The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

vrpr 0.2.2

• Fixes the installation failure in CRAN’s additional clang23 checks (LLVM 23 libc++, which no longer provides <algorithm> transitively): the vendored Client.cpp and VehicleType.cpp use std::any_of without including it. tools/vendor.R now adds the include, as it already did for <iterator>.

vrpr 0.2.1

• Fixes the installation failure on the CRAN macOS builders that use Apple’s MacOSX11.3 SDK (r-release-macos-x86_64, r-oldrel-macos-arm64 and r-oldrel-macos-x86_64): a lambda in the vendored SearchSpace.cpp captured a structured binding, which older Apple clang rejects (allowed only from C++20). tools/vendor.R now copies the binding into a plain reference before the lambda.

vrpr 0.2.0

Upgrades the vendored solver core to PyVRP 0.14.0 (from 0.13.4), a major upstream release that reworks the data model and the search engine.

New features

• Pickup and delivery (shipments): add_shipments() adds paired pickup/delivery visits (same vehicle, pickup first, same trip), enabling the PDP/PDPTW family of problems. Solutions report num_shipments and num_missing_shipments.
• New unplanned() accessor: a tidy view of the optional clients and shipments left out of the solution (prize collecting).
• routes() now also reports activity ("client", "pickup" or "delivery"), shipment and trip columns; client rows keep the previous columns, so pure client instances read as before.

Upstream changes tracked

• The search engine follows PyVRP 0.14’s design: a single operator family (Relocate/Swap plus explicit optional-client/shipment/group operators) replaces the old node/route operator split; the granular neighbourhood is now computed by upstream C++ code (default num_neighbours = 50).
• The penalty manager follows PyVRP 0.14: midpoint initial penalties, violation-magnitude registration, update every 500 solutions and penalty_decrease = 0.90. The init_load/init_tw/init_dist arguments of ils_params() were removed accordingly.
• Internally, routes are sequences of typed activities; reload depots (multi trip) appear as depot activities inside the route.

Breaking changes

• ils_params() lost init_load, init_tw and init_dist (see above); num_neighbours now defaults to 50 (PyVRP’s default).
• routes() gains columns; code that relied on the exact column set should select explicitly.

vrpr 0.1.1

Portability fixes for CRAN builders; no user-facing changes.

• Fixed compilation with LLVM 23’s libc++ (CRAN’s clang23 additional checks), which no longer provides std::back_inserter transitively: the vendored search/LocalSearch.cpp and Solution.cpp now include <iterator> explicitly.
• Fixed compilation on the CRAN macOS builders that use Apple’s MacOSX11.3 SDK (r-release-macos-x86_64 and r-oldrel-macos), whose libc++ <concepts> lacks std::convertible_to: a small compatibility shim (vrpr_compat.h) supplies the concept where the standard library does not.
• Route::Iterator in the vendored sources now declares all five iterator member typedefs, as required by pre-C++20 std::iterator_traits on the same older libc++ (verified by compiling every translation unit against the MacOSX11.3 SDK’s libc++ headers).
• These fixups are applied by tools/vendor.R on top of the verbatim PyVRP sources, so they survive re-vendoring.

vrpr 0.1.0

First release.

• A tidy, pipe-friendly R interface to the ‘PyVRP’ vehicle routing solver. PyVRP’s high-performance C++ core is vendored and rewired through ‘cpp11’, with no ‘Python’ runtime dependency.
• Supports the capacitated VRP, time windows (VRPTW), heterogeneous fleets, multiple depots (MDVRP), prize-collecting (optional clients and mutually exclusive client groups), simultaneous pickup and delivery / backhaul, and multi-trip routes.
• read_vrplib() and read_solomon() read standard VRPLIB/TSPLIB and Solomon instances; plot() draws solutions with ‘ggplot2’.
• The solver is a faithful port of PyVRP’s iterated local search; objective and solution-quality parity with PyVRP is verified in tools/benchmark/.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.