Context Footprint

A static analysis tool for measuring architectural context exposure in codebases.

Context Footprint is a research prototype that computes a Context Footprint (CF) metric for functions and types in a codebase.
CF approximates the amount of code context that must be traversed to analyze a given symbol, based on language-level dependencies and abstraction boundaries.

This tool is developed alongside an ongoing academic study and is intended for measurement, comparison, and empirical analysis, rather than production use.

---

Status

⚠️ Research Preview

• The CF definition and traversal rules are still evolving.
• APIs, outputs, and heuristics may change before publication.
• The repository currently serves reproducibility and early feedback purposes.

---

What the Tool Does

Given a semantic index of a codebase, the tool can:

• Compute the distribution of CF values across all functions or types
• Identify symbols with unusually large context exposure
• Query the CF of a specific symbol
• Print the source code that contributes to a symbol’s CF

CF is computed via conservative graph traversal over language-level dependencies, with configurable pruning rules.

---

Supported Languages

The tool consumes semantic data (JSON) produced by language-specific extractors (e.g. LSP-based), and is therefore language-agnostic in principle.

Tested languages include:

• Python
• TypeScript

Support for additional languages depends on the availability of semantic data extractors that output the SemanticData JSON format.

---

Installation

Option 1: uv / pip (recommended)

Install as a Python tool—includes the cf-extract command for Python project extraction:

uv tool install cftool
# or: pip install cftool

Requires Python 3.9+.

Option 2: Cargo

Build from source:

git clone https://github.com/yourusername/context-footprint.git
cd context-footprint
cargo build --release

Requires Rust 1.70+.

Prerequisites

• A semantic data JSON file for the target project (e.g. from cf-extract for Python)

---

Basic Usage

1. Generate semantic data

For Python projects, use the bundled extractor:

cf-extract /path/to/python/project > semantic_data.json

Or use another extractor (e.g. LSP-based) that outputs the SemanticData JSON format.

2. Analyze CF distribution

cftool semantic_data.json stats
# or with cargo build: ./target/release/cftool semantic_data.json stats

3. Find symbols with highest CF

cftool semantic_data.json top --limit 10

4. Query a specific symbol

cftool semantic_data.json compute "<symbol-id>"

5. Inspect contributing context

cftool semantic_data.json context "<symbol-id>"

---

Output

The tool reports CF values as token counts, using a configurable size function. Output includes percentile distributions and summary statistics for large codebases.

Example:

Functions - Context Footprint Distribution:
  Count: 856
  Median: 245 tokens
  90th percentile: 20,567 tokens

---

How CF Is Computed (Brief)

1. A directed dependency graph is constructed from the semantic data (JSON).

2. Starting from a target symbol, dependencies are traversed conservatively.

3. Traversal stops at:

   • External libraries
   • Explicit abstraction boundaries defined by the pruning policy

4. The size of the reachable subgraph is summed.

The default pruning policy is intentionally conservative and favors soundness over precision.

For a formal definition, see docs/design.md.

---

Project Structure

The implementation separates core analysis logic from language-specific adapters:

src/
├─ domain/        # Graph model and traversal logic
└─ adapters/      # Size functions, doc scoring, test detection

---

License

Apache 2.0

---

Acknowledgements

Semantic data is consumed as JSON (e.g. from LSP-based extractors).