Changelog¶

All notable changes to this project are documented here. The format follows Keep a Changelog.

Unreleased¶

Documentation site built with Sphinx and published to GitHub Pages: getting started, a user guide for every module, executed examples, a generated API and command-line reference, background, and a list of known issues and workarounds (#1).
AppliedOperation.bound_operations: the operations the CRSs themselves declared when the route is bound, so a pair of bound CRSs reports both codes instead of none.
operation= accepts a pyproj.crs.CoordinateOperation as the operation itself, applied as stated.
A longitude beyond a full turn is refused with CoordinateOutOfRangeError, as an impossible latitude already was.

PROJ 9.9.0 with EPSG v13.102 and proj-data 1.25, up from PROJ 9.8.1 with EPSG v12.029 and proj-data 1.24; pyproj stays at 3.8.0. PROJ 9.9.0 refuses a proj.db with database layout 1.6, so a database built on PROJ 9.8.1 must be rebuilt. Every workaround on the known-issues page was re-checked and is still needed.
Examples, docs and tests name EPSG:11559 instead of EPSG:9484, which EPSG v13.102 deprecated and restated from ETRS89. The test dataset follows the v13.102 changes to EPSG:28991, EPSG:2180 and four deprecated vertical operations; no expected coordinate changed.
available_operations can list many more candidates for a pair whose datums have no area of use in common: PROJ 9.9.0 no longer drops chains whose steps do not overlap when extents are ignored (OSGeo/PROJ#4699). Such a candidate has no area_of_use.

Engineering CRSs now honour the xy value order. PROJ’s always_xy never normalises an engineering CRS, and PROJ reads the origin of a Similarity transformation stated in a northing-first projected CRS in declared order even where the authority stated it easting-first (EPSG’s own EPSG:1035 does); both transposed coordinates silently. The package now adds the missing swap and settles the ordinate convention from the operation’s area of use, refusing the operation when it cannot. Documented as a workaround in transformation.py and on the known-issues page.
A grid the compiled pipeline must read is checked on disk, so a stated operation (OSDU payload) or a bound-CRS chain whose grid is absent raises MissingGridError at construction rather than failing at the first point or being reported as an unavailable operation.
to_dataframe() no longer produces duplicate column labels for a CRS whose axes share an abbreviation (EPSG:3388).
The devcontainer no longer patches the installed proj.db in place or links grids into the installed data directory; local grids are read from local/grids/ through PROJ_DATA, and a patched database is written only under the repository. tests/test_environment.py guards this.

First version.