Package {scip}

Title:	Interface to the SCIP Optimization Suite
Version:	1.10.1-1
Description:	Provides an R interface to SCIP (Solving Constraint Integer Programs), a framework for mixed-integer programming (MIP), mixed-integer nonlinear programming (MINLP), and constraint integer programming (2025, <doi:10.48550/arXiv.2511.18580>). Supports linear, quadratic, SOS, indicator, and knapsack constraints with continuous, binary, and integer variables. Includes a one-shot solver interface and a model-building API for incremental problem construction.
License:	Apache License (≥ 2)
URL:	https://bnaras.github.io/scip/, https://scipopt.org/
BugReports:	https://github.com/bnaras/scip/issues
Depends:	R (≥ 4.0)
Imports:	methods, Matrix
Suggests:	tinytest, slam, knitr, rmarkdown
VignetteBuilder:	knitr
SystemRequirements:	CMake (>= 3.11), GNU make, C++17
Encoding:	UTF-8
NeedsCompilation:	yes
Config/roxygen2/version:	8.1.0
Packaged:	2026-10-04 15:57:56 UTC; naras
Author:	Balasubramanian Narasimhan [aut, cre], SCIP Optimization Suite Authors [cph] (SCIP, SoPlex, PaPILO libraries)
Maintainer:	Balasubramanian Narasimhan <naras@stanford.edu>
Repository:	CRAN
Date/Publication:	2026-10-04 23:00:09 UTC

scip: Interface to the SCIP Optimization Suite

Description

Provides an R interface to SCIP (Solving Constraint Integer Programs), a framework for mixed-integer programming (MIP), mixed-integer nonlinear programming (MINLP), and constraint integer programming (2025, doi:10.48550/arXiv.2511.18580). Supports linear, quadratic, SOS, indicator, and knapsack constraints with continuous, binary, and integer variables. Includes a one-shot solver interface and a model-building API for incremental problem construction.

Author(s)

Maintainer: Balasubramanian Narasimhan naras@stanford.edu (ORCID)

Authors:

• Balasubramanian Narasimhan naras@stanford.edu (ORCID)

Other contributors:

• SCIP Optimization Suite Authors (SCIP, SoPlex, PaPILO libraries) [copyright holder]

See Also

Useful links:

• https://bnaras.github.io/scip/

• https://scipopt.org/

• Report bugs at https://github.com/bnaras/scip/issues

Convert a matrix to CSC (Compressed Sparse Column) format

Description

Convert a matrix to CSC (Compressed Sparse Column) format

Usage

make_csc_matrix(x)

Arguments

Value

A list with components i (row indices, 0-based), p (column pointers, 0-based), x (values), nrow, ncol

Print method for scip_control objects

Description

Print method for scip_control objects

Usage

## S3 method for class 'scip_control'
print(x, ...)

Arguments

Value

Invisible x.

Add an indicator constraint to a SCIP model

Description

If binvar = 1 then sum(coefs * x[vars]) <= rhs.

Usage

scip_add_indicator_cons(model, binvar, vars, coefs, rhs, name = NULL)

Arguments

Value

Integer; 1-based constraint index.

Add a linear constraint to a SCIP model

Description

Adds lhs <= sum(coefs * x[vars]) <= rhs.

Usage

scip_add_linear_cons(model, vars, coefs, lhs = -Inf, rhs = Inf, name = NULL)

Arguments

Value

Integer; 1-based constraint index.

Add a quadratic constraint to a SCIP model

Description

Adds lhs <= linexpr + quadexpr <= rhs where quadexpr = sum(quadcoefs[k] * x[quadvars1[k]] * x[quadvars2[k]]).

Usage

scip_add_quadratic_cons(
  model,
  linvars = integer(0),
  lincoefs = double(0),
  quadvars1 = integer(0),
  quadvars2 = integer(0),
  quadcoefs = double(0),
  lhs = -Inf,
  rhs = Inf,
  name = NULL
)

Arguments

Value

Integer; 1-based constraint index.

Add a SOS1 constraint to a SCIP model

Description

At most one variable in the set can be nonzero.

Usage

scip_add_sos1_cons(model, vars, weights = NULL, name = NULL)

Arguments

Value

Integer; 1-based constraint index.

Add a SOS2 constraint to a SCIP model

Description

At most two adjacent variables in the set can be nonzero.

