M4 — pls4all source split plan

Source: Explore-agent analysis of cpp/src/core/model.cpp (4541 lines, 180 KB) + cpp/src/core/extra_pls.cpp (1647 lines, 80 KB).

Totals

File	Lines	Functions	Helpers (anon namespace)
model.cpp	4541	18	~1500 lines (L18–L1524)
extra_pls.cpp	1647	19	~290 lines (L18–L306)
Total	6188	37	~1790

Per-file function table

model.cpp

#	Function	Line	Length	Target
1	fit_pls_regression_nipals	1528	291	models/pls/nipals.cpp
2	fit_pls_canonical_nipals	1819	10	models/pls/canonical.cpp
3	fit_pls_canonical_svd	1829	10	models/pls/canonical.cpp
4	fit_pls_svd	1839	161	models/pls/plssvd.cpp
5	fit_opls_nipals	2000	350	models/opls/opls.cpp
6	fit_pls_regression_orthogonal_scores	2350	265	models/pls/oscores.cpp
7	fit_pls_regression_kernel	2615	280	models/pls/kernel.cpp
8	fit_pls_regression_svd	2895	166	models/pls/svd.cpp
9	fit_pls_regression_power	3061	166	models/pls/power.cpp
10	fit_pls_regression_randomized_svd	3227	170	models/pls/randomized_svd.cpp
11	fit_pcr_svd	3397	288	models/pcr/pcr.cpp
12	fit_pls_regression_simpls	3685	200	models/pls/simpls.cpp
13	predict_into	3885	36	models/core/dispatch.cpp
14	fit_di_pls	3921	242	models/classification/pls_da.cpp
15	fit_pls_sparse_simpls	4163	225	models/sparse/sparse_simpls.cpp
16	fit_model	4388	46	models/core/dispatch.cpp
17	transform_into	4434	36	models/core/dispatch.cpp
18	model_array	4470	72	models/core/dispatch.cpp

extra_pls.cpp

#	Function	Line	Length	Target
1	fit_sparse_pls_da	307	60	models/sparse/sparse_pls_da.cpp
2	fit_group_sparse_pls	367	78	models/sparse/group_sparse.cpp
3	fit_fused_sparse_pls	445	36	models/sparse/fused_sparse.cpp
4	fit_cppls	481	70	models/specialized/cppls.cpp
5	fit_weighted_pls	551	66	models/specialized/weighted.cpp
6	fit_robust_pls	617	61	models/specialized/robust_huber.cpp
7	fit_ridge_pls	678	59	models/specialized/ridge.cpp
8	fit_continuum_regression	737	63	models/specialized/continuum.cpp
9	fit_pls_glm	800	47	models/glm/pls_glm.cpp
10	fit_pls_qda	847	98	models/classification/pls_qda.cpp
11	fit_pls_cox	945	85	models/classification/pls_cox.cpp
12	fit_pds	1030	111	transfer/pds.cpp
13	fit_ds	1141	102	transfer/ds.cpp
14	fit_mir_pls	1243	90	models/local/mir_pls.cpp
15	fit_missing_aware_nipals	1333	65	models/specialized/missing_aware_nipals.cpp
16	approximate_press	1398	79	models/core/dispatch.cpp
17	fit_bagging_pls	1477	54	models/ensembles/bagging.cpp
18	fit_boosting_pls	1531	52	models/ensembles/boosting.cpp
19	fit_random_subspace_pls	1583	65	models/ensembles/random_subspace.cpp

High-risk cross-cutting helpers

The ~1790 lines of helpers (anonymous namespaces) MUST be extracted FIRST, into shared _common.cpp files, before any algorithm extraction:

Helper	Location	Used by	Extract to
simple_simpls	extra_pls.cpp:97	model.cpp fit_pls_sparse_simpls AND extra_pls sparse variants	models/pls/_common.cpp
center_scale_in_place	model.cpp:161	~15 model.cpp functions	models/core/_common.cpp
dominant_svd_pair, largest_symmetric_eigenvector	model.cpp helpers	SVD / Kernel / Canonical / OPLS	models/pls/_common.cpp
canonicalize_direction_sign, canonicalize_svd_pair_sign	model.cpp:274,691,772,919,953,1064,1086,1175,3801	SVD variants + OPLS chassis	models/pls/_common.cpp
soft_threshold	extra_pls.cpp:87	all 4 sparse variants	models/sparse/_common.cpp
copy_matrix, column_means, subtract_means, squared_norm	extra_pls.cpp helpers	sparse + specialized	models/sparse/_common.cpp
invert_square, dot, matrix_vector_product	model.cpp helpers	most algorithms	models/core/_common.cpp
validate_fit_request, copy_matrix_checked	model.cpp helpers	all fit_* functions	models/core/_common.cpp

Proposed extraction order (7 phases)

Phase	Scope	Commits
1	Helpers → _common.cpp (3 files)	3
2	Core PLS solvers (nipals, simpls, svd, plssvd, power, randomized_svd, kernel, canonical)	8
3	PCR + OPLS	2
4	Classification (pls_da, pls_qda, pls_cox)	3
5	Sparse (4 files)	4
6	Specialized (6) + ensembles (3) + transfer (2) + local + glm	12
7	Core dispatch + header consolidation	2

Total: ~34 commits if each extraction is its own commit (best for blame preservation). Per-commit gate: n4m_tests 265/265 still green and no function moves are functional changes.

Notes from the agent’s analysis

• Lines 18–1524 of model.cpp form a tight ~1500-line anonymous-namespace helper block with a clear boundary; lines 18–306 of extra_pls.cpp form a ~290-line one. Both extract cleanly.

• No explicit // === ALGORITHM === section dividers exist; functions appear in topological order by dependency.

• The 18 model.cpp functions reach line 4541; the 19 extra_pls.cpp functions end at line 1647. Both files have no unaccounted top-level definitions.

• Helpers do not cross between the two source files (each set is file-local). The simple_simpls cross-call from model.cpp → extra_pls.cpp via fit_pls_sparse_simpls is the ONE exception — extracting it first to models/pls/_common.cpp resolves it cleanly.