Skip to contents

Changelog

Source: NEWS.md

spesim 0.5.2

Performance

  • Vectorised environment-acceptance step. spesim_simulate_neutral_recruitment() previously called .env_accept_prob() once per candidate point via vapply (40 calls per simulation step), with two redundant costs inside every call:

    • .to_grid_index() recomputed the grid midpoints vector ((grid_vals[-1] + grid_vals[-length(grid_vals)]) / 2) on every invocation.
    • .resolve_env_col() re-ran its column-name matching on every invocation.

    Both costs are now eliminated. Grid midpoints (mids_gx / mids_gy) are precomputed once at setup; the resolved column name (env_col) is stored directly in grad_lookup at setup time. The scalar helper is replaced by a single vectorised function, .env_accept_prob_vec(), that accepts the whole candidate matrix and returns a probability vector via one findInterval call and one cbind-indexed matrix lookup over all candidates at once. Self-time of the acceptance step at the benchmark configuration (N = 1,500, S = 15, hybrid model, profvis at 10 ms interval) dropped from ≈ 2,740 ms to ≈ 100 ms — a 27× reduction.

  • Vectorised candidate displacement sampling. The candidate-position proposal loop — vapply over 40 displacements per step, each calling .propose_displacement() → .step_magnitude() → a scalar rnorm/rexp/runif — is replaced by two batch functions: .propose_displacement_batch() and .propose_linear_pos_batch(). Each samples all batch_size step magnitudes and angles in a single RNG call (e.g. stats::rnorm(40) instead of 40 × stats::rnorm(1)), then computes candidate positions with vectorised arithmetic. The vapply closure overhead and 40 individual R-level dispatches to .step_magnitude() disappear entirely. Combined with the acceptance-step improvement above, total profiled wall time for the benchmark configuration fell from 17,690 ms (v0.5.1) to 2,600 ms — a 6.8× overall speedup.

Bug fixes

  • calculate_quadrat_environment() now always returns a character site column. When quadrats$quadrat_id was integer the resulting site column inherited that type, causing dplyr::left_join() to error with “Can’t join x$site with y$site ” when joined against the always-character site produced by create_abundance_matrix(). The merge argument is now as.character(quadrats$quadrat_id), making the two functions type-consistent in all cases. This silently prevented the spesim-method-testing and spesim-ordination-dbrda vignettes from rebuilding under R CMD check.

spesim 0.5.1

Performance

  • C++ point-in-polygon engine. All polygon domain checks previously used sf::st_as_sf() + sf::st_within() on every call, incurring full CRS comparison and GEOS setup overhead. They now delegate to a compiled C++ ray-casting implementation (pip_cpp), with ring coordinates extracted once and cached for the duration of any tight loop. Measured speedups at representative call sizes: 6–13× faster for batches of 1–40 points, 112× faster for the sequential-inhibition (Strauss) single-point loop.

  • Vectorised initial community placement. spesim_simulate_neutral_recruitment() previously sampled one position per individual inside a for loop (J individual calls to sf::st_within). Placement is now done in a single vectorised call to .sample_uniform_in_domain(), eliminating J − 1 polygon coordinate extractions and sf round-trips.

  • Vectorised sf point construction. .as_sfc_points() replaced a lapply(sf::st_point(...)) loop with a single sf::st_as_sf() call, which is the recommended vectorised path.

  • Shared pip tester across all helpers. .sample_uniform_in_domain(), .simulate_strauss_points(), and .linear_pos_to_xy() all accept an optional pip_fn argument so callers that already hold a pre-built tester avoid redundant ring-coordinate extraction.

spesim 0.4.6

Maintenance

  • Added rnaturalearthdata to Suggests to satisfy CRAN dependency declaration (used in vignette via requireNamespace()).

spesim 0.4.5

Bug fixes + promised features now active

  • Fixed generate_fisher_log_series() for n_species = 1 (now sums to n_individuals).
  • Fixed fast Strauss MH sampler state updates (accept/reject now consistent).
  • Fixed Thomas and Strauss/Geyer fast helpers to be polygon-aware (bbox simulate + clip + thin/top-up).
  • Implemented environmental filtering + local interspecific interactions during species assignment as documented.
  • Distance–decay now aligns site_coords to abund_matrix by site id and uses a robust Sorensen dissimilarity for presence/absence.
  • ZSM_M is now active: when supplied as a finite value in (0, 1), SAD_MODEL = "zsm" switches from the theta-only Ewens (Chinese restaurant) sampler to a Moran-style death–birth with immigration heuristic; lower m produces a more uneven SAD. The NEWS 0.4.2 entry below stating ZSM_M is “ignored” is superseded by this release.

spesim 0.4.2

Vignettes / neutral SAD clarification

  • Updated point-process recipes vignette with an explicit explanation of the current neutral-theory SAD helper (what it can and cannot do), plus worked examples and advanced panels.
  • SAD_MODEL = "zsm" is now documented and implemented as a theta-only Ewens sampler; ZSM_M is currently ignored (accepted for forward compatibility).

spesim 0.4.1

Maintenance

  • Build/check hygiene improvements (CRAN incoming note reductions).

spesim 0.4.0

Runner refactor + stable results

  • spesim_run() is now the recommended, consolidated entry point and returns a stable S3 results object (spesim_result).
  • run_spatial_simulation() is retained as a deprecated wrapper around spesim_run() for backwards compatibility.
  • Vignettes and README updated to reflect the new workflow.

spesim 0.3.19

Startup message

spesim 0.3.18

Concave domain complexity control

  • create_sampling_domain(shape = "concave") gains concave_lobes (2–5) to control large-scale outline complexity; lower values yield simpler concave polygons.

spesim 0.3.17

Default domain shape diversity

  • create_sampling_domain() now supports multiple shape families (concave/irregular, convex hulls, ellipses/circles, and boxy rectangles/squares), and the default uses a mixed generator to better mimic real-world landscape patch shapes.

spesim 0.3.16

Default domain variability in spesim_run()

  • spesim_run() now generates a new random default domain for each run when domain = NULL and no seed is supplied.
  • When seed is supplied, the default domain is reproducible across runs.

