Refactor DFT+U occupation reader for strict compact input

[lanshuyue]

Reminder

• I have read AGENTS.md and docs/developers_guide/agent_governance.md.
• I have linked an issue or explained why this PR does not need one.
• I have added adequate unit tests and/or case tests, or explained why not.
• I have listed the exact verification commands run and their results.
• I have described user-visible behavior changes, including INPUT parameter changes.
• I have explained core-module impact for ESolver, HSolver, ElecState, Hamilt, Operator, Psi, or other source/ changes.
• I have requested any needed governance exception below.

Linked Issue

No linked issue. This is a focused follow-up to closed PR #8155, addressing the maintainer request to reduce the size of dftu_base_io.cpp and resubmit the reader refactor when ready.

Unit Tests and/or Case Tests for my changes

• Commands run:
  • cmake --build build --target MODULE_PW_dftu_base_test -j4
  • cmake --build build --target module_pwdft MODULE_DFTU_base_io MODULE_DFTU_op -j4
  • cmake --build build --target io_input MODULE_IO_input_test_para -j4
  • OMP_NUM_THREADS=1 ctest --test-dir build --output-on-failure -R '^(MODULE_PW_dftu_base_test|MODULE_DFTU_base_io|MODULE_DFTU_op|MODULE_IO_input_test_para)$'
  • make -f source/Makefile BUILD_DIR=/tmp/abacus_make_verify CXX=mpiicpx ... OPENMP=ON -j2
  • git diff --check upstream/develop...HEAD
  • python3 tools/03_code_analysis/agent_governance_check.py --base upstream/develop --head HEAD --format text
  • python3 tools/03_code_analysis/code_quality_score.py --format json source/source_pw/module_pwdft/dftu_base_io.cpp source/source_pw/module_pwdft/dftu_base_io_read.cpp
• Result summary:
  • All requested CMake build targets passed.
  • The focused CTest targets passed.
  • The legacy Makefile compiled dftu_base_io_read.cpp and linked ABACUS.mpi successfully.
  • git diff --check passed.
  • Before this PR, dftu_base_io.cpp scored 0 (raw score -52) and was 1345 lines, while read_occup_m had cyclomatic complexity 40. After extraction, dftu_base_io.cpp is 996 lines and scores 6; dftu_base_io_read.cpp is below 500 lines and scores 87. The IO/reader average is 46.5. The remaining reader deductions include complexity findings for read_occup_atom_header and read_occup_block_label.
  • Governance check has no blocking findings. Its global-reference warnings are migration-neutral: 2 references moved to the new reader file and 2 were removed from the old file (net 0). The documentation warning is addressed by updating the init_occ_mat INPUT help metadata; generated documentation is not edited directly.
• Checks not run, with reason:
  • Full ABACUS integration suite was not run. The change is limited to DFT+U occupation-matrix restart parsing and has focused PW, LCAO, parser-error, and INPUT-metadata unit coverage.

What's changed?

• Extract read_occup_m and its parser helpers into dftu_base_io_read.cpp, keeping the public DFT+U IO declarations together in dftu_base_io.h.
• Accept only the current compact occupation-matrix format written by occ_mat.txt, with both full-line and inline # comments supported; legacy Atom= / L= / spin= input is rejected with a filename-and-line diagnostic.
• Replace the unsafe fixed char word[20] token buffer and ignore(150, '\n') calls with line-based std::string parsing.
• Require exact compact headers, atom order, DFT+U channels, spin/SOC labels, finite values, matrix dimensions, complete atom coverage, and no trailing non-comment content.
• Parse complete matrix blocks into temporary storage before updating OccupationMatrix.
• Store parsed matrices through OccupationMatrix::set_flat() instead of accessing ModuleBase::matrix::c directly or using nested element-wise set() loops.
• Remove the now-redundant init_occ_mat and init_chg reader arguments; file selection remains the caller's responsibility.
• Remove the legacy dm_onsite.txt fallback and update the init_occ_mat help text to document the compact-only input contract.
• Add coverage for long and inline comments, long tokens, legacy-header rejection, short/long matrix rows, missing SOC blocks, compact PW SOC, and compact LCAO real spin-basis storage.

Known Limitation and Follow-up

• This PR does not change the current init_chg = file requirement for a DFT+U occupation-matrix file. If occ_mat.txt is absent, ABACUS still terminates before read_occup_m() is called.
• Allowing init_chg = file to continue without an occupation-matrix file requires a separate implementation that can initialize or reconstruct the DFT+U occupation matrix safely from the available calculation state.
• The missing-file error must remain in place until that fallback initialization is implemented and tested. A follow-up change should remove the hard requirement only together with that functionality; it is intentionally out of scope for this compact-reader refactor.

Governance Notes

• INPUT/docs changes: No INPUT parameter or default changes. The init_occ_mat help metadata now documents the compact-only file format and current filename search order. Generated YAML/Markdown documentation is intentionally not committed.
• Core module impact: DFT+U restart I/O only. Both PW Pauli storage and LCAO real spin-basis storage remain supported; only legacy text-file compatibility is removed. No ESolver/HSolver public interface changes.
• Global dependency rationale: GlobalV::MY_RANK and GlobalV::ofs_running were moved with read_occup_m; PR-level added=2, removed=2, net=0.
• Exceptions requested: none.

[mohanchen]

@ABACUS-PR-BOT Could you review the PR?

[ABACUS-PR-BOT]

Reviewed 8e86f9304d1eddf35fc922252913f782bc283f52 statically with an independent reviewer. No new implementation defect was identified; the inline comments are optional test improvements.

Non-blocking: Please refresh the quality-check entry in the PR description. The command names dftu_base_io_read.cpp, but the extracted file is dftu_base_read.cpp; the checked-in scorer reports 87/100 with complexity entries for read_occup_atom_header/read_occup_block_label, rather than 93/100 with none (the IO/reader average is 46.5). Correcting the filename and recorded results is sufficient.

Fresh whitespace and trusted governance checks returned exit 0, with explained migration/documentation warnings. No builds or runtime tests were run; CI diagnostics were unavailable.

[ABACUS-PR-BOT]

Non-blocking: Please add a focused multi-atom reader case, including an uncorrelated species, plus missing/duplicate/out-of-order atom variants. The reader now enforces complete ordered coverage, but its unit fixtures all contain one Fe atom, so these paths lack a regression check. A follow-up after merge is reasonable.

[ABACUS-PR-BOT]

Non-blocking: This fixture ends after spin 1, so removing the extra-column check would still make EXPECT_EXIT pass when the reader reaches the missing spin-2 block. Please make the rest of the file valid and change only the row under test; the short-row fixture has the same ambiguity. This can be a follow-up to make these cases catch the row-validation regressions they name.

[lanshuyue]

Thanks, fixed in 1b16b3b. Both the short- and long-row fixtures now contain a complete valid spin-2 block, so the row under test is the only malformed part of each file. The focused tests and the related DFT+U test targets pass.