GeoThinneR 2.1.1

Changed

• Clarified printed message and documentation for the optional spatial coverage metric when s2 (>= 1.1.0) is not available.

Fixed

• Fixed integer overflow warning with large datasets when using distance-based thinning with local kd-trees method. The issue was in the split() function used to assign points to grid cells. (#4)

GeoThinneR 2.1.0

Added

• Added is_lonlat() helper function to check for valid longitude/latitude ranges.
• A warning is printed when coordinates lie outside the typical global longitude/latitude ranges.
• summary() now allows to choose which trial to summarise using the trial argument.
• Support for the priority parameter in the "distance" thinning method. When multiple candidate points have the same number of neighbors, the point with the lowest priority is removed (#1).
• Support for NA values in the priority vector. These are treated as the lowest priority and trigger a warning.

Changed

• The "grid" and "precision" methods now handle priority ties by randomly selecting among equally prioritized points.
• Updated the vignette section on the priority parameter.
• Added s2 (>= 1.1.0) to Suggests in the DESCRIPTION file.

Fixed

• print() now reports both the number of trials run and returned (#3).
• summary() no longer errors when s2 is missing or too old. Now spatial coverage is set to NA with a message (#3).

GeoThinneR 2.0.0

Breaking changes. This version is backward-incompatible with earlier versions of GeoThinneR.

Major changes

• Introduced a new GeoThinned S3 class to represent thinning results. All calls to thin_points() now return a GeoThinned object instead of a list of logical vectors or data frames.
• Added new helper methods for GeoThinned objects:
  • print() and summary() for inspection.
  • plot() for visual comparison of thinned vs. original data.
  • largest() to retrieve the trial with the most retained points.
  • largest_index() to get the index of the best trial.
  • get_trial() to access specific trials.
• Redesigned thin_points() to return structured output and provide a more consistent user interface.
• Output structure and behavior of all thinning methods has been updated for compatibility with the new object-oriented design.
• We have improved the thinning algorithms by implementing two new modifications of the kd-tree method: local kd-trees (local_kd_tree) and restricted search (k_estimation).
• The round and hashing method (round_hash) now is deprecated and has been removed from the package because of presenting lower accuracy than the other distance-based methods.
• The thinning method argument method now includes the three main thinning strategies implemented in GeoThinneR for better understanding and usability. These are: "distance", "grid", and "precision".
• The search_type argument has been included for distance-based thinning (method = "distance") to select between the search algorithms ("brute", "kd_tree", "local_kd_tree", and "k_estimation").
• The n argument for the grid-based thinning (method = "grid") now lets you specify the number of points to retain per grid cell.

Minor changes

• The long_col argument now is called lon_col for consistency with the lat_col argument.
• The default column names (lon_col and lat_col) have been changed from "decimalLongitude" and "decimalLatitude" to "lon" and "lat".
• Improved documentation, error messages, and reproducibility behavior across all functions.
• Updated the package vignette to reflect all changes.

GeoThinneR 1.1.0

• Removed non-CRAN dependency (rtree) and associated functions related to R-Trees structure thinning method (r_tree_thinning()).
• Minor improvements to documentation.

GeoThinneR 1.0.0

• First release