acdeveloper-sci/forti4d

Purpose and Entry Points

Forti4D is a Python static-analysis toolkit for mixed F77/F90/F95 source trees, designed around regex and logical-line processing rather than a full Fortran parser. The CLI/package orchestration entry point is forti4d/pipeline.py; runtime paths and environment overrides are centralized in forti4d/config.py.

Pipeline Orchestration

forti4d/pipeline.py runs a dependency-ordered 19-stage batch pipeline, supports selecting/resuming stages, and materializes reports under the configured output directory. Stages communicate primarily through CSV/text artifacts rather than an in-memory shared model. Logging is initialized through forti4d/lib/logging_setup.py; parallel execution support is in forti4d/lib/parallel.py.

Source Normalization and Parsing Boundary

The core parsing boundary is the custom logical-line reader in forti4d/lib/reader_logical.py, which normalizes F77 fixed-form and F90 free-form continuation behavior. Regex grammars in forti4d/lib/patterns_v1.py and forti4d/lib/patterns_v2.py, plus statement classification in forti4d/lib/kinds.py, replace external parser dependencies. This scope intentionally supports F77–F95 but not Fortran 2003+ or C-preprocessor expansion.

Corpus Inventory and Dependency Graph

Inventory establishes program-unit identities, types, and line ranges in forti4d/analyzers/inventory.py. Dependency extraction then resolves CALL/USE/function relationships, ambiguities, external references, includes, and file-level relationships in forti4d/analyzers/dependencies.py. These outputs form the graph and unit identity foundation used by structural and reachability analyses.

Line-Level Audit and Scope Attribution

forti4d/analyzers/profiler.py produces per-file audit CSVs that classify logical lines and density metrics. Multiple analyzers consume this audit layer: forti4d/analyzers/complexity.py, forti4d/analyzers/common_blocks.py, and scope-management analyzers. Attribution relies on inventory line ranges, assigning statements to the innermost containing program unit.

Structural, Risk, and Modernization Analysis

Topology and migration interpretation are layered over inventory, dependency, and audit outputs: forti4d/analyzers/block_analysis.py models block topology; forti4d/analyzers/structure_analysis.py identifies graph roles such as hubs and entry points; forti4d/analyzers/cross_analysis.py derives migration strategy. Entry points include explicit PROGRAM units and implicit main programs.

Symbol and Legacy-Feature Scope Layer

The E4 ScopeManager layer extracts declarations and legacy semantics within unit boundaries: forti4d/analyzers/symbols.py, forti4d/analyzers/derived_types.py, and forti4d/analyzers/equivalences.py. It captures variables, signatures, IMPLICIT rules, derived types, and EQUIVALENCE alias groups; it complements graph-level analysis rather than resolving modern OOP or procedure-pointer semantics.

Consolidation and Deliverables

Late stages derive reachability/dead-code status, SLOC, clone relationships, graph render inputs, and ranked migration risk through forti4d/analyzers/reachability.py, forti4d/analyzers/consolidate.py, and forti4d/analyzers/prioritization.py. forti4d/analyzers/html_report.py creates a self-contained report, while forti4d/analyzers/visual_graph.py emits Graphviz DOT; Graphviz is optional and only needed to render those files. The only stated Python third-party runtime dependency is Loguru.