Skip to content

Cohesion analyzer reference

The cohesion analyzer computes LCOM-HS (Henderson-Sellers) per file and a variable sharing ratio per function to identify files and modules with low internal cohesion.

For the conceptual model — what counts as a variable, what LCOM-HS measures and how to read the scores — see Understanding cohesion metrics. To run it, see the Quick start.

codefang run -a static/cohesion --format text .
codefang run -a static/cohesion --format json --per-file .

Configuration options

The cohesion analyzer uses the UAST directly and has no analyzer-specific configuration options.

Option Type Default Description
(none) -- -- Uses UAST; no analyzer-specific config

The run-level walk flags apply (--languages, --include-generated, --include-vendored, --per-file).

Metrics and thresholds

Metric Meaning
Total Functions Every function found, same-named ones included; methods are reported as Type.name
LCOM Score Function-weighted mean of the per-file LCOM-HS values (0 = cohesive, 1 = not)
Cohesion Score 1 - LCOM Score; the section score (6/10 = 0.60)
Avg Cohesion Mean per-function cohesion (shared variables / distinct variables)

Metrics render with two decimals. Per-function cohesion is bucketed as:

Bucket Range Issue severity
Excellent >= 0.6 not listed
Good [0.4, 0.6) good
Fair [0.3, 0.4) fair
Poor < 0.3 poor

The distribution always sums to Total Functions. The issues list holds every function below Excellent, ordered by cohesion ascending, then name; text shows the first five (-v for all).

The summary message follows the Cohesion Score: Excellent >= 0.7, Good >= 0.4, Fair >= 0.3, otherwise Poor.

No data

Files without functions (struct-only files, empty files, Markdown) are excluded from every number. When no analyzed file has a function — including an existing empty directory — the section reports No cohesion data available, Total Functions 0, score 0 (0/10), no distribution and an empty issues list, and the command exits 0. Unreadable or unparseable files are skipped and counted in the progress: skipped N … line on stderr.

Example output

The fixture is two Go files — share.go with f using {x, y}, g using {x} and h using {x, y, z} (LCOM ⅓), and disjoint.go with f2 using {p} and g2 using {q} (LCOM 0.5) — plus a struct-only types.go.

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ COHESION                                                         Score: 6/10 ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

  Summary: Good overall cohesion with room for improvement

  Key Metrics
  ────────────────────────────────────────────────────────────────────────────
  Total Functions     5           LCOM Score          0.40
  Cohesion Score      0.60        Avg Cohesion        0.53

  Distribution
  ────────────────────────────────────────────────────────────────────────────
  Excellent (>=0.6)  ████████████████████████░░░░░░░░░░░░░░░░  60%  (3)
  Good (0.4-0.6)     ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░   0%  (0)
  Fair (0.3-0.4)     ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░   0%  (0)
  Poor (<0.3)        ████████████████░░░░░░░░░░░░░░░░░░░░░░░░  40%  (2)

  Top Issues
  ────────────────────────────────────────────────────────────────────────────
  f2                        disjoint.go                         0.00
  g2                        disjoint.go                         0.00

--format compact renders the one-line form:

COHESION     [██████░░░░] 6/10  Good overall cohesion with room for improvement

--format json emits the section report (text and compact render the same section). With --per-file, a files array sits between issues and score: one entry per analyzed file with that file's own metrics, distribution and issues; a function-free file carries "status": "No functions found".

{
  "overall_score_label": "6/10",
  "sections": [
    {
      "title": "COHESION",
      "score_label": "6/10",
      "status": "Good overall cohesion with room for improvement",
      "metrics": [
        { "label": "Total Functions", "value": "5" },
        { "label": "LCOM Score", "value": "0.40" },
        { "label": "Cohesion Score", "value": "0.60" },
        { "label": "Avg Cohesion", "value": "0.53" }
      ],
      "distribution": [
        { "label": "Excellent (>=0.6)", "percent": 0.6, "count": 3 },
        { "label": "Good (0.4-0.6)", "percent": 0, "count": 0 },
        { "label": "Fair (0.3-0.4)", "percent": 0, "count": 0 },
        { "label": "Poor (<0.3)", "percent": 0.4, "count": 2 }
      ],
      "issues": [
        { "name": "f2", "location": "disjoint.go", "value": "0.00", "severity": "poor" },
        { "name": "g2", "location": "disjoint.go", "value": "0.00", "severity": "poor" }
      ],
      "score": 0.6
    }
  ],
  "overall_score": 0.6
}

--format yaml and --format bin (a CFB1 envelope over the compact JSON of the same value) emit the computed metrics: every function with its quality level (cohesion ascending), the bucket counts (only buckets that occur), the functions below Good with a risk level (HIGH below Fair, MEDIUM otherwise), and the aggregate.

function_cohesion:
    - name: f2
      source_file: disjoint.go
      cohesion: 0
      quality_level: Poor
    - name: g2
      source_file: disjoint.go
      cohesion: 0
      quality_level: Poor
    - name: h
      source_file: share.go
      cohesion: 0.6666666666666666
      quality_level: Excellent
    - name: f
      source_file: share.go
      cohesion: 1
      quality_level: Excellent
    - name: g
      source_file: share.go
      cohesion: 1
      quality_level: Excellent
distribution:
    excellent: 3
    poor: 2
low_cohesion_functions:
    - name: f2
      source_file: disjoint.go
      cohesion: 0
      risk_level: HIGH
      recommendation: Consider splitting into multiple focused functions
    - name: g2
      source_file: disjoint.go
      cohesion: 0
      risk_level: HIGH
      recommendation: Consider splitting into multiple focused functions
aggregate:
    total_functions: 5
    lcom: 0.4
    lcom_variant: LCOM-HS (Henderson-Sellers)
    cohesion_score: 0.6
    function_cohesion: 0.5333333333333333
    health_score: 60
    message: Good overall cohesion with room for improvement

The HTML plot output includes two charts:

  1. Function Cohesion Scores — A bar chart showing per-function cohesion scores, sorted worst-first. Color-coded: green (≥0.6), yellow (0.4–0.6), orange (0.3–0.4), red (<0.3). Includes reference lines at threshold boundaries.

  2. Cohesion Distribution — A pie chart showing the distribution of functions across quality categories (Excellent, Good, Fair, Poor).

The raw report behind the plots lists every function with its start_line, line_count (the function's line span), variable_count (distinct variables), cohesion, the three assessments and _source_file.

See also