Tuning for Large Repos

Large repositories (50k+ files, deep dependency trees, multiple language ecosystems) can take longer to scan. Here’s how to speed things up.

Use parallel workers

The single biggest speedup for large repos:

nfr-review report ./my-repo --workers 8

Workers parallelise collector execution across threads. Start with the number of CPU cores and adjust based on I/O vs. CPU balance.

Exclude test directories

By default, test directories are excluded from analysis. If you’ve disabled this, re-enable it:

nfr-review report ./my-repo --exclude-tests

This skips scanning tests/, test/, __tests__/, spec/, and fixture directories, significantly reducing file I/O for repos with large test suites.

Skip unnecessary collectors

If you know certain collectors aren’t relevant, skip them:

collectors:
  skip:
    - graphify        # graph analysis is expensive for large repos
    - jacoco-report   # skip if no JaCoCo reports
    - gatling         # skip if no Gatling reports

Each skipped collector saves one pass over the file tree.

Use tech overrides

Auto-detection scans the full repo tree for technology markers. If you know your stack, declare it explicitly to skip detection:

tech:
  java: true
  spring_boot: true
  kubernetes: true
  python: false       # skip Python analysis
  go: false           # skip Go analysis

Container mode in CI

The container execution mode avoids pip dependency resolution on every run:

- uses: JimAKennedy/nfr-review@v1
  with:
    execution: "container"
    path: .
    workers: 4

Image layers are cached, so subsequent runs start faster.

Selective scanning

For very large monorepos, scan only the directories that changed:

# In CI, identify changed paths
CHANGED=$(git diff --name-only origin/main...HEAD | cut -d/ -f1-2 | sort -u)

# Scan each changed directory
for dir in $CHANGED; do
  if [ -d "$dir" ]; then
    nfr-review run "$dir" --output-dir results/
  fi
done

Timeouts

If a scan is taking too long in CI, set a step timeout:

- name: Run NFR Review
  timeout-minutes: 15
  uses: JimAKennedy/nfr-review@v1
  with:
    path: .
    workers: 4