Version 2.0 GeoTensor to implement numpy API #21
Conversation
Co-authored-by: Copilot <[email protected]>
There was a problem hiding this comment.
Pull Request Overview
This PR introduces version 2.0 of the GeoTensor functionality with added support for the NumPy API and several documentation improvements. Key changes include:
- Updating the package version in pyproject.toml to 2.0.0a2.
- Revising navigation and content in the documentation (mkdocs.yml, various markdown files) to reflect the new GeoTensor features and usage examples.
- Updating example code in docs/index.md, README.md, and related files to use RasterioReader for Sentinel-2 imagery instead of the previously used S2_SAFE_reader.
Reviewed Changes
Copilot reviewed 15 out of 15 changed files in this pull request and generated no comments.
Show a summary per file
| File | Description |
|---|---|
| pyproject.toml | Version bump update |
| mkdocs.yml | Added new navigation entries for GeoReader concepts and new modules |
| docs/modules/read_module.md | Updated API sections and headings |
| docs/modules/plot_module.md | New module documentation for plotting functions |
| docs/modules/geotensor_module.md | Expanded documentation for GeoTensor and its NumPy API support |
| docs/index.md | Revised example code and updated imagery source details |
| docs/Sentinel-2/run_in_gee_image.ipynb | Added installation instructions and updated Python version info |
| README.md | Revised example code and updated imagery source details |
Comments suppressed due to low confidence (3)
docs/index.md:35
- Typo: 'croping' should be corrected to 'cropping'.
See also read.read_from_bounds, read.read_from_polygon for different ways of croping an image
README.md:35
- Typo: 'croping' should be corrected to 'cropping'.
See also read.read_from_bounds, read.read_from_polygon for different ways of croping an image
docs/modules/geotensor_module.md:5
- Typo: 'subseting' should be corrected to 'subsetting'.
allow for operations like reprojection, subseting from spatial coordinates, mosaicking or vectorization.
There was a problem hiding this comment.
Pull Request Overview
This PR implements GeoTensor version 2.0 with a refactored numpy API for GeoTensor, along with several documentation updates and minor internal logic adjustments.
- Update package version and breaking changes in pyproject.toml
- Adjust navigation and documentation to highlight new GeoTensor and API features
- Modify vectorization and plotting functions to support new masking and extent calculations
Reviewed Changes
Copilot reviewed 19 out of 19 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| pyproject.toml | Updated version to "2.0.0a3" reflecting the new major release |
| mkdocs.yml | Added new navigation items for updated modules and plotting docs |
| georeader/vectorize.py | Changed type checking logic for handling GeoData vs numpy arrays |
| georeader/plot.py | Revised extent calculation for non-rectilinear transforms and mask handling |
| docs/* | Updated documentation to include new GeoTensor API and usage examples |
| Makefile | Added target to publish alpha docs under a separate documentation path |
| if not hasattr(binary_mask, "transform"): | ||
| binary_mask_np = binary_mask | ||
| else: | ||
| binary_mask_np = binary_mask.values | ||
| binary_mask_np = np.array(binary_mask) | ||
|
|
||
| assert transform is None, "transform only must be used if input is np.ndarray" | ||
| transform = binary_mask.transform |
There was a problem hiding this comment.
Using hasattr() here as a replacement for isinstance(binary_mask, np.ndarray) might not capture all intended types. Consider confirming that this condition correctly distinguishes raw numpy arrays from GeoData objects.
| if max_val_mask is None: | ||
| max_val_mask = color_array.shape[0] | ||
|
|
||
| assert (max_val_mask - min_val_mask) == color_array.shape[0], f"max_val_mask - min_val_mask must be equal to the number of colors {max_val_mask} - {min_val_mask} != {color_array.shape[0]}" |
There was a problem hiding this comment.
Using 'assert' for validating the mask range could be problematic if assertions are disabled in production environments. Consider raising a ValueError with a clear error message instead.
This PR has many new features for the
GeoTensorclass. In particularGeoTensornow inherits fromnp.ndarrayand therefore impllements most of numpy API.In theory the only breaking change is the one mention below. Since GeoTensor is a central class of the package, I decided to create a new major version.
See GeoTensor module docs and GeoTensor numpy API tutorial for detailed example of new features.
Alpha docs of this version at: https://spaceml-org.github.io/georeader/alpha/
Breaking changes
torch.Tensoras inner value forGeoTensor, I don't think it was a feature used.These changes are already as an alpha version in pypi. In order to install this alpha version:
Pending numpy API features
These functions will work but the propagated
transformwith not be correct:reshapeto allow flatteing time and band axis.np.rot90)np.transposeto allow transposing x and y axes (and transposing thetransform)np.flip,np.rollaxis,np.moveaxis....