Impact & Analysis

NestWeaver’s analysis commands help you understand the consequences of changes before you make them. The impact command traces the blast radius of a symbol through the dependency graph with confidence-weighted scoring, while pr-impact analyzes an entire PR’s changed files and assigns a risk level. These are structural analyses based on the graph — no LLM required.

Command reference

Command	Description	Key Flags
impact	Trace blast radius of a symbol	--depth, --confidence, --repo, --json
pr-impact	PR blast radius with risk scoring	--files, --depth, --json
affected-tests	Select tests for changed files	--files, --base-ref, --json
dead-code	Detect unreachable symbols	--min-confidence, --json
hubs	Find most-connected hub nodes	--top, --json, --config
bridges	Find architectural chokepoints	--top, --json, --config
contracts	Inspect API contract graph	(subcommand)
ranking	Inspect ranking priors	explain, rank subcommands
eval	Offline retrieval-quality evaluation	run, compare subcommands
export	Export graph to external format	--format, --output, --top
clusters	Detect community clusters	--resolution, --json

Blast radius analysis

# Check what depends on a symbol
nestweaver impact "processPayment" --depth 5

# Filter to a specific repo in a multi-repo setup
nestweaver impact "processPayment" --repo payments-api --json

# Set minimum edge confidence threshold
nestweaver impact "processPayment" --confidence 0.8

impact traverses incoming CALLS, IMPORTS, EXTENDS, and IMPLEMENTS edges from the target symbol. Each affected symbol gets an impact_score that decays multiplicatively through edges — low-confidence paths are pruned automatically. The default depth is 3.

PR impact analysis

# Auto-detect changed files from git diff
nestweaver pr-impact

# Specify files explicitly
nestweaver pr-impact --files src/auth.rs,src/db.rs

# Deeper analysis
nestweaver pr-impact --depth 5 --json

pr-impact maps changed files to their symbols, runs transitive impact analysis, groups results by cluster, and assigns a risk level: Low, Medium, High, or Critical. When no --files are given, it uses git diff --name-only to detect changes automatically.

Test selection

# Select tests affected by specific file changes
nestweaver affected-tests --files src/auth.rs,src/db.rs

# Diff against a base branch
nestweaver affected-tests --base-ref main --json

affected-tests performs static, call-graph-based regression test selection. It maps changed files to symbols, reverse-traverses CALLS/IMPORTS to depth 3, and buckets dependent test files into priority tiers. This is a prioritized signal, not a provably safe subset — it misses reflection, DI, codegen, and data-driven tests.

Dead code detection

# Find potentially dead code
nestweaver dead-code

# Only high-confidence results
nestweaver dead-code --min-confidence high --json

dead-code walks forward from every entry point following CALLS, IMPORTS, EXTENDS, IMPLEMENTS, and MEMBER_OF edges. Symbols not reached are reported as potentially dead, with confidence levels: low, medium, high. Entry points are determined from manifest files and visibility modifiers.

Graph topology

# Find hub nodes (highest degree centrality + PageRank)
nestweaver hubs --top 20

# Find bridge/chokepoint nodes (highest betweenness centrality)
nestweaver bridges --top 20

# Detect community clusters (Leiden algorithm)
nestweaver clusters --resolution 0.5 --json

Hubs are central abstractions many parts of the codebase depend on. Bridges are architectural chokepoints — many shortest paths pass through them, so changing a bridge has outsized blast radius. Clusters use Leiden community detection; results are cached in a sidecar file.

Graph export

# Export as Cypher (Neo4j)
nestweaver export --format cypher

# Export as GraphML (Gephi/yEd)
nestweaver export --format graphml --output graph.xml

# Export as Mermaid flowchart (top 30 symbols by PageRank)
nestweaver export --format mermaid --top 30

# Export as MessagePack (for WASM mode)
nestweaver export --format msgpack --output graph.msgpack

Supported formats: Cypher (Neo4j), GraphML (Gephi/yEd), Mermaid, MessagePack.