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.
complexity.py, sloc.py, and clones.py are the core metrics scripts of Tier 3. complexity.py counts decision points in each unit to compute McCabe cyclomatic complexity, giving a direct measure of how difficult a unit is to test and maintain. sloc.py precisely counts lines of code by classifying every physical line as blank, comment, code, or continuation, then aggregates logical statement counts per unit. clones.py compares same-named units across files using normalized token sequences to identify whether duplicates are exact copies, maintained variants, or fully independent diverged implementations. All three outputs feed into consolidate.py for the final per-unit summary.
complexity.py — McCabe Cyclomatic Complexity
complexity.py is step 8 of the pipeline. It requires the audit/ directory produced by profiler.py.
<FORT_OUT>/audit/<filename>_DEBUG.csv for each source file<FORT_OUT>/inventory_report.csv
Output: report_complexity.csv
One row per program unit, sorted by CC descending.
| Column | Description |
|---|
File | Source file name |
Unit | Unit name |
Type | Unit type |
CC | McCabe cyclomatic complexity |
Level | Complexity level: LOW, MEDIUM, HIGH, or CRITICAL |
Start_Line | First line of the unit |
End_Line | Last line of the unit |
Total_Lines | Physical line count of the unit |
CC Calculation
CC = 1 + (number of decision points in the unit)
| Statement Kind | Count |
|---|
IF_CONSTRUCT | +1 (block IF and single-line IF) |
ELSE_STMT | +1 only if ELSE IF / ELSEIF; plain ELSE = 0 |
DO_CONSTRUCT | +1 |
SELECT_CONSTRUCT | 0 (branches counted via CASE) |
CASE_STMT | +1 unless CASE DEFAULT or CLASS DEFAULT |
WHERE_CONSTRUCT | +1 |
FORALL_CONSTRUCT | +1 |
Complexity Scale
| Level | CC Range | Meaning |
|---|
LOW | 1–10 | Simple, easy to test |
MEDIUM | 11–20 | Moderate complexity |
HIGH | 21–50 | High complexity, refactoring recommended |
CRITICAL | > 50 | Very high risk, hard to maintain |
Scope resolution assigns each decision point to the innermost unit whose [Start_Line, End_Line] range contains the statement’s line number. CC is computed from the audit CSVs, not directly from source — profiler.py must be run first.
sloc.py — Source Lines of Code
sloc.py is step 14 of the pipeline.
- Fortran source files in
FORT_SRC
<FORT_OUT>/inventory_report.csv
Output: report_sloc.csv
One row per program unit, sorted by SLOC_net descending.
| Column | Description |
|---|
File | Source file name |
Unit | Unit name |
Type | Unit type |
LOC | Total physical lines in the unit’s line range |
N_Blank | Blank physical lines |
N_Comments | Comment-only physical lines |
N_Continuation | Continuation lines (2nd, 3rd… physical line of a multi-line statement) |
SLOC_physical | LOC - N_Blank - N_Comments (lines with actual code, including continuations) |
SLOC_net | SLOC_physical - N_Continuation (logical statements only) |
Pct_Comment | N_Comments / LOC × 100 |
Line Classifications
| Category | Condition |
|---|
COMMENT | LogicalLine.is_comment = True |
BLANK | Not a comment, and LogicalLine.text is empty or whitespace |
CODE | First physical line of a non-comment, non-blank logical line |
CONTINUATION | 2nd, 3rd, … physical line of a multi-line statement |
SLOC_net equals the number of logical statements in the unit — the most accurate size measure for comparing units, since it is independent of coding style (how many continuation lines are used per statement).
Pct_Comment measures documentation density. Values below 5% on units with more than 50 logical statements indicate poorly documented code.
sloc.py’s LOC column counts physical lines by reading each source file directly. This differs from the Total_Lines value in inventory_report.csv (End_Line - Start_Line + 1), which is a structural range estimate. Both are valid but answer different questions — the discrepancy is expected and not an error.
clones.py — Duplicate Unit Detection
clones.py is step 15 of the pipeline.
<FORT_OUT>/dep_00_ambiguities.csv (from dependencies.py)<FORT_OUT>/inventory_report.csv- Fortran source files in
FORT_SRC
Output: report_clones.csv
One row per pair of same-named units. Groups with N copies produce N×(N-1)/2 rows (e.g. 3 copies → 3 pairs). Rows are sorted DIVERGED first, then SIMILAR, then IDENTICAL.
| Column | Description |
|---|
Unit | Unit name |
Type | Unit type |
File_A | First file |
File_B | Second file |
SLOC_A | Normalized line count of unit in File_A |
SLOC_B | Normalized line count of unit in File_B |
Similarity_Pct | Similarity percentage (0–100) |
Status | IDENTICAL, SIMILAR, or DIVERGED |
Clone Classifications
| Status | Condition |
|---|
IDENTICAL | Similarity ratio = 1.00 (byte-for-byte identical after normalization) |
SIMILAR | Ratio ≥ 0.80 (default threshold) |
DIVERGED | Ratio < 0.80 |
Normalization
Before comparison, each unit’s source is normalized:
- Logical lines outside
[Start_Line, End_Line] are excluded.
- Comment lines and blank lines are removed.
- Each remaining line is uppercased and whitespace-collapsed to a single space.
This makes comparison insensitive to formatting, comment additions, and case conventions while preserving structural differences.
report_clones.csv is always written — even when dep_00_ambiguities.csv is absent or empty (no duplicate unit names in the corpus). In that case the file contains headers only. This is a valid outcome, not an error.
