Changelog

Source: NEWS.md

speed 0.0.11

Major Changes

• Added the linked_cols argument to speed(), naming columns that should be rearranged along with the swap column, for example a variety_name label belonging to a numeric variety code. For hierarchical designs, pass a named list to link different columns at different levels. (#105)
• speed() now stops as soon as a design reaches the lowest score its layout allows, applicable only to the default objective_function(). This can be turned off per level with optim_params(stop_at_optimal = FALSE). summary() now reports the lower bound score alongside the achieved one.

Minor Changes

• The startup version check now only runs in interactive sessions. Set the SPEED_NO_VERSION_CHECK environment variable to disable it entirely.
• summary() now reports why each level stopped - the optimum was reached, no further improvement was found, no swap was possible, or the iteration cap was hit. The reason is also recorded as stop_reason in each level’s metadata.
• speed() now warns when a swap_all = TRUE group holds no two treatments of equal replication, and stops a level immediately when no group in it can be swapped.

Bug Fixes

• speed() no longer fails with “supplied seed is not a valid integer” when seed is left unset in a session that has not yet used the random number generator and random_initialisation is toggled on.
• optim_params(random_initialisation = ) no longer moves plots holding no treatment.
• optim_params(random_initialisation = ) now respects swap_all = TRUE, so it no longer breaks up the wholeplots or strips of a split-plot or strip-plot design.
• speed() now gives a clear error when grid_factors is malformed, instead of failing with “missing value where TRUE/FALSE needed”.
• A two sided formula passed to spatial_factors is now rejected instead of silently accepted, and hierarchical designs now run the same spatial_factors, iterations and seed checks as simple ones.
• Named lists for hierarchical arguments such as iterations are now split per level under the optimise argument, as they already were when swap was a named list. A level the list leaves out now takes that argument’s own default rather than the default for spatial_factors.
• Columns that take no part in the optimisation are no longer converted to factors and back, so a class that cannot be rebuilt with as.<class>(), such as Date, is now returned unchanged instead of as character. (#122)
• Each level of a hierarchical design now starts from the best design found so far, rather than whichever design the previous level’s search last accepted. A level could otherwise be optimised, and reported on, against a design that was never returned, or replace an earlier level’s better result with a worse one.
• A design that ties the best score found now replaces it, so a search that ends on a score plateau returns a random arrangement from that plateau rather than the systematic input it was given. Early stopping still counts only strict improvements.
• The swapped treatments passed to an objective function are now labels rather than factor codes, so objective_function_piepho() updates the plots that actually moved.

speed 0.0.10

Major Changes

• Added a summary() method for "design" objects, reporting structure and replication, a decomposed optimisation score, and design-quality diagnostics. (#73)
• grid_factors gains an optional by element naming the column that separates a design into several grids, e.g. list(dim1 = "row", dim2 = "col", by = "site") for a multi-environment trial. Each grid is scored on its own.

Minor Changes

• Designs whose row/col columns are not numeric, or where two plots share a coordinate, now fail with a message naming the problem.

Bug Fixes

• Design metrics are now built from each plot’s row/col coordinates rather than the order of the rows in the data frame. Designs generated with objective_function_piepho() should be regenerated.
• Multi-site designs are no longer scored as one pooled grid, which discarded plots whose coordinates collided and counted adjacencies between sites. Use grid_factors$by to name the grouping column.
• objective_function_piepho() now scores evenness of distribution per grid and reports each grid separately. A grid with no treatment replicated within it contributes 0 rather than Inf.
• calculate_efficiency_factor() now errors for a design whose treatment contrasts are not estimable, instead of returning an impossible value above 1. The row-column model gained an intercept, which does not change results that were already valid.
• summary() no longer errors on designs that cannot be placed on a single grid; the affected diagnostics report why they are unavailable instead.
• calculate_nb() no longer errors on designs with missing plots when pair_mapping is not supplied.
• calculate_adjacency_score() now recycles a single ring_weights value across every entry of ring_dists, so the default is usable with more than one ring.
• swap_all = TRUE no longer changes the replication of a design when an earlier level has unbalanced a swap group mid-search. Only treatments with matching replication are exchanged.

speed 0.0.9

Major Changes

• Deprecated the splits argument of initialise_design_df() in favor of initialise_split_design_df(). Passing splits now warns with the equivalent suggested call.

Bug Fixes

• speed() now errors when swap_all = TRUE is used on a design with unequal within-group replication, instead of silently swapping treatments with different replication counts.
• speed() no longer returns numeric/integer columns (e.g. treatment, row, col) as their internal factor level codes instead of their original values.
• speed() no longer emits a “Setting row names on a tibble is deprecated” warning when passed a tibble.
• speed() now accepts designs with vctrs-backed multi-class columns (e.g. from the edibble package) instead of erroring; such columns are now returned as character.

speed 0.0.8

Major Changes

• Added ring_dists, ring_weights, and ring_type arguments to calculate_adjacency_score for weighting matches at larger adjacent ring radii; can be passed via speed().

speed 0.0.7

Major Changes

• Added splits argument to initialise_design_df to support split-plot designs (#92).
• Added main_weight and interaction_weight arguments to objective_function_factorial to tune the trade-off between main-treatment and interaction balance (#90).

Minor Changes

• Fixed autoplot.design where 'block' column was required when providing another column for block (#88).

speed 0.0.6

Major Changes

• Extended random_initialise to handle hierarchical (multi-level) optimise lists by shuffling within each level’s grouping.

speed 0.0.5

Major Changes

• Added objective_function_factorial for factorial designs, combining main-treatment and interaction balance scores (#78).

speed 0.0.4

Major Changes

• Added vignettes for MET (#70) and factorial (#71) designs.

speed 0.0.3

Major Changes

• Optimisation parameters were changed from options to arguments to enable better reproducibility of designs (#65
• Enabled one stage MET designs
• Added contributing guide and code of conduct (#59)

See changelog for further details.

speed 0.0.2

Major Changes

• Enabled more complex designs and added some vignettes with examples and detailed use.

See changelog for further details.

speed 0.0.1

First version.