `docs/version_history.md`
- Binder demo of example notebook due to issues with Binder servers.

Added

Features
- Full compatibility with numpy 2.0 for pds4 products
(pds3 products already compatible in last release)

Changed
- `pds4_tools` is now vendored within `pdr` to apply changes for numpy 2.0
compatibility and avoid hot patches

Removed
- `pds4_tools` dependency
- `_patches` module

Added

Features
- Compatibility with numpy 2.0 for pds3 products and fits pds4 products
(non-fits PDS4 product numpy 2.0 compatibility awaits changes in
`pds4_tools` dependency)
- Support for VAX_REAL data types in qubes
- `pdr.Data.__contains__` method
[see issue 57](https://github.com/MillionConcepts/pdr/issues/57)
- Unit tests are now backed by `pytest` fixtures for improved flexibility and
compatibility
- ROW_PREFIX_BYTES are now accounted for when calculating offset positions

Dataset Support
- Galileo NIMS qubes
- MSL
- EMS EDRs and RDRs
- additional navcam and hazcam RDRs
- Rosetta
- CONSERT
- ALICE supplementary tables
- MIDAS RDRs
- additional COSIMA tables
- LASP Mars datasets
- Venus Express (VEX)
- Phoenix
- ASE EDRs, RDS
- Telltale Experiment (TT) products
- Atmospheric Opacity (AO) products
- TEGA LED EDRs and SC RDRs
- MGN RSS solar wind experiment
- MGS RSS solar conjunction data
- MEX VMC EDRs and RDRs
- Mariner 10 MAG, PLS, and POS products
- WFF/ATM derived DEM images
- Misc additional products from:
- Cassini RPWS, MIMI, and ISS
- Galileo Probe, MAG, SSI
- IHW
- ground-based datasets at SBN and ATM
- Stardust
- LRO Mini-RF
- NEAR NLR
- ICE
- Juno JADE
- Pioneer Venus Orbiter OMAG
- Vega
- Voyager LECP
- MEX PFS and MARSIS
- MER navcam/hazcam/pancam RDRs

Other
- (GitHub-specific) CI for unit tests / coverage reports

Changed
- PDR now associates FITS HDUs with PDS data objects by matching byte ranges
as described in PDS labels to HDU byte ranges as described in their FITS
headers, rather than by performing fuzzy string match between HDU names and
PDS object names
- PDS4 FITS files are now opened using `astropy.io.fits` on the backend
instead of `pds4_tools`, mirroring the PDS3 FITS behavior
- Non-filled-out FITS cards are represented as None, rather than special
astropy objects without stable byte representations
- Assorted backend refactoring, linting, and minor improvements
- Unit tests no longer unnecessarily write to the source tree
- All string objects loaded directly from products by `Data` ('DESCRIPTION',
'LABEL', etc.) canonicalize line endings to \n

Fixed
- More compatibility fixes for Python 3.9, mostly related to type annotations
- Improved reliability of associating compressed files with filenames that
imply uncompressed files (and vice versa)
- Addressed some cases in which PDR wasn't accurately determining lengths for
HEADER objects
- HISTOGRAM_IMAGE objects will now be read as images instead of attempting
to be read as tables. (IMAGE_HISTOGRAMS will still be read as tables)
- setup.py also correctly pins `numpy < 2.0.0`
- The hotpatch for `pds4-tools` Python 3.12 incompatibility no longer
has persistent side effects; for instance, `conda` will no longer
complain about checksum mismatches.
- Bug reading PDS4-labeled FITS files

Removed
- `pdr.Data.__iter__` method is being deprecated
[see issue 57](https://github.com/MillionConcepts/pdr/issues/57)
- `pdr.Data.__getitem__` method is being deprecated
[see issue 57](https://github.com/MillionConcepts/pdr/issues/57)

Changed
- Pinned numpy version to less than 2.0 until compatibility issues with new
release are resolved.