Generalized SMC Algorithms

.Rbuildignore

@@ -1,34 +1,36 @@
+^.*\.Rproj$
+^\.Rproj\.user$
 README.Rmd
 \.git
 \.gitignore
+^\.github$
 \.travis\.yml
 \.DS_Store
-\.\.Rcheck
-builder.sh
+^\.vscode$
+^\.cache$
+^\.claude$
+^\.clang-format$
+^src/\.clang-format-ignore$
+^src/cpp_code_overview\.md$
+^air\.toml$
+^jarl\.toml$
+^compile_commands\.json$
 ^.*\.o$
-^doc$
+^\.\.Rcheck$
 ^Meta$
+^doc$
+builder.sh
 ^_pkgdown\.yml$
 ^docs$
 ^pkgdown$
-^.*\.Rproj$
-^\.Rproj\.user$
-^cran-comments\.md$
-^CRAN-RELEASE$
 ^R/dev_helpers\.R$
-^inst/enumpart/enumpart\.exe$
 ^data-raw$
 ^explore$
-^\.github$
-^LICENSE\.md$
+^inst/enumpart/enumpart$
+^inst/enumpart/enumpart\.exe$
+^cran-comments\.md$
+^CRAN-RELEASE$
 ^CRAN-SUBMISSION$
 ^revdep$
-^\.cache$
-^\.claude$
-^\.vscode$
-^\.clang-format$
-^src/\.clang-format-ignore$
+^LICENSE\.md$
 ^CONTRIBUTING\.md$
-^air\.toml$
-^jarl\.toml$
-^compile_commands\.json$

.github/workflows/check-standard.yaml

@@ -4,7 +4,7 @@ on:
   push:
     branches: [main, master]
   pull_request:
-    branches: [main, master]
+  workflow_dispatch:
 
 name: R-CMD-check
 

.gitignore

@@ -1,29 +1,38 @@
+# R / RStudio
 .Rproj.user
-.vscode
-.DS_Store
 .Rhistory
 
-.claude/
+# OS / editors
+.DS_Store
+.vscode
 .cache/
+.claude/
 CLAUDE.md
+compile_commands.json
 
-*.so
+# Compiled objects
 *.o
-*.rds
-*.tmp
+*.so
+*.dll
 *.gch
 *.pch
-*.dll
 
-inst/doc
-inst/enumpart/enumpart
+# R data / temp
+*.rds
+*.tmp
+
+# Build artifacts
 Meta
 doc
+inst/doc
+src-i386/
+
+# pkgdown / vignette caches
+docs
 docs/articles/*_cache/*
 vignettes/*_cache/*
-src-i386/
+
+# Package-specific
+inst/enumpart/enumpart
 inst/enumpart/enumpart.exe
 redist/tests/testthat/_snaps
-docs
-
-compile_commands.json

DESCRIPTION

@@ -1,12 +1,13 @@
 Package: redist
-Version: 4.3.2.9000
-Date: 2026-02-28
+Version: 5.0.0.1
 Title: Simulation Methods for Legislative Redistricting
 Authors@R: c(
     person("Christopher T.", "Kenny", email = "ctkenny@proton.me", 
            role = c("aut", "cre"), comment = c(ORCID = "0000-0002-9386-6860")),
     person("Cory", "McCartan", email = "mccartan@psu.edu", role = "aut",
            comment = c(ORCID = "0000-0002-6251-669X")),
+    person("Philip", "O'Sullivan", email = "posullivan@fas.harvard.edu", role = "aut",
+           comment = c(ORCID = "0000-0002-9665-2462")),
     person("Ben", "Fifield", email = "benfifield@gmail.com", role = "aut",
            comment = c(ORCID = "0000-0002-2247-0201")),
     person("Kosuke", "Imai", email = "imai@harvard.edu", role = "aut",
@@ -61,11 +62,17 @@ Suggests:
     spelling,
     igraph,
     pkgload
-LinkingTo: Rcpp, RcppArmadillo, RcppThread, cli, redistmetrics
+LinkingTo: 
+    Rcpp,
+    RcppArmadillo,
+    RcppEigen,
+    RcppThread,
+    redistmetrics,
+    cli
+License: GPL (>= 2)
 Encoding: UTF-8
 Language: en-US
 LazyData: true
-License: GPL (>= 2)
 SystemRequirements: C++17
 NeedsCompilation: yes
 BugReports: https://github.com/alarm-redist/redist/issues