Plotting side-effects

spesim 0.3.15

Ordination teaching vignette (db-RDA)

  • New vignette: constrained ordination on simulated data using distance-based RDA (db-RDA) via vegan::capscale().

spesim 0.3.14

Build hygiene and vignette snippet checks

  • Removed pkgdown/vignette build warnings by aligning vignette titles and setting highlight: null for html vignettes.
  • Added a test to ensure all vignette code chunks are syntactically valid (knitr::purl() + parse()).

spesim 0.3.13

Performance notes

  • New vignette: “Performance notes” (spesim-performance) with conservative interactive ranges and practical tips.

spesim 0.3.12

Method-testing vignette

  • New vignette: “Method testing with spesim” (spesim-method-testing).

spesim 0.3.11

Method-testing bundle helper

  • New spesim_method_test() convenience function returns res, audit, and optional standard tables/plots for teaching and method-testing.

spesim 0.3.10

Report: configurable conceptual audit

  • generate_full_report() gains include_audit and audit_top_n to control the report’s conceptual audit section.
  • Report audit now runs spesim_audit(..., diagnostics = "nn") for deterministic, lightweight behaviour.

spesim 0.3.9

Filtering audits: occupancy-aware diagnostics

  • audit_environmental_filtering() now reports occupancy_sites (number of quadrats with abundance > 0).
  • Low occupancy is flagged as too_sparse_to_assess, helping prevent over-interpretation of noisy correlations.

spesim 0.3.8

Filtering diagnostics: plotting helper

  • Add plot_filtering_response() to visualise environment–abundance relationships for gradient-responsive species.
  • The plot shows the configured optimum (dashed line) and tolerance band (shaded) in natural units.
  • Environmental gradients vignette updated with a filtering sanity-check section.

spesim 0.3.7

Sampling audits: effective sampling frame (area-based)

  • Add estimate_sampling_frame() and include a conservative boundary constraint summary in placement_audit:
    • safe_area_fraction = area(safe_domain)/area(domain), where safe_domain is the domain buffered inward by half the quadrat diagonal.
    • buffer_dist (half-diagonal) and quadrat_size recorded for context.
  • Quadrat placement audits and audit_sampling_scheme() now report the effective sampling frame fraction alongside exclusion/rejection rates.

spesim 0.3.6

Sampling audits: standardise placement_audit

  • Quadrat placement functions now attach their placement_audit attribute via shared helpers and normalise a minimal common schema.
  • audit_sampling_scheme() now validates/normalises placement_audit before reporting.
  • Quadrat placement vignette updated with guidance on inspecting placement_audit for boundary exclusion.

spesim 0.3.5

Method-testing tutor: edge-corrected spatial diagnostics (optional)

  • Extend spesim_audit() with diagnostics = c("nn", "spatstat").
  • Add optional spatstat-based, edge-corrected spatial summaries in the true irregular window:
    • Ripley-style departure summary via L(r)-r (border correction)
    • pair correlation function extrema (Ripley correction)
  • New helper: spesim_spatstat_diagnostics() (used internally; requires spatstat.geom + spatstat.explore).

spesim 0.3.4

Method-testing tutor: regime classification

  • Add spesim_regime() to convert audit metrics into teaching-friendly green/amber/red regime classifications.
  • spesim_audit() now includes a regime element and print.spesim_audit() prints the classification summary.

spesim 0.3.3

Method-testing audits

  • Add spesim_audit() plus supporting helpers to quantify whether a realised run matches the intended qualitative regime:
    • nearest-neighbour spatial diagnostics per species (clustered/CSR-like/inhibited)
    • environment–abundance checks for gradient-responsive species (optima/tolerances vs realised quadrat environments)
    • sampling scheme boundary exclusion / rejection summaries (when available)
  • Quadrat placement functions now attach a placement_audit attribute with rejection/exclusion metrics (scheme-dependent).
  • The text report now includes a Conceptual audit section summarising these checks.

spesim 0.3.2

Documentation

  • Add Start here article to improve pkgdown discoverability.
  • Add a model card vignette clarifying what spesim does/does not model and how to interpret results.
  • Add a validation & sanity checks vignette with suggested diagnostics (including optional spatstat checks).
  • Clarify that the default sampling domain is synthetic/unitless (users can supply a real-world sf domain).

Reproducibility

  • Reset set.seed(P$SEED) before generating advanced panels and text reports when write_outputs = TRUE, improving determinism for vegan-based components.

spesim 0.3.1

Usability

spesim 0.3.0

  • Add spesim_demo() convenience wrapper for teaching/quick exploration.
  • Improve pkgdown build cleanliness (vignette titles, NEWS headings).

spesim 0.2.0-dev

  • Implemented Thomas (Rcpp), Strauss (Rcpp fast), and Geyer (Rcpp) point processes for clustered species.