[Not ready] add warning when mixing nodata/mask/alpha

.github/workflows/ci.yml

@@ -14,7 +14,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: [3.6, 3.7, 3.8, 3.9]
+        python-version: [3.7, 3.8, 3.9]
 
     steps:
       - uses: actions/checkout@v2
@@ -31,9 +31,8 @@ jobs:
       - name: Run Tox
         run: tox -e py
 
-      # Run pre-commit (only for python-3.8)
+      # Run pre-commit
       - name: run pre-commit
-        if: matrix.python-version == 3.8
         run: pre-commit run --all-files
 
       - name: Upload Results

CHANGES.md

@@ -1,3 +1,146 @@
+# 3.0.0 (TDB)
+
+* add `crs` property in `rio_tiler.io.base.SpatialMixin` (https://github.com/cogeotiff/rio-tiler/pull/429)
+* add `geographic_bounds` in `rio_tiler.io.base.SpatialMixin` to return bounds in WGS84 (https://github.com/cogeotiff/rio-tiler/pull/429)
+
+```python
+from rio_tiler.io import COGReader
+
+with COGReader("https://rio-tiler-dev.s3.amazonaws.com/data/fixtures/cog.tif") as cog:
+    print(cog.bounds)
+    >> (373185.0, 8019284.949381611, 639014.9492102272, 8286015.0)
+
+    print(cog.crs)
+    >> "EPSG:32621"
+
+    print(cog.geographic_bounds)
+    >> (-61.28762442711404, 72.22979795551834, -52.301598718454485, 74.66298001264106)
+```
+
+* Allow errors to be ignored when trying to find `zooms` for dataset in `rio_tiler.io.COGReader`. If we're not able to find the zooms in selected TMS, COGReader will defaults to the min/max zooms of the TMS (https://github.com/cogeotiff/rio-tiler/pull/429)
+
+```python
+from pyproj import CRS
+from morecantile import TileMatrixSet
+
+from rio_tiler.io import COGReader
+
+# For a non-earth dataset there is no available transformation from its own CRS and the default WebMercator TMS CRS.
+with COGReader("https://rio-tiler-dev.s3.amazonaws.com/data/fixtures/cog_nonearth.tif") as cog:
+    >> UserWarning: Cannot dertermine min/max zoom based on dataset informations, will default to TMS min/max zoom.
+
+    print(cog.minzoom)
+    >> 0
+
+    print(cog.maxzoom)
+    >> 24
+
+# if we use a `compatible TMS` then we don't get warnings
+europa_crs = CRS.from_authority("ESRI", 104915)
+europa_tms = TileMatrixSet.custom(
+    crs=europa_crs,
+    extent=europa_crs.area_of_use.bounds,
+    matrix_scale=[2, 1],
+)
+with COGReader(
+    "https://rio-tiler-dev.s3.amazonaws.com/data/fixtures/cog_nonearth.tif",
+    tms=europa_tms,
+) as cog:
+    print(cog.minzoom)
+    >> 4
+
+    print(cog.maxzoom)
+    >> 6
+```
+
+* compare dataset bounds and tile bounds in TMS crs in `rio_tiler.io.base.SpatialMixin.tile_exists` method to allow dataset and TMS not compatible with WGS84 crs (https://github.com/cogeotiff/rio-tiler/pull/429)
+* use `httpx` package instead of requests (author @rodrigoalmeida94, https://github.com/cogeotiff/rio-tiler/pull/431)
+* warn user when dataset has Nodata value in addition to a Mask or an Alpha band (https://github.com/cogeotiff/rio-tiler/pull/402)
+
+**breaking changes**
+
+* update morecantile requirement to version >=3.0 (https://github.com/cogeotiff/rio-tiler/pull/418)
+* remove python 3.6 support (https://github.com/cogeotiff/rio-tiler/pull/418)
+* remove `max_size` defaults for `COGReader.part` and `COGReader.feature`, which will now default to full resolution reading.
+
+```python
+# before
+with COGReader("my.tif") as cog:
+    img = cog.part(*cog.dataset.bounds, dst_crs=cog.dataset.crs, bounds_crs=cog.dataset.crs)
+    # by default image should be max 1024x1024
+    assert max(img.width, 1024) # by default image should be max 1024x1024
+    assert max(img.height, 1024)
+
+# now (there is no more max_size default)
+with COGReader("my.tif") as cog:
+    img = cog.part(*cog.dataset.bounds, dst_crs=cog.dataset.crs, bounds_crs=cog.dataset.crs)
+    assert img.width == cog.dataset.width
+    assert img.height == cog.dataset.height
+```
+
+* deprecate `.metadata` and `.stats` methods (https://github.com/cogeotiff/rio-tiler/pull/425)
+* add `.statistics` method in base classes (https://github.com/cogeotiff/rio-tiler/pull/427)
+
+* remove `rio_tiler.io.base.SpatialMixin.spatial_info` and `rio_tiler.io.base.SpatialMixin.center` properties (https://github.com/cogeotiff/rio-tiler/pull/429)
+
+* Reader's `.bounds` property should now be in dataset's CRS, not in `WGS84` (https://github.com/cogeotiff/rio-tiler/pull/429)
+
+```python
+# before
+with COGReader("my.tif") as cog:
+    print(cog.bounds)
+    >>> (-61.287001876638215, 15.537756794450583, -61.27877967704677, 15.542486503997608)
+
+# now
+with COGReader("my.tif") as cog:
+    print(cog.bounds)
+    >>> (683715.3266400001, 1718548.5702, 684593.2680000002, 1719064.90736)
+
+    print(cog.crs)
+    >>> EPSG:32620
+
+    print(cog.geographic_bounds)
+    >>> (-61.287001876638215, 15.537756794450583, -61.27877967704677, 15.542486503997608)
+```
+
+* Use `RIO_TILER_MAX_THREADS` environment variable instead of `MAX_THREADS` (author @rodrigoalmeida94, https://github.com/cogeotiff/rio-tiler/pull/432)
+* remove `band_expression` in `rio_tiler.io.base.MultiBandReader` (https://github.com/cogeotiff/rio-tiler/pull/433)
+* change `asset_expression` input type from `str` to `Dict[str, str]` in `rio_tiler.io.base.MultiBaseReader` (https://github.com/cogeotiff/rio-tiler/pull/434)
+
+```python
+# before
+with STACReader("mystac.json") as stac:
+    img = stac.preview(
+        assets=("data1", "data2"),
+        asset_expression="b1*2",  # expression was applied to each asset
+    )
+
+# now
+with STACReader("mystac.json") as stac:
+    img = stac.preview(
+        assets=("data1", "data2"),
+        asset_expression={"data1": "b1*2", "data2": "b2*100"},  # we can now pass per asset expression
+    )
+```
+
+* add `asset_indexes` in `rio_tiler.io.base.MultiBaseReader`, which replaces `indexes`. (https://github.com/cogeotiff/rio-tiler/pull/434)
+
+```python
+# before
+with STACReader("mystac.json") as stac:
+    img = stac.preview(
+        assets=("data1", "data2"),
+        indexes=(1,),  # indexes was applied to each asset
+    )
+
+# now
+with STACReader("mystac.json") as stac:
+    img = stac.preview(
+        assets=("data1", "data2"),
+        asset_indexes={"data1": 1, "data2": 2},  # we can now pass per asset indexes
+    )
+```
+
 # 2.1.3 (2021-09-14)
 
