How dblalock/bolt works
Step 2. Identify files and enumerate functions.
Step 4. Use proper visual cues.
Step 5. Tools and automation.
• For C++, Doxygen (with “EXTRACT_ALL” enabled and “CALL_GRAPH” set to YES in the configuration) along with Graphviz will emit call graphs showing which functions call which.
• For Python, specialized tools (such as pycallgraph) can help generate similar function–level call graphs.
• Combining these tools with manual annotation ensures you capture both automated dependency information and any design decisions that you want to emphasize.
A sample (simplified) textual sketch would be:
┌────────────────────────────────────────────┐
│ bolt_api.py (Python wrapper) │
│-------------------------------------------- │
│ + __init__() │
│ + fit() │
│ + transform() ────► calls bolt_native() │
│ + knn() ────► calls transform() │
└────────────────────────────────────────────┘
│ ▲
│ │ (SWIG)
▼ │
┌─────────────────────────────┐ ┌─────────────────────────────┐
│ bolt_native.cpp (C++) │─────►│ quantize/mithral.cpp │
│-----------------------------│ │-----------------------------│
│ BoltEncoder::BoltEncoder() │ │ + encode() │
│ BoltEncoder::fit() │ │ + transform_block() │
│ BoltEncoder::nativeCall() │ │ ... │
└─────────────────────────────┘ └─────────────────────────────┘
• Of functions within each file