Usage

scip_add_sos2_cons(model, vars, weights = NULL, name = NULL)

Arguments

Value

Integer; 1-based constraint index.

Add a variable to a SCIP model

Description

Add a variable to a SCIP model

Usage

scip_add_var(model, obj, lb = 0, ub = Inf, vtype = "C", name = NULL)

Arguments

Value

Integer; 1-based variable index.

Add multiple variables to a SCIP model

Description

Add multiple variables to a SCIP model

Usage

scip_add_vars(model, obj, lb = 0, ub = Inf, vtype = "C", names = NULL)

Arguments

Value

Integer; 1-based index of first variable added.

SCIP solver control parameters

Description

Create a list of control parameters for the SCIP solver. Parameters are organized into logical groups: output, limits, tolerances, presolving, LP, branching, heuristics, separating, and emphasis. Any SCIP parameter can also be set directly using its native path via ....

Usage

scip_control(
  verbose = TRUE,
  verbosity_level = 3L,
  display_freq = 100L,
  time_limit = Inf,
  node_limit = -1L,
  stall_node_limit = -1L,
  sol_limit = -1L,
  best_sol_limit = -1L,
  mem_limit = Inf,
  restart_limit = -1L,
  gap_limit = 0,
  abs_gap_limit = 0,
  feastol = 1e-06,
  dualfeastol = 1e-07,
  epsilon = 1e-09,
  presolving = TRUE,
  presolve_rounds = -1L,
  presolve_emphasis = "default",
  lp_threads = 1L,
  lp_iteration_limit = -1L,
  lp_scaling = TRUE,
  branching_score = "p",
  heuristics_emphasis = "default",
  separating_emphasis = "default",
  emphasis = "default",
  threads = 1L,
  ...
)

Arguments

Details

The four emphasis arguments (emphasis, presolve_emphasis, heuristics_emphasis, separating_emphasis) are SCIP meta-settings: each one adjusts a whole family of native parameters at once, the same way SCIP's interactive shell commands set emphasis, set presolving emphasis, set heuristics emphasis and set separating emphasis do. They are applied before any individual parameter, in that order, so a native parameter given via ... always overrides whatever the emphasis chose for it.

Value

A named list of class "scip_control" with components:

verbose

Logical; verbosity flag.

scip_params

Named list; all parameters as SCIP native paths.

Output

verbose

Logical; print solver output. Default TRUE.

verbosity_level

Integer 0–5; verbosity detail (0 = none, 3 = normal, 5 = full). Default 3. Ignored if verbose = FALSE.

display_freq

Integer; display a status line every this many nodes (-1 = never). Default 100.

Termination Limits

time_limit

Numeric; maximum solving time in seconds. Default Inf (no limit).

node_limit

Integer; maximum number of branch-and-bound nodes. Default -1L (no limit).

stall_node_limit

Integer; nodes without improvement before stopping. Default -1L (no limit).

sol_limit

Integer; stop after finding this many feasible solutions. Default -1L (no limit).

best_sol_limit

Integer; stop after this many improving solutions. Default -1L (no limit).

mem_limit

Numeric; memory limit in MB. Default Inf (no limit).

restart_limit

Integer; maximum restarts. Default -1L (no limit).

Tolerances

gap_limit

Numeric; relative MIP gap tolerance; solver stops when the gap falls below this value. Default 0 (prove optimality).

abs_gap_limit

Numeric; absolute gap between primal and dual bound. Default 0.

feastol

Numeric; feasibility tolerance for LP constraints. Default 1e-6.

dualfeastol

Numeric; dual feasibility tolerance. Default 1e-7.

epsilon

Numeric; absolute values below this are treated as zero. Default 1e-9.

Presolving

presolving

Logical; enable presolving. Default TRUE.

presolve_rounds

Integer; maximum presolving rounds (-1 = unlimited). Default -1L.

presolve_emphasis

Character; presolving emphasis setting: "default", "aggressive", "fast", or "off". Default "default".

LP

lp_threads

Integer; number of threads for LP solver. Default 1L.

lp_iteration_limit

Integer; LP iteration limit per solve (-1 = no limit). Default -1L.

lp_scaling

Logical; enable LP scaling. Default TRUE.

Branching

branching_score

Character; branching score function: "s" (sum), "p" (product), "q" (quotient). Default "p".

