Forti4D Changelog: Version History and Release Notes

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/acdeveloper-sci/forti4d/llms.txt

Use this file to discover all available pages before exploring further.

All notable changes to forti4d are documented here. The format follows Keep a Changelog. Semantic Versioning is used from v0.7.0 onward; earlier releases (v0.1–v0.6) were tagged incrementally before SemVer was adopted.

---

​

v0.8.0 — 2026-08-01

​

Added

• In-process library API (Bloque 1): import forti4d; forti4d.run_pipeline(...) runs the full pipeline without subprocess — no more shelling out to python -m forti4d.pipeline. All 19 analyzers now follow a consistent analyze_xxx() / write_xxx() / main() contract, returning results in memory in addition to (never instead of) writing their CSV/report files. CLI behavior, flags, and output files are unchanged.
• Opt-in per-file parallelism (Bloque 2): --workers N / FORT_WORKERS parallelizes the three heaviest, fully-independent-per-file steps — inventory, profiler, blocks — via ProcessPoolExecutor. Default is workers=1 (sequential); behavior is unchanged unless explicitly requested. forti4d/lib/parallel.py::pmap() guarantees sequential/parallel output parity by construction — both modes run the exact same code path.
• Real logging via loguru (Bloque 3): all internal print() diagnostics across the 19 analyzers replaced with logger.debug/info/success/warning/error() calls. Every run now writes a full-detail log to <results_dir>/forti4d.log (always DEBUG+, regardless of --quiet); new --log-file PATH and --no-log-file CLI flags override or disable it. Library callers (forti4d.run_pipeline()) stay silent by default — call the new forti4d.configure_logging() to opt in to console/file output. loguru is forti4d’s first third-party dependency.

​

Changed

• --quiet redefined: now a console log-level filter (WARNING+ instead of INFO+) rather than an all-or-nothing toggle. Previously a successful --quiet run discarded all internal diagnostic output with no trace anywhere; now it’s always captured in forti4d.log.
• run_pipeline()’s on_step_end callback signature changed from (name, success, elapsed, error, output) to (name, success, elapsed, error) — the output param (captured stdout) is gone now that step output goes through loguru instead of redirect_stdout. run_pipeline() also drops its quiet= parameter (superseded by configure_logging()).

​

Fixed

• clones.py::build_file_index: indexed Fortran source files by basename (f.name) instead of relative path — silently dropped all but one file when multiple files share the same name in different subdirectories (e.g., main.F in 19 subdirs of FVCOM); changed to f.relative_to(path).
• dependencies.py::load_inventory_enhanced: used File (basename) as the identifier key for inventory entries, causing dep_00_ambiguities.csv and dep_01_master_data.csv to list only the basename when different files in subdirectories share the same name — changed to Relative_Path, with fallback to File for backward compatibility with older inventory files.
• clones.py: pairwise comparison called read_logical_lines() once per file pair, re-parsing the same source file multiple times when it appeared in several ambiguity groups — added a per-run parse cache (_parse_cache) so each file is parsed at most once per pipeline run.
• inventory.py::write_inventory: skipped creating inventory_report.csv entirely when the source directory contained no Fortran units (if rows: guard) — violated the pipeline invariant; now always writes the file, with headers only when the inventory is empty.
• clones.py::analyze_clones: returned None for report_clones when the inventory was empty, causing write_clones to skip writing the output file — violated the pipeline invariant; now returns an empty list so write_clones always writes report_clones.csv (headers only).
• cross_analysis.py::to_float(): crashed on native int/float values for Fan_In/Fan_Out when fed in-memory (non-CSV-string) data from the new library API — added an isinstance(val, str) guard before .strip().
• visual_graph.py: treated 0 (native int, from in-memory data) as falsy but "0" (CSV string) as truthy, silently dropping the Fi=0 label for units with zero Fan-In — changed to explicit != "" comparisons.

​

Documentation

• README.md: Requirements section now lists loguru instead of claiming stdlib-only; Quick Start documents the new --quiet semantics and --no-log-file.

​

Tests

• tests/test_output_manifest.py, tests/test_lib_pipeline.py: new — byte-for-byte regression gate and library-vs-CLI parity check.
• tests/test_parallel_helper.py, tests/test_parallel_parity.py: new — pmap() order/parity guarantees and workers=1 vs workers=2 output parity, scoped and full-pipeline.
• tests/test_logging_setup.py, tests/test_logging_behavior.py: new — raw loguru API assumptions, configure_logging() behavior, and CLI/library end-to-end logging behavior via subprocess.

---

​

v0.7.4 — 2026-07-31

​

Fixed

• dependencies.py: always write all six dep_* output files (dep_01 through dep_06) even when the corpus has no cross-file dependencies — previously the files were omitted for pure library/module corpora, causing downstream failures in structure_analysis, reachability, and clones; also moves dep_06_include_files.csv out of the incorrectly nested file_deps_map block.

​

Documentation

• doc/scripts/dependencies.md: correct constant names in Configuration table (OUT_* → *_OUT); add “Always written” note to all seven dep_* outputs.

​

Tests

• tests/fixtures_lib_only/ + tests/test_empty_outputs.py: mini-corpus of two module-only .f90 files and three tests verifying that dep_00_ambiguities.csv, report_clones.csv, and report_reachability.csv are written empty (headers only) for a pure-library corpus — exercises the code paths fixed in v0.7.2, v0.7.3, and v0.7.4.

---

​

v0.7.3 — 2026-07-31

​

Fixed

• reachability.py: always write report_reachability.csv (headers only if no entry points exist) — previously the file was omitted for pure library or module-only corpora that contain no PROGRAM or IMPLICIT-MAIN units.

