Repository navigation
1.5.0 - 2026-11-18
This is a relatively big release, special thanks to @sergei-malyshko-navvis, @dsamaey and @SpaceOceanCorp for helping out 🎉
Changed
- PROJ alignment sweep #133
+fis the flattening and+rfthe reciprocal flattening, as in PROJ. The two were swapped, so definitions that passed a1/fvalue under+f(e.g.+f=298.257222101) built a nonsensical ellipsoid; they must now use+rf. PROJ rejects such a+fvalue outright- Transverse Mercator (
+proj=tmerc) uses the exact Poder/Engsager algorithm for ellipsoids, which has been PROJ's default since 6.0, and accepts+approxto select the Evenden/Snyder series. Inside a normal zone the two agree to well under a millimetre; 40 degrees from the central meridian they differ by ~1.5 km, and beyond ~80 degrees the series returns garbage where the exact algorithm reports the point as outside the projection domain - Equidistant Cylindrical (
+proj=eqc) implements the ellipsoidal method (EPSG:1028) as well as the spherical one (EPSG:1029), matching PROJ 9.8.0. Northings on an ellipsoid are now meridional arc lengths rather thana * phi
Projection.projectInversethrows aProjectionExceptionfor projections that have no inverse (Airy, August, Boggs, and others). It used to return its input unchanged, soinverseProject()silently passed projected coordinates off as radians #141- Azimuthal projections (
ortho,gnom,aeqd,stere) built from a PROJ string without+lat_0/+lon_0are centred on (0, 0), PROJ's default. They used to keep the constructor centres: (45, 45) fororthoand the north pole for the others. Code that calls the constructors directly is unaffected #139 - Updated
LICENSEandNOTICE.md#132
Fixed
- Polar stereographic without
+lat_tstakes true scale at the pole, as in PROJ. It used to default to 0, which halved the coordinates:+proj=stere +lat_0=90 +k=0.994 +ellps=WGS84put (0, 85) at y = -280343 m instead of -555457 m. Definitions with+lat_ts=90(e.g. EPSG:32661) are unchanged #139 - Spherical precision fixes, as in PROJ #143
- Aitoff, Winkel Tripel and spherical
aeqd: the forward transform computes the angular distance withhypot/atan2instead ofacos, so it stays accurate near the origin/centre.aeqdnow rejects only points within 1e-14 of the antipode (the tolerance was 1e-8) - Mollweide: 30 Newton iterations instead of 10, which ran out before converging and snapped the point to the pole
- Hammer:
initialize()divided+Mby+Win place, so calling it a second time gave a different projection
- Aitoff, Winkel Tripel and spherical
- Oblique Mercator: compute the natural-origin (uc) offset from the central-line azimuth (
+alpha) instead of the rectified bearing (+gamma), so the projection centre maps to the false easting/northing when+gammadiffers from+alpha(e.g. an explicit+gamma=0with a non-zero azimuth) #130 - PROJ alignment sweep #133
- Oblique Mercator:
+alphaalone no longer behaves as+gamma=0, and+gammaalone no longer falls back to an azimuth of -45 degrees. Both parameters are now tracked as given, as PROJ does, and the two-point form (+lat_1/+lon_1/+lat_2/+lon_2) is reachable - Oblique Mercator: use
atan2in the forward transform, as PROJ does; the previousatan(y/x)plus a fixed+PIcorrection picked the wrong branch for points more than 90 degrees from the central line, and the near-meridian fallback had a spuriousBfactor +Rdefines a sphere. It used to set only the semi-major axis and leave the ellipsoid's eccentricity in place, which shifted every projection with an ellipsoidal branch (merc,tmerc,laea,aea,cea,leac,lcc,stere,sterea,somerc,aeqd,omerc,geos,cass) by tens to hundreds of kilometres- Mercator:
+lat_tswas ignored; it now sets the scale factor as in PROJ (cos(lat_ts)on a sphere,msfn(lat_ts)on an ellipsoid) - Equal Area Cylindrical: a supplied
+k_0was overwritten bycos(lat_ts)even when+lat_tswas absent - The Snyder conics (
euler,murd1,murd2,murd3,pconic,vitk1) ignored+lat_1/+lat_2and used hardcoded 30/60 degree parallels; their shared inverse also used the unshifted northing inatan2and dropped the sign flip for a negative cone constant - Equidistant Conic (