 * Make sure output data is of type `Uint8` when applying a colormap (https://github.com/cogeotiff/rio-tiler/pull/423)

README.md

@@ -89,8 +89,21 @@ At the low level, `rio-tiler` is *just* a wrapper around the [rasterio.vrt.Warpe
 
     with STACReader("item.json") as stac:
         print(stac.assets)  # available asset
-        img = stac.tile(x, y, z, assets="asset1", indexes=(1, 2, 3))  # read tile for asset1 and indexes 1,2,3
-        img = stac.tile(x, y, z, assets=("asset1", "asset2", "asset3",), indexes=(1,))  # create an image from assets 1,2,3 using their first band
+        img = stac.tile(  # read tile for asset1 and indexes 1,2,3
+            x,
+            y,
+            z,
+            assets="asset1",
+            indexes=(1, 2, 3),
+        )
+
+        img = stac.tile(  # create an image from assets 1,2,3 using their first band
+            x,
+            y,
+            z,
+            assets=("asset1", "asset2", "asset3",),
+            asset_indexes={"asset1": 1, "asset2": 1, "asset3": 1},
+        )
     ```
 
 - [Mosaic](https://cogeotiff.github.io/rio-tiler/mosaic/) (merging or stacking)
@@ -136,26 +149,6 @@ $ pip install -U pip
 $ pip install -e .
 ```
 
-#### GDAL>=3.0 / PROJ>=6.0 performances issue
-
-`rio-tiler` is often used for dynamic tiling, where we need to perform small tasks involving cropping and reprojecting the input data. Starting with GDAL>=3.0 the project shifted to PROJ>=6, which introduced new ways to store projection metadata (using a SQLite database and/or cloud stored grids). This change introduced a performance regression as mentioned in https://mapserver.gis.umn.edu/id/development/rfc/ms-rfc-126.html:
-
-> using naively the equivalent calls proj_create_crs_to_crs() + proj_trans() would be a major performance killer, since proj_create_crs_to_crs() can take a time in the order of 100 milliseconds in the most complex situations.
-
-We believe the issue reported in [issues/346](https://github.com/cogeotiff/rio-tiler/issues/346) is in fact due to :point_up:.
-
-To get the best performances out of `rio-tiler` we recommend for now to use GDAL **2.4** until a solution can be found in GDAL or in PROJ.
-
-Note: Starting with rasterio 1.2.0, rasterio's wheels are distributed with GDAL 3.2 and thus we recommend using rasterio==1.1.8 if using the default wheels, which include GDAL 2.4.
-
-Links:
-
-- http://rgdal.r-forge.r-project.org/articles/PROJ6_GDAL3.html
-- https://mapserver.gis.umn.edu/id/development/rfc/ms-rfc-126.html
-- https://github.com/OSGeo/gdal/issues/3470
-- https://github.com/OSGeo/gdal/issues/1662
-
-
 ## Plugins
 
 #### [**rio-tiler-pds**][rio-tiler-pds]