Spatial Parallel Computing by Hierarchical Data Partitioning

Geospatial data computation is parallelized by grid, hierarchy, or raster files. Based on 'future' (Bengtsson, 2024 ) and 'mirai' (Gao et al., 2025 ) parallel back-ends, 'terra' (Hijmans et al., 2025 ) and 'sf' (Pebesma et al., 2024 ) functions as well as convenience functions in the package can be distributed over multiple threads. The simplest way of parallelizing generic geospatial computation is to start from par_pad_*() functions to par_grid(), par_hierarchy(), or par_multirasters() functions. Virtually any functions accepting classes in 'terra' or 'sf' packages can be used in the three parallelization functions. A common raster-vector overlay operation is provided as a function extract_at(), which uses 'exactextractr' (Baston, 2023 ), with options for kernel weights for summarizing raster values at vector geometries. Other convenience functions for vector-vector operations including simple areal interpolation (summarize_aw()) and summation of exponentially decaying weights (summarize_sedc()) are also provided.

Latent Variable Models Diagnostics

Diagnostics and visualization tools for latent variable models fitted with 'lavaan' (Rosseel, 2012 ). The package provides fast, parallel-safe factor-score prediction (lavPredict_parallel()), data augmentation with model predictions, residuals, delta-method standard errors and confidence intervals (augment()), and model-based latent grids for continuous, ordinal, or mixed indicators (prepare()). It offers item-level empirical versus model curve comparison using generalized additive models for both continuous and ordinal indicators (item_data(), item_plot()) via 'mgcv' (Wood, 2017, ISBN:9781498728331), residual diagnostics including residual correlation tables and plots (resid_cor(), resid_corrplot()) using 'corrplot' (Wei and Simko, 2021 < https://github.com/taiyun/corrplot>), and Q–Q checks of residual z-statistics (resid_qq()), optionally with non-overlapping labels from 'ggrepel' (Slowikowski, 2024 < https://CRAN.R-project.org/package=ggrepel>). Heavy computations are parallelized via 'future'/'furrr' (Bengtsson, 2021 ; Vaughan and Dancho, 2018 < https://CRAN.R-project.org/package=furrr>). Methods build on established literature and packages listed above.

Genetic Population Level Functions

This collection of gene representation-independent functions implements the population layer of extended evolutionary and genetic algorithms and its support for the R-package 'xega' < https://CRAN.R-project.org/package=xega>. The population layer consists of functions for initializing, logging, observing, evaluating a population of genes, as well as of computing the next population. For parallel evaluation of a population of genes 4 execution models - named Sequential, MultiCore, FutureApply, and Cluster - are provided. They are implemented by configuring the lapply() function. The execution model FutureApply can be externally configured as recommended by Bengtsson (2021) . Configurable acceptance rules and cooling schedules (see Kirkpatrick, S., Gelatt, C. D. J, and Vecchi, M. P. (1983) , and Aarts, E., and Korst, J. (1989, ISBN:0-471-92146-7) offer simulated annealing or greedy randomized approximate search procedure elements. Adaptive crossover and mutation rates depending on population statistics generalize the approach of Stanhope, S. A. and Daida, J. M. (1996, ISBN:0-18-201-031-7). For 'xega''s architecture, see Geyer-Schulz, A. (2025) .

R Fortunes

A collection of fortunes from the R community.