NEWS

growkar 0.99.4

The introductory vignette no longer guards its plotting chunks with a flag defined in an earlier chunk. R CMD build also tangles each vignette, and during tangling the chunk code is not evaluated, so eval = has_graphics could not be resolved and reported an error for every guarded chunk. The ggplot2 availability check is now written inline in the chunk options, which resolves both when the vignette is woven and when it is tangled.

Changes made in response to the second round of Bioconductor package review (issue #4227).

Added BiocBaseUtils to Imports and replaced the package's own helpers with its equivalents.
Removed the internal growkar_require_suggested(). The plot_*() functions and select_palette() now check their optional graphics dependencies with BiocBaseUtils::checkInstalled().
Removed the internal growkar_numeric_or_na(). Scalar arguments are now validated with BiocBaseUtils::isScalarNumber(), BiocBaseUtils::isScalarCharacter(), and BiocBaseUtils::isTRUEorFALSE() in as_tidy_growth_data(), detect_exponential_phase(), compute_growth_rate(), summarize_growth_metrics(), validate_growth_data(), validate_growth_experiment(), and the store_metadata arguments of growth_metrics(), phase_windows(), and fit_growth_models().
validate_growth_data() no longer coerces min_points_per_sample with as.integer(); a non-scalar or non-numeric value is now an error.

select_palette() no longer assembles its palettes from RColorBrewer. The eight qualitative palettes it offers are shipped by base R and are returned by [grDevices::palette.colors()] with identical colour values, so RColorBrewer has been dropped from Suggests altogether. The plot_*() functions now check only for ggplot2, and palette_name still accepts the same names ("Dark2", "Set1", and so on).
Each analysis result is now written to metadata() under exactly one key. fit_growth_models() previously stored its fits and parameters twice (as model_fits/growth_model_fits and model_parameters/ growth_model_parameters), and every analysis duplicated its settings into both analysis_params and its own *_parameters entry. The model_fits, model_parameters, and analysis_params keys have been removed; the documented growth_* and exponential_phase_* keys are unchanged.
validate_growth_experiment() now calls methods::validObject() instead of repeating the checks already performed by the GrowthExperiment validity method. It checks only the finiteness requirement that the class itself does not impose.
Replaced the superseded purrr::map_dfr() with purrr::list_rbind() at all six call sites.

SummarizedExperiment input must now carry numeric time values in rowData(se)$time. The previous fallback, which silently coerced assay row names (or the tidyomics .feature labels) to numeric, has been removed; the error message shows how to set the column explicitly. Coercion of character values is now confined to as_tidy_growth_data(), the import adapter for vendor plate-reader exports, where non-numeric entries are expected and dropped rows are reported with a warning.

Removed the internal growkar_tidy_from_summarized_experiment(). Conversion of a GrowthExperiment to long form is now a short adapter over tidySummarizedExperiment's own as_tibble() output, reading the .feature and .sample labels directly.
as_tidy_growth_data() recognises .sample and .feature as column aliases, so tibbles obtained from a SummarizedExperiment with the tidyomics verbs can be passed back to growkar unchanged.
Removed the loadNamespace() call on tidySummarizedExperiment, which is an Imports dependency. Its namespace, and therefore its as_tibble() method, is now loaded through a regular importFrom() directive. The vignette no longer guards the tidyomics section behind an availability check.

Changes made in response to the Bioconductor package review (issue #4227).

Added the S4 class GrowthExperiment, which extends SummarizedExperiment with a validity method enforcing the canonical growth-assay layout. It is now the only data container in the package.
Removed the growkar_data S3 class and its constructor as_growkar(), along with the setOldClass()/setAs() pair used to coerce it. It duplicated the container constructor and was the source of the S3/S4 tension noted in review.
Replaced as_summarized_experiment() with the S4 constructor GrowthExperiment() and standard coercion as(x, "GrowthExperiment"). The constructor carries a metrics argument, preserving the only capability that was unique to as_growkar(). as(x, "SummarizedExperiment") is available in the other direction.
Converted the growkar_fit S3 class to the formal S4 class GrowthFit, with a validity method, show(), summary(), coef(), fitted(), residuals(), and nobs() methods, and the new accessors fit_sample(), fit_model(), fit_status(), fit_converged(), and fit_data().
extract_params() and augment_growth_fit() are now S4 generics with methods for GrowthFit.
The package now defines no S3 classes and registers no S3 methods.

Moved ggplot2 and RColorBrewer from Imports to Suggests. The data-structure and analysis layers install and run without a graphics stack; the plot_*() functions check for the suggested packages at call time and raise an informative error when they are unavailable.

Added tidySummarizedExperiment to Suggests and documented tidy manipulation and display of growkar objects through the tidyomics stack rather than reimplementing those transformations.
Repositioned as_tidy_growth_data() explicitly as an import adapter for vendor plate-reader exports, which tidyomics does not cover.
Fixed a malformed code chunk in the introductory vignette.

Refactored the package around SummarizedExperiment as the canonical data model for microbial growth phenotyping workflows.
Added SE-native accessor and analysis helpers, including growth_assay(), timepoints(), sample_data(), growth_metrics(), phase_windows(), and fit_growth_models().
Standardized metadata conventions for derived summaries, exponential-phase windows, model fits, and analysis parameters stored in metadata(se).
Updated core analysis and plotting functions so tidy and wide inputs are standardized into the canonical SummarizedExperiment representation before downstream analysis.
Added plot_growth_curve_facets() for averaged, sample-family faceted growth curve visualization in multi-sample datasets.
Reworked the README, vignette, examples, and package metadata to present growkar as a Bioconductor-native package for high-throughput microbial growth phenotyping.