)

* no `map` before `any` or `all`

* get rid of more maps

* revert a change in utils

* revert some more changes

* Revert "get rid of more maps"

This reverts commit 429da5b.

* Update Rasters.jl

export Extents.extent

* Update src/Rasters.jl

* Update src/Rasters.jl

---------

Co-authored-by: Rafael Schouten <[email protected]>

to fix the load error

* Print a styled error message so it's more eye-catching

* Mention that packages need to be loaded *manually*

Co-authored-by: Rafael Schouten <[email protected]>

---------

Co-authored-by: Rafael Schouten <[email protected]>

* add continue_patch commits here

* fix refdims values in testts

* fix Vararg warnings

* bugfix concat

* fix concat with parent

* min bound DD at 0.28.2

* compare parents in reproject

* fix comparison

* bugfix comparisons

* bugfix tests

* fix allocs?

* fix allocs

* run code once before testing allocations

* allow more allocations for 1.11

* optimize cellsize for lon-lat projections

* never reverse order of intervalbounds

* add a test with a reverse order dimension

* use approx so tests pass

* combine two broadcasts

* change cellsize to cellarea and default to m2

* fix a test

* use mapped dims if those are lon-lat

* fix typo in test

* remove some code scribbles

* radius not R

* better check if a crs is lon-lat

* even better check if projection is lon-lat

* define AG = ArchGDAL in main file

* use sind and cosd instead of deg2rad

* use isintervals(dims)

* consistently use AG instead of ArchGDAL

* add area_in_crs keyword

* make the example easier to read

* use Eriksson's formula instead of Girard's theorem

* test on a 5x5m grid

* get rid of accidental line in test

* put back order = :trad

* optimizations

* depend on GeometryOpsCore, reexport common manifolds

* Refactor the toplevel `cellarea` to take a method as the first arg

* Remove kwargs from cellarea

* Fix dispatch and propagate kwargs

* Update Project.toml

* fix typo

* Update extensions.jl

* Update Rasters.jl

* fix tests

* Switch `reproject` to use Proj instead of ArchGDAL

This commit creates a new `reproject` method that uses Proj to transform an array of values in place using `proj_trans_generic`, only allocating a Float64 output array.  It also performs the reproject check before actually transforming the array.

This decreases runtime on my machine from 4.168 ms (ArchGDAL) to 665.187 μs (Proj), and allocations from 35167 allocs: 1.254 MiB (ArchGDAL) to 108 allocs: 42.859 KiB (Proj).

* Fix invocations to ArchGDAL GFT converts in tests

* Add `+type=crs` to all Proj-strings

This is a weakness of this method compared to the old ArchGDAL method.  Does this need a fix in Proj?

* Implement cellarea in Proj

* Add Proj to test project

* Switch cellarea tests to Proj

* Bump Proj compat to 1.7.2 to get the WKT support

* remove cellarea and reproject.jl from archgdal extension

* throw extension error with reproject

Co-authored-by: Rafael Schouten <[email protected]>

---------

Co-authored-by: Anshul Singhvi <[email protected]>
Co-authored-by: Rafael Schouten <[email protected]>

* add missing deps

* typeo

* add a constructor to load future worldclim as a rasterstack

* pass keywords to raster and use bioclim keys

* add a raster constructor for future bioclim

* add some tests

* bump rasterdatasources compat

* tests need proj

* use RDS 0.7

* fix a broken test

---------

Co-authored-by: Rafael Schouten <[email protected]>

* Update utils.jl

* Update ncdatasets_source.jl

* extract tweaks

* just write kw... in docs

Co-authored-by: Rafael Schouten <[email protected]>

* use Type{G}

Co-authored-by: Rafael Schouten <[email protected]>

* add a missing where G

* merge _rowtype changes from other branch

* optimization for extract points with skipmissing

---------

Co-authored-by: Rafael Schouten <[email protected]>

* Update commondatamodel.jl

* Add test

* move resample tests to be last

* Include commondatamodel and zarr tests

* use Zarr from ZarrDatasets.jl

* Don't run Zarr tests

* Add the julia cache action to speed up CI

* extract tweaks

* add rasters.sample

* get rid of some toy code

* just write kw... in docs

Co-authored-by: Rafael Schouten <[email protected]>

* use Type{G}

Co-authored-by: Rafael Schouten <[email protected]>

* add a missing where G

* neater formatting

* better reuse _rowtypes code

* more code style

Co-authored-by: Rafael Schouten <[email protected]>

* extract tweaks

* add rasters.sample

* get rid of some toy code

* neater formatting

* better reuse _rowtypes code

* more code style

Co-authored-by: Rafael Schouten <[email protected]>

* merge _rowtype changes from other branch

* optimization for extract points with skipmissing

* streamline _rowtype

* correctly use _rowtype

* improve type stability

* switch to using dimindices internally

* let users specify geometry type

* just use a broadcast

* don't need to be compatible with julia < 1.9

Co-authored-by: Rafael Schouten <[email protected]>

* deal with type stability better

* allow not providing `n`

* tweak the docstring

---------

Co-authored-by: Rafael Schouten <[email protected]>

* bugfix

* add tests