​

Documentation

• doc/scripts/reachability.md: note that the output file is always written and that downstream consumers (consolidate.py, cross_analysis.py) handle the empty-file case gracefully.

---

​

v0.7.2 — 2026-07-31

​

Fixed

• dependencies.py: always write dep_00_ambiguities.csv (headers only if no duplicates found) — previously the file was omitted when the corpus had no units with the same name in multiple files, causing clones.py to treat a valid zero-ambiguity run as a missing-file error.
• clones.py: treat absent dep_00_ambiguities.csv as zero ambiguities instead of aborting with an ERROR message; create report_clones.csv empty in that case; fix typo _wrtite_empty_csv → _write_empty_csv.

​

Documentation

• doc/scripts/dependencies.md: note that dep_00_ambiguities.csv is always written, even when empty.
• doc/scripts/clones.md: clarify that absent or empty dep_00_ambiguities.csv is a valid outcome; note that prioritization.py handles the empty case.

---

​

v0.7.1 — 2026-07-21

​

Fixed

• sloc.py: replaced max() with += when accumulating per-file LOC in the console summary — max() returned only the largest unit per file, causing Total LOC to appear smaller than Physical SLOC (arithmetically impossible).
• executive_summary.py: group files by Relative_Path instead of File (basename) — basename collisions across subdirectories caused incorrect file counts and merged LOC for files sharing the same name in different directories.
• pipeline.py: validate FORT_SRC directory before executing steps — provides a clear error message when the source path is missing or invalid.
• inventory.py + 9 analyzers (profiler, sloc, complexity, symbols, derived_types, equivalences, common_blocks, consolidate, dependencies): add Relative_Path column to inventory and thread it through all file-access and debug-CSV naming — fixes “file not found” errors for source files located in subdirectories.
• reader_logical.py: replace extension-only format detection with a content-based heuristic (detect_fortran_format) — correctly handles HPC code that uses .F extension but is written in free-form F90+ syntax.

​

Documentation

• README.md, doc/mi4d.md: document C preprocessor directive limitation (directives not evaluated; conditional branches always captured; macros and #include not resolved).
• doc/scripts/reader_logical.md: document detect_fortran_format() and the two-phase format detection logic.
• doc/scripts/executive_summary.md: update File column description in file_statistics.csv to reflect relative path usage.

---

​

v0.7.0 — 2026-04-02

​

Added

• Complete English documentation: README rewrite (PyPI mindset, CLI usage, Known Limitations, Future Work), CONTRIBUTING.md, doc/architecture.md, doc/interpretation.md, doc/mi4d.md (MI4D conceptual overview).
• doc/scripts/: 22 script reference files — all translated, renamed to English, and updated to reflect current column names and constants.
• uv.lock: lockfile for reproducible installs.
• .gitignore: standard Python/editor exclusions.

​

Changed

• Complete English translation of all source code: CSV column names, output filenames, config constants, print statements, internal identifiers.
• SemVer versioning adopted (prior releases tagged as v0.1–v0.6).
• Development branch renamed: desarrollo → dev.

​

Fixed

• inventory.py: closed_unit.tipo → closed_unit.type (AttributeError blocking the full pipeline).
• profiler.py, sloc.py: FOLDER_CODE → CODE_PATH (ImportError).
• complexity.py, consolidate.py, sloc.py, executive_summary.py: r['Archivo'] → r['File'] in print/report generation (KeyError).

​

Removed

• All Spanish-language .md documentation files (replaced by English versions).

---

​

v0.6 — TBD

​

Added

• Synthetic Fortran fixtures corpus for testing (tests/fixtures/).
• pytest test suite: 67 tests covering the full pipeline.

​

Fixed

• inventory.py: detection of anonymous INTERFACE blocks.
• Two parser bugs exposed by the fixtures corpus.

---

​

v0.5 — TBD

​

Changed

• Full restructure into a proper Python package (forti4d/).
• Source reorganized into lib/ and analyzers/ subdirectories.
• pyproject.toml added; forti4d CLI entry point configured.

---

​

v0.4 — TBD

​

Added

• html_report.py: self-contained HTML report from prioritization data (step 19 — pipeline grows to 19 steps).
• INCLUDE directive tracking in profiler.py and inventory.py.

---

​

v0.3 — TBD

​

Changed

• E4 scope signals (Implicit_None, Has_Equiv) integrated into cross_analysis.py migration strategy assignment and prioritization.py risk scoring.

---

​

v0.2 — TBD

​

Added

• symbols.py (step 10): variable, parameter, and implicit symbol extraction.
• derived_types.py (step 11): derived TYPE definition extraction.
• equivalences.py (step 12): EQUIVALENCE aliasing groups via union-find.
• E4 scope risk signal (Score_E4) in prioritization.py.
• EQUIVALENCE summary columns in consolidate.py.
• Pipeline grows to 18 steps.

---

​

v0.1 — TBD

​

Added

• Initial 15-step static analysis pipeline for Fortran (F77/F90/F95).
• Core scripts: inventory.py, dependencies.py, profiler.py, block_analysis.py, structure_analysis.py, cross_analysis.py, executive_summary.py, complexity.py, common_blocks.py, sloc.py, clones.py, consolidate.py, visual_graph.py, prioritization.py.
• pipeline.py: single-command orchestrator with --from, --only, --skip, --quiet flags.
• config.py: centralized path configuration via FORT_SRC/FORT_OUT environment variables.
• reader_logical.py: F77 fixed-form and F90 free-form continuation handling.