+proj=eqdc) was not PROJ's algorithm at all: it used a Lambert-Conformal-style formulation with a hardcoded eccentricity of 0.822719, a unit radius, and hardcoded standard parallels, and its inverse was never wired into the framework. Rewritten from PROJ'seqdc.cpp - Bonne:
+lat_1was ignored and the standard parallel was pinned at 90 degrees (the Werner limit) - Polyconic: the ellipsoidal branch was unreachable (
sphericalwas forced true), its forward transform used an uninitialised value in place of the longitude, and its inverse used1/eswhere PROJ uses1 - es - Cassini: sign error in the ellipsoidal forward easting series (
C1 - ...instead ofC1 + ...) - Rectangular Polyconic:
+lat_tsand+lat_0were ignored - Transverse Mercator: the spherical branch applied the scale factor twice to the easting
- Winkel Tripel:
+lat_1was ignored, and the Winkel constant was being passed as the latitude of origin - Krovak: PROJ's fixed defaults (Bessel 1841,
lat_049°30′N,lon_042°30′ of Ferro − 17°40′,k_00.9999) are applied when the definition omits them, and+czechselects the native westing/southing orientation - Bipolar Conic: dropped a spurious
lon_0default of −90 degrees that PROJ does not have - Near-sided Perspective (
+proj=nsper):+hthrewNoSuchElementException, the aspect (oblique/equatorial/polar) was hardcoded to equatorial, and the far-side-of-the-globe domain check was commented out - Space Oblique Mercator for Landsat (
+proj=lsat):+lsatand+pathwere rejected as unsupported and the satellite/path were hardcoded to 1/120 - Hammer:
+Wand+Mwere rejected as unsupported, and the initialisation overwrote whatever was set with the defaults - Lagrange:
+Wwas rejected as unsupported, and the default was 1.4 instead of PROJ's 2 - Urmaev Flat-Polar Sinusoidal (
+proj=urmfps):+nwas rejected as unsupported - Lambert Equal Area Conic (
+proj=leac):+souththrewNoSuchElementException putp2,nell,mbtfpq,mbt_fps: the Newton iteration ran on the output northing instead of the latitude, so it never converged and the unconverged latitude was usedwag1/urmfps,wag2,mbtfpp: the transformed latitude was computed and then discarded, and the raw latitude used in its place- Loximuthal: a missing
abs()in the near-parallel test sent every point south of+lat_1down the wrong branch - Robinson: latitudes landing exactly on a 5-degree table node selected the node below through
floor()rounding +units=ch,+units=fath,+units=linkand+units=us-chwere silently treated as metres because those units, though defined, were missing from the lookup table; the Indian units (ind-yd,ind-ft,ind-ch) were added+vunitsand+vto_meterare accepted instead of raisingUnsupportedParameterException. The EPSG resource ships 158 compound (horizontal + vertical) definitions carrying+vunits, so 144 of them -OSGB36 / British National Grid + ODN height,Amersfoort / RD New + NAP height, theETRS89 / NTM zone N,ETRS89 / UTM zone NandNAD83(CSRS) / UTM zone Nfamilies,SWEREF99 xx + RH2000,Tokyo + JSLD, and others - could not be constructed at all. The vertical unit is recorded but not applied: proj4j carries the vertical ordinate in metres, as it did before. The horizontal transform is unaffected either way, which is all PROJ does with it for the 2D part
- Oblique Mercator:
Added
- Equal Earth projection (
+proj=eqearth) and EPSG codes 8857-8859 (WGS 84 / Equal Earth Greenwich, Americas, Asia-Pacific) #138 - PROJ alignment sweep #133
ProjAlignmentTest, pinning 67 cases against values generated from raw PROJ pipelines and cross-checked on PROJ 9.5.1 and 9.6.0
- Inverse transforms for Aitoff (closed form), Winkel Tripel (PROJ's Newton-Raphson, Bildirici 2016) and Hammer (closed form, taking
+Wand+Minto account) #141 - CRS area of use:
CoordinateReferenceSystem.getAreaOfUse()returns anAreaOfUsewith the EPSG bounds (decimal degrees) and area name, ornullfor parameter-defined CRSs and codes without metadata. The data ships inproj4j-epsg(area_of_use.tsv, from PROJ database v10.074) #142 SECURITY.md#131