Changes in version 0.2.3                        

Breaking changes

  - Disabled GPU for store_analyze_combos() — benchmarks showed GPU
    (Vulkan via ggmlR) is ~24x slower than CPU (C++) for this workload.
    ggmlR remains as optional dependency for other use cases.

New features

  - short_path_bfs.cpp — short_path_bfs() rewritten entirely in C++ with
    depth-limited BFS and multi-index lookup, replacing the previous R
    implementation for significantly better performance.

Bug fixes

  - Fixed non-ASCII characters in R code (replaced Russian text with
    English in diagnostic messages)
  - Fixed documentation mismatches: sort_by parameter now properly
    documented in find_best_random_combinations() and
    find_path_iterative()

                        Changes in version 0.2.2                        

Breaking changes

  - find_path_iterative() now uses C++ StateStore backend instead of
    data.frame lists — memory usage is linear instead of quadratic
  - analyze_top_combinations() still works but the new
    store_analyze_combos() writes directly to C++ store without
    intermediate data.frames

New features

  - C++ StateStore — compact hash-indexed state storage
    (src/state_store.h, src/state_store.cpp):
    
      - Flat vector<int> for states, separate metadata vectors,
        unordered_map hash index
      - Capacity starts at 10,000 and doubles on demand (amortized O(1)
        insert)
      - Operations encoded as int (1/2/3) instead of strings for cache
        efficiency
      - Incremental hash: keys computed only for new states, never
        recomputed
      - state_store_find_intersections() — O(min(N,M)) set intersection
        via hash
      - state_store_find_best_match() — Manhattan distance on flat array
      - state_store_filter_middle() — skip first/last steps per combo
      - state_store_reconstruct_path() — full C++ path reconstruction
      - state_store_to_dataframe() — convert to data.frame for debugging

  - store_analyze_combos() — C++ cycle expansion writing directly to
    StateStore, bypassing all R-level list/data.frame creation

  - sort_by parameter for find_best_random_combinations():
    
      - Flexible vector of sorting criteria: "longest", "shortest",
        "most_unique", "least_unique", "most_repeated", "least_repeated"
      - Criteria combine freely: e.g. c("shortest", "most_unique")
      - New repetition_ratio column in output
      - sort_by parameter propagated to find_path_iterative()

  - Bridge state output — find_path_iterative() and find_path_bfs() now
    print bridge state chains with 1-based numbering and labels

  - Direct vs hub distance check in find_path_bfs() — compares Manhattan
    distance start<->final vs hub_s<->hub_f, skips BFS hubs when direct
    is closer

Improvements

  - CelestialCoords struct extracted to shared header
    (celestial_coords.h) for reuse across C++ files
  - find_path_bfs() now returns bridge_states_start and
    bridge_states_final in its result, including BFS hub states

                 Changes in version 0.2.1 (2026-03-01)                  

Breaking changes

  - Removed arrow dependency — all functions now return plain data.frame
    instead of Arrow Tables
  - data.table moved from Imports to Suggests — package works without
    it, uses data.table for speed when available

Improvements

  - Significantly faster installation: removed heavy arrow dependency
    (~50+ MB)
  - Lightweight by default: only Rcpp is required

                        Changes in version 0.2.0                        

New features

  - GPU acceleration via ggmlR Vulkan backend (optional):
      - cayley_gpu_available(), cayley_gpu_init(), cayley_gpu_status(),
        cayley_gpu_free() — GPU infrastructure with lazy initialization
      - calculate_differences(..., use_gpu = TRUE) — Manhattan distance
        on GPU (sub -> abs -> sum_rows)
      - apply_operations_batch_gpu() — batch permutation operations via
        matrix multiplication on GPU
      - manhattan_distance_matrix_gpu() — pairwise N*M Manhattan
        distance matrix using 3D tensors
  - Sparse BFS: sparse_bfs(), reconstruct_bfs_path() — sparse BFS with
    hybrid hub/random selection
  - BFS pathfinding: find_path_bfs() — find path via BFS highways +
    iterative connector
  - Path shortening: short_path_bfs() — greedy BFS hopping to shorten
    existing paths
  - Bidirectional BFS pathfinding: bidirectional_bfs() for shortest path
    between permutation states
  - Iterative path solver: find_path_iterative() for finding paths via
    cycle expansion
  - Celestial coordinates: convert_LRX_to_celestial(),
    calculate_angular_distance_z(), calculate_midpoint_z(),
    find_closest_to_coords() — map operation counts to spherical
    coordinates
  - Combination analysis: analyze_top_combinations() for full cycle
    analysis of top operation sequences
  - State utilities: generate_state(), generate_unique_states_df(),
    select_unique(), check_duplicates(), save_bridge_states(),
    find_combination_in_states(), convert_digits()
  - Path utilities: invert_path(), validate_and_simplify_path(),
    reconstruct_full_path()
  - Distance metrics: manhattan_distance(), breakpoint_distance(),
    short_position()
  - Simple operation variants: shift_left_simple(),
    shift_right_simple(), reverse_prefix_simple()

Improvements

  - C++ implementations of all core operations via Rcpp with OpenMP
  - GPU functions fall back to CPU automatically when ggmlR/Vulkan is
    unavailable

                 Changes in version 0.1.0 (2025-11-25)                  

  - Initial CRAN submission.
  - Basic TopSpin operations: shift_left(), shift_right(),
    reverse_prefix()
  - Cycle analysis functions: get_reachable_states(),
    get_reachable_states_light()
  - Optimization tools: find_best_random_combinations()