Heuristics

heuristics_emphasis

Character; heuristic emphasis setting: "default", "aggressive", "fast", or "off". Default "default".

Separating

separating_emphasis

Character; cutting-plane separation emphasis setting: "default", "aggressive", "fast", or "off". Default "default".

Emphasis

emphasis

Character; global solver emphasis, choosing a predefined parameter profile for a problem class or solving goal. One of "default", "cpsolver" (constraint-programming style search, no LP relaxation), "easycip" (solve easy problems fast), "feasibility" (find a feasible solution fast), "hardlp" (cope with hard LP relaxations), "optimality" (prove optimality fast), "counter" (solution counting), "phasefeas", "phaseimprove", "phaseproof" (the three phases of SCIP's three-phase approach), "numerics" (increased numerical safety), or "benchmark" (do not try to avoid the memory limit). Default "default".

Parallel

threads

Integer; number of threads for concurrent solving. Default 1L. Note: concurrent solving may require a SCIP build compiled with parallel support (e.g., PARASCIP=true); not all installations provide this.

See Also

scip_solve, scip_set_param

Examples

## Quick solve with 60-second time limit
ctrl <- scip_control(time_limit = 60)

## Quiet solve with 1% gap tolerance
ctrl <- scip_control(verbose = FALSE, gap_limit = 0.01)

## Aggressive heuristics, no presolving
ctrl <- scip_control(heuristics_emphasis = "aggressive", presolving = FALSE)

## Aggressive presolving and separation, as for hard combinatorial MIPs
ctrl <- scip_control(presolve_emphasis = "aggressive",
                     separating_emphasis = "aggressive")

## Global profile: find a feasible solution quickly
ctrl <- scip_control(emphasis = "feasibility")

## Pass a native SCIP parameter directly
ctrl <- scip_control("conflict/enable" = FALSE, "separating/maxrounds" = 5L)

Get solver information

Description

Get solver information

Usage

scip_get_info(model)

Arguments

Value

A list with solve_time, nodes, iterations, gap, sol_count.

Get number of solutions found

Description

Get number of solutions found

Usage

scip_get_nsols(model)

Arguments

Value

Integer.

Get objective value of best solution

Description

Get objective value of best solution

Usage

scip_get_objval(model)

Arguments

Value

Numeric; objective value, or NA if no solution.

Get the k-th solution from the solution pool

Description

Get the k-th solution from the solution pool

Usage

scip_get_sol(model, k)

Arguments

Value

A list with objval and x.

Get the best solution

Description

Get the best solution

Usage

scip_get_solution(model)

Arguments

Value

A list with objval and x.

Get solver status

Description

Get solver status

Usage

scip_get_status(model)

Arguments

Value

Character; status string (e.g., "optimal", "infeasible").

Create a SCIP model

Description

Creates a new SCIP optimization model for incremental problem construction.

Usage

scip_model(name = "scip_model")

Arguments

Value

An external pointer representing the SCIP model.

Free a SCIP model

Description

Explicitly frees the SCIP model and all associated memory. The model is also freed automatically when garbage collected.

Usage

scip_model_free(model)

Arguments

Value

Invisible NULL.

Solve a SCIP model

Description

Solve a SCIP model

Usage

scip_optimize(model)

Arguments

Value

Invisible NULL. Use scip_get_status and scip_get_solution to retrieve results.

Set objective sense

Description

Set objective sense

Usage

scip_set_objective_sense(model, sense = "minimize")

Arguments

Value

Invisible NULL.

Set a SCIP parameter

Description

Set a SCIP parameter

Usage

scip_set_param(model, name, value)

Arguments

Value

Invisible NULL.

Solve a linear or mixed-integer program using SCIP

Description

One-shot interface to the SCIP solver. Formulates and solves:

\min_{x} \; obj' x

subject to constraint rows defined by A, b, sense, with variable types vtype and bounds lb, ub.

Usage

scip_solve(obj, A, b, sense, vtype = "C", lb = 0, ub = Inf, control = list())

Arguments

Value

A named list with components:

status

Character; solver status (e.g., "optimal", "infeasible", "unbounded").

objval

Numeric; optimal objective value (or NA if no solution).

x

Numeric vector; primal solution (or NULL if no solution).

sol_count

Integer; number of solutions found.

gap

Numeric; relative optimality gap.

info

List with additional solver information (solve_time, iterations, nodes).