Crate srcmetrics 

Language-independent source code metrics.

srcmetrics parses source code with tree-sitter, converts it to a small language-independent intermediate representation (ir), and computes metrics from that representation only (metrics). Every metric is reported separately under its id (metrics::specs); none are combined into a single readability score. What each metric means is documented at https://lunainsidious.github.io/srcmetrics/metrics/.

Supported languages: C, C++, Go, Java, JavaScript (JSX), Python, Rust, TypeScript (TSX).

Analyze source text

let result = srcmetrics::analyze::analyze_source(
    "example.py",
    "def greet(name):\n    if name:\n        return 'Hello, ' + name\n    return 'Hello'\n",
)?;
let srcmetrics::result::FileResult::Ok { functions, .. } = &result.files[0] else { unreachable!() };
assert_eq!(functions[0].name.as_deref(), Some("greet"));
assert_eq!(functions[0].metrics.metrics["complexity.cyclomatic"], Some(2.0));

Analyze a directory

analyze::analyze walks a directory (respecting .gitignore), analyzes every file with a supported extension and records run metadata (commit, parser and metric definition versions). Files that fail to parse are kept in the result with status: "error" instead of being dropped.

let result = srcmetrics::analyze::analyze(std::path::Path::new("src"), None)?;
println!("{}", serde_json::to_string_pretty(&result).unwrap());

A metric that cannot be computed (e.g. an average over zero functions) is null in result::MetricsOutput::metrics, with the reason in result::MetricsOutput::unavailable; it is never reported as 0.

Modules

analyze

Analysis pipeline: path -> IR -> metrics -> AnalysisResult (ADR-0010).

csv

CSV output (ADR-0016): one row per project / file / function.

error

Analysis errors (ADR-0005, design principle P7).

ir

Common Intermediate Representation (ADR-0003).

lang

Language Adapters: source code -> Common IR (ADR-0004).

metrics

Metric Engine. Calculators read only the Common IR (design principles P2, P4).

model

Experimental readability model: ridge regression fitted on user-provided labels (ADR-0019). There are no built-in weights: a model exists only when trained on labels (design principle P5).

report

Self-contained HTML report (ADR-0020): no scripts, no external resources.

result

Analysis result: the serialized output format (ADR-0010, design principle P10).

stats

Statistical analysis of analysis results (ADR-0018).