Performance Improvement to Windowed Reading

[cmarshak]

Main thrust is the change to function merge_tile_datasets_within_extent so that merging and reading can be done efficiently. Formerly read full tile for each path/url. Now only reads data required for merge.

Changed

• Update merge_tile_datasets_within_extent (formely named merged_tile_datasets) to only read data within provided extents
  • Requires extent (i.e. list[float]) as input now.
• Internally, swap use of m x n arrays (with total dimensions 2) to the 3 dimensional arrays c x m x n. Specifically, use band interleaved by pixel (BIP) format where c is the number of channels. Although the API remains unchanged (outputs 2 dimensional array), the intermediate functions are slightly more general and applicable.
  • merged.py - all functions now accept BIP (3d arrays) and return them
  • geoid.py - all functions return and expect BIP (3d arrays) including the input dem array.
• Improved performance of merge by reading only the extent that is required.
• Typing for 3.9+
• Use pyproject.toml for installation.

Added

• Support for 1/3 arc second 3Dep
• Tests using golden datasets and mocked tiles/geoid - ensures correctness of transformations

Removed

• Support for NED1 and 3Dep 1 arcsecond
• Support for Python 3.7 and 3.8
• setup.py

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB