Testing suite overhaul#98
Open
grantdadams wants to merge 20 commits into
Open
Conversation
Addresses the testing-suite recommendations in discussion #75. Structure & robustness: - Flatten the themed tests-*/ subfolders into a flat tests/testthat/ with test-<theme>-* names (testthat only auto-discovers the top level). Delete the fragile source()-based test-subdir-folders.R runner. - Fix the skip policy: pure-R unit tests run everywhere; the fit_mod() integration tests carry a file-top skip_on_cran() so R CMD check stays CRAN-fast while the coverage job (NOT_CRAN=true) runs the full suite. This stops ~53 files being silently skipped in CI. - Fix the latent expect_all_true(): it was called 52x but never defined. Add it as a custom expectation in helpers.R and de-qualify the calls. Docs: - Add tests/testthat/README.md: theme -> source map, run/skip/snapshot/ coverage instructions, and the FIMS-style test-file template. - Document the multi-species operating-model simulator and the calc_*_nll reference helpers. Coverage & CI: - Add covr to Suggests; add a TMB-patched test-coverage workflow (system deps + NOT_CRAN=true so the full suite counts) uploading to Codecov. - Add the dev branch to the R-CMD-check triggers; add coverage/check badges. Verified: full suite 3744 pass / 0 fail (NOT_CRAN=true); R CMD check tests OK with 0 errors. Coverage baseline ~52%.
- test-plot-smoke.R: fit a small single- and multi-species model once and assert every exported plotter (plot_biomass/ssb/recruitment/depletion/ index/catch/F/selectivity/mortality/data/comp, the plot() S3 method, and the predation plotters) runs without error, drawing to a throwaway device. This is the regression net for the upcoming base->ggplot rewrite. - plot_mortality: replace deprecated ggplot2 size= with linewidth= in element_rect()/geom_contour()/geom_line() (ggplot2 >= 3.4.0). Verified: 21 pass / 0 fail / 0 warn (NOT_CRAN=true).
First batch of the plotting overhaul (base graphics -> ggplot2, standardized + colourblind-safe). - R/7-plot_helpers.R: shared infrastructure for all plotters -- standard theme_classic-based theme, viridis colour/fill scales, model-list coercion, ggsave helper, and the unified "return the ggplot object" contract (so plots are composable and their data is testable). - plot_timeseries (engine for plot_biomass/ssb/recruitment/depletion/ ssb_depletion/exploitable_biomass): replace the base-graphics multi-panel renderer with a faceted ggplot built from a tidy data frame. The exact quantity/CI extraction is unchanged, so plotted values are identical; the function now returns the ggplot. Drop the vestigial .save_par() call. - tests/testthat/test-plot-accuracy.R: assert each plotter's $data equals the same quantity from as.data.frame.Rceattle() (the canonical tidy output), aligned on the plotted years. Verified: plot accuracy + smoke tests pass (NOT_CRAN=true); plotted biomass matches source/1e6 exactly.
Batch 2 of the plotting overhaul. - plot_index / plot_catch: rewrite the base-graphics survey-index and fishery-catch fit plots as faceted ggplots (observed points + lognormal 95% bars, predicted line per model). Shared assembly (.fleet_fit_df) and renderer (.render_fleet_fit) remove the duplicated extraction/plot loops. plot_catch keeps MSE-projection ribbons. Both return the ggplot. - plot_index gains a `log` argument; plot_logindex is now a deprecated thin forwarder to plot_index(log = TRUE) (drops ~220 duplicated lines). - plot_indexresidual: rewrite the "pearson" branch as a ggplot of log(predicted/observed) by year (faceted by fleet); the "osa" branch still routes to osa_residuals()/plot.rceattle_osa(). - Old base-graphics-only arguments (line_col/right_adj/top_adj/single.plots/ alpha/lwd/ymax) are retained as ignored args for back-compatibility. - Tests: accuracy (plotted Predicted == quantities$index_hat / catch_hat) and smoke (incl. plot_index(log) and the plot_logindex deprecation). Verified: plot accuracy + smoke tests pass (NOT_CRAN=true); index/catch predictions match source exactly.
- plot_maturity: faceted ggplot of maturity-at-age (line per model), theme + viridis, returns the ggplot. - Bug fix: the function read data_list$pmature, which does not exist (the field is data_list$maturity), so it errored on real fits. Now reads the correct field. It had no test coverage, which is why this went unnoticed. - Tests: smoke + accuracy (plotted Maturity == data_list$maturity). Part of batch 3 of the plotting overhaul; the persp-based selectivity surfaces follow.
Batch 3 of the plotting overhaul. Selectivity/M age x year surfaces are now drawn as year-coloured at-age curves (faceted by fleet) instead of 3D persp -- the "multi-line by year" convention from r4ss's SSplotSelex, and far more legible/accessible. - plot_selectivity: year-coloured selectivity-at-age curves, faceted by fleet, sex as linetype. Returns the ggplot. - plot_selectivity_vs_maturity: terminal-year fishery selectivity vs input maturity-at-age, faceted by species (SPR-debugging view). - Tests: smoke + accuracy (terminal selectivity == sel_at_age). Conventions cross-checked against r4ss / WHAM / SAM.
