cleanup

[FBartos]

Reorganized the R backend into smaller, purpose-specific files and consolidated duplicated helpers without changing analysis behavior or output.

[juliuspfadt]

build builder bot build for glory

[Copilot]

Copilot wasn't able to review this pull request because it exceeds the maximum number of lines (20,000). Try reducing the number of changed lines and requesting a review from Copilot again.

[juliuspfadt]

I was going through the jasp files and found this. no idea if that is because of your PR:
error.jasp.zip

[juliuspfadt]

See above. I am a bit lost as to how to review this, I checked the .jasp files and did some minor clicking around. I would still have some Ai review this, other than that it seems fine.

[juliuspfadt]

Review summary

Verified the "no behavior change" claim both empirically (ran the test suite, root-caused the one failure) and structurally (diffed every one of the ~800 top-level functions' old body vs. new body across all 8 file families, via parallel review).

Test results: agentTestAll() — 289 pass, 2 fail, 10 skip. Both failures are in test-other-Meta-Regression - dat.bangertdrowns2004.R (analysis 4, heterogeneity-coefficient table) and are not caused by this PR — confirmed by loading the pre-refactor code (base commit eda15c5) directly and rerunning the same test, which produces the identical "new" numbers rather than the stale hardcoded expectations. That table already carries skip_on_os("linux") # numerical precision issues; it's a near-singular scale-model fit (SE ≈ 71–88k) sensitive to environment/BLAS drift, unrelated to this refactor.

Behavior-preservation audit: All 8 file families (bayesianmetaanalysisdeprecated, classicalmetaanalysis, classicalmetaanalysismultilevelmultivariate, forestplot, robustbayesianmetaanalysis, funnelplot/effectsizecomputation/selectionmodels, predictionperformance+misc) checked out clean — no dropped functions, no duplicated definitions anywhere in the repo, no style regressions. The genuine consolidations (.maShowMetaforRCode gaining a makeCallText param, .mammVarianceCovarianceMatrixDetails(), the new metaanalysis-export-helpers.R/metaanalysis-r-code-helpers.R) are all correctly wired to their call sites.

Minor, non-blocking nits:

• R/classicalmetaanalysis-plots.R:56-68 and R/classicalmetaanalysis-model-summary-tables.R:307-321 — leftover runs of 15-16 blank lines from cutting functions out of the monolith. Cosmetic only.
• The PR description says "consolidated duplicated helpers," but for prediction-performance specifically the one real duplicate (the metamisc::valmeta() arg-construction block in classical vs. Bayesian fit functions) wasn't actually merged — still copy-pasted in both files post-move. Not a bug, just a missed opportunity / slightly imprecise description.
• Pre-existing, unrelated to this PR: library(metaSEM) in R/metaanalyticsem.R:25 violates the repo's no-library() convention — worth a future cleanup.

Recommendation: Approve. Well-executed, verified behavior-preserving refactor. Suggest one final clean agentTestAll() run before merge (a vdiffr snapshot blip showed up in one isolated single-analysis run but didn't reproduce in the full suite — likely test-run-order noise, not a real regression).

[FBartos]

Yeah, some of the meta-sem visualizations do not work (not a new issue) and shall be fixed in a later separate PR.

The goal of this PR was just rearrange the files because they became too long which is inefficient for AI maintenance.