NAMESPACE

@@ -59,19 +59,24 @@ S3method(ungroup,redist_plans)
 S3method(weights,redist_plans)
 export(add_constr_compet)
 export(add_constr_custom)
+export(add_constr_custom_plan)
 export(add_constr_edges_rem)
 export(add_constr_fry_hold)
 export(add_constr_grp_hinge)
 export(add_constr_grp_inv_hinge)
 export(add_constr_grp_pow)
 export(add_constr_incumbency)
 export(add_constr_log_st)
+export(add_constr_min_group_frac)
 export(add_constr_multisplits)
+export(add_constr_plan_incumbency)
+export(add_constr_plan_splits)
 export(add_constr_polsby)
 export(add_constr_pop_dev)
 export(add_constr_segregation)
 export(add_constr_splits)
 export(add_constr_status_quo)
+export(add_constr_total_plan_splits)
 export(add_constr_total_splits)
 export(add_reference)
 export(as_redist_map)
@@ -82,15 +87,19 @@ export(compare_plans)
 export(competitiveness)
 export(county_splits)
 export(distr_compactness)
+export(est_norm_biased)
+export(est_norm_unbiased)
 export(filter)
 export(freeze)
 export(get_adj)
 export(get_existing)
+export(get_existing_seats)
 export(get_mh_acceptance_rate)
 export(get_plans_matrix)
 export(get_plans_weights)
 export(get_pop_tol)
 export(get_sampling_info)
+export(get_seats_matrix)
 export(get_target)
 export(group_frac)
 export(is_contiguous)
@@ -204,7 +213,6 @@ importFrom(cli,cli_inform)
 importFrom(cli,cli_text)
 importFrom(cli,cli_warn)
 importFrom(doRNG,"%dorng%")
-importFrom(dplyr,"%>%")
 importFrom(dplyr,.data)
 importFrom(dplyr,all_of)
 importFrom(dplyr,arrange)

NEWS.md

@@ -1,9 +1,48 @@
-# redist (development)
-
+# redist 5.0.0
+* Replaces old SMC weights with generally lower variance optimal weights.
+* Adds the option to add Mergesplit MCMC steps at any point during an SMC run. 
+Adding mergesplit steps can help achieve convergence for plans with a larger 
+number of districts without increasing the sample size. 
+* Improves SMC and Mergesplit MCMC performance by pre-allocating and reusing as 
+much memory as possible while drawing spanning trees.
+* Introduces new methods for sampling plans for both SMC and Mergesplit MCMC. 
+The final output is still plans from the same distribution as before but new 
+sampling spaces and splitting methods will sometimes perform better under some
+scenarios.
+* Introduces a new method for splitting in SMC - generalized region splits.
+Instead of splitting off one district at a time this allows for splitting into
+two arbitrary sized regions. For an equal sample size generalized region splits 
+tends to converge slower but it is typically much faster (up to twice as fast or 
+more) since on average it draws spanning trees on smaller subgraphs then 
+single district splits. 
+* Adds support for sampling multimember district plans with both SMC and 
+mergesplit MCMC under some mild conditions. The district seat sizes (how many 
+legislators a district can have) must be a range of values e.g. (3,4,5) and no 
+district seat size can be the sum of two others. 
+* When counties are used `redist_mergesplit` now samples from the same target 
+distribution as `redist_smc` (it guarantees no more than the number of districts
+minus 1 splits). 
+* `redist_mergesplit` inputs now work differently. 
+    * `nsims` is now the number of plans saved. 
+    * `warmup` is the number of steps to run the chain for before collecting any samples.
+    * `thin` means we will run the chain for `thin - 1` steps between saving plans
+    * Overall the chain will be run for `warmup + nsims * thin` and return `nsims` plans.
+* Adds the option to incorporate rejection sampling for all constraints in SMC
+and mergesplit MCMC. Any constraint can now include a threshold argument `thresh` 
+where for a newly split plan if either of the two new regions has a raw score 
+greater than or equal to `thresh` then the plan will be automatically reject. 
+This amounts to giving plans where any region has a score above `thresh` a 
+probability of 0. 
+* Updates the target distribution when counties are turned on. For more details 
+see the forthcoming working paper. 
+* The mergesplit backend for `redist_shortburst` now uses uniform edge sampling 
+with forest space for the backend instead of sampling with graph space and all
+`k` related parameters have been removed. 
+
+# redist 4.4.0
 * Old `enumpart` functions are deprecated in favor of `redist_enumpart()` which provides a more consistent syntax.
 
 # redist 4.3.2
-
 * Allows for parallel flip with `chains` argument in `redist_flip()`.
 * Fixes URL issues causing a note on CRAN. We have pointed the links to our website directly.