Batch 4 (part 1) of the plotting overhaul. - plot_mortality: default to year-coloured M-at-age curves faceted by species (type = "lines"), with type = "heatmap" for the age x year tile. Drops the base-graphics persp/contour branches; returns the ggplot. M2 = TRUE plots M1 + M2, FALSE plots M1; `log` and `maxage` retained. - Tests: accuracy (plotted M == quantities$M2_at_age).
plot_f was a near-duplicate of plot_timeseries for F_spp. It is now a thin wrapper: plot_timeseries(output = "F_spp") for the line + CI machinery, plus Ftarget/Flimit reference lines per species. Removes ~225 duplicated lines and gives F the same consistent 95% CI band convention. Returns the ggplot. Tests: F accuracy cross-checked against as.data.frame()$F_spp.
Completes batch 4 of the plotting overhaul. - plot_m_at_age: total M (M1 + M2) at a fixed age as a time series, faceted by species (line per model, sex as linetype). Returns the ggplot. - plot_m2_at_age_prop: proportion of predation mortality (M2) on each prey age attributable to each predator, coloured by predator and faceted by prey species/sex. The M2_prop extraction is unchanged. Returns the ggplot. - Tests: smoke (multispecies) + accuracy (plot_m_at_age == M_at_age).
- plot_b_eaten: total biomass eaten as prey per species as a faceted time series (MSE ribbons when applicable). Returns the ggplot. - plot_b_eaten_prop: biomass eaten of each prey by predator, coloured by predator and faceted by prey species. Returns the ggplot. - plot_ration: population consumption (ration x biomass-at-age, summed over age) as a faceted time series. Explicit sex indexing fixes the original base-graphics drop bug (rowSums on a 1-sex slice). Returns the ggplot. - plot_diet_comp (already ggplot): now returns the per predator/prey/year cowplot grids invisibly. Verified it plots observed (Stomach_proportion_by_weight) vs fitted (diet_hat[,2]) diet proportions, paired row-for-row via residuals(source = "diet"). - Tests: smoke (multispecies) + accuracy (plot_b_eaten == sum B_eaten_as_prey).
…gplot Final batch of the plotting overhaul. - plot_stock_recruit: SSB vs recruitment points + fitted SRR curve (mean / Beverton-Holt / Ricker), faceted by species, with a 95% normal data ellipse (add_ci). Returns the ggplot. - plot_data: r4ss-style data-availability timeline rebuilt as a faceted ggplot (year x fleet, faceted by data type, point area = relative quantity). Accepts a fit or a raw data_list. Returns the ggplot. - plot_form: functional-response forms drawn with geom_line (1-D forms) or geom_raster (2-D predator x prey), returning the ggplot. (Needs a multispecies params object, as before.) - Tests: smoke (plot_data, plot_stock_recruit) + accuracy (plot_stock_recruit SSB/R == quantities$ssb/R).
devtools::document(): update man pages whose roxygen changed (plot_catch, plot_index, plot_indexresidual, plot_logindex, plot_mortality, plot_selectivity_vs_maturity, plot_stock_recruit) and add man pages for the new internal plotting helpers (.as_model_list, .model_labels, .rceattle_theme, .rceattle_scale, .save_ggplot). NAMESPACE unchanged.
plot_logindex was a thin scale variant of plot_index. Now that plot_index takes log = TRUE, remove plot_logindex entirely rather than keeping a deprecated shim: drop the function + its export/man page, update the diagnostics and model-options vignettes, and drop the smoke-test case. Note: this removes an exported function from a released package; callers should switch to plot_index(..., log = TRUE).
The base-graphics index/catch fits used gplots::plotCI for error bars; the ggplot rewrite uses geom_errorbar, so gplots is no longer used. Removing it from Imports clears the "declared Imports should be used" R CMD check NOTE.
- Version 4.6.0 -> 4.7.0. - NEWS.md: new 4.7.0 section covering the ggplot plotting overhaul (return ggplot objects, viridis/theme_classic, plot_index(log=), SR ellipses, the test-suite/coverage work), the breaking changes (plot_logindex removed, plotters return ggplot, gplots dropped), and the bug fixes. - model-options vignette: note plotters are ggplot2-based and return the ggplot object (composable, optionally saved via file=).
Add *.gcda / *.gcno (generated by covr::package_coverage()) to .gitignore so the coverage workflow's instrumentation output is never accidentally committed. (Also picks up the .mcp.json ignore already pending in the working tree.)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Moved all plotting functions to standardized ggplot2 that plots, optionally saves, and returns ggplot2 object
Standardized testing following @Bai-Li-NOAA recommendations
This was a mix of Claude and my edits