Präzi: From Package-based to Call-based Dependency Networks

The replication package contains:

• Scripts for generating call-based dependency networks
• Scraper for mining toolchain & build status on Docs.rs
• Analysis scripts for producing results of the paper
• Links to datasets

Generating CDNs using rust-callgraphs constructed call graphs

There are two options for generating a CDN:

1. Static CDN: a one-time generation that uses the resolved dependency versions available in the call graphs.
2. Dynamic CDN: on-the-fly generation based on user-provided timestamps. Constructs API-mappings based on entry and exit points of packages.

Generate a one-time static CDN

# Annontate and prune (i.e., remove std calls) call graphs 
./ufify/run.sh

# Create a CDN and PDN 
./gen/run.sh

Preperation for dynamic CDNs

./api-pair-extract/run.sh

Working with Call-based Dependency Networks

Installation Prerequisites

Dependencies

• Rust toolchain
• Python 3
• pandas
• numpy
• networkx

Python bindings to Rust's semver library

1. Generate an FFI-compatible C-ABI library in Rust

cd bindings
cargo build --release

2. Replace lookup path to the library

After compilation, a generated library (libsemver_ffi.so) should be available under the target folder:

target/release/libsemver_ffi.so

Replace the path in evolysis-rustcg.py to match your absolute path.

On-the-fly generation and analysis

Running the evolysis-rustcg.py will automatically create 3 PDNs (index, docs.rs, RustPräzi) and 1 CDN (RustPräzi). After generation, you can run analysis such as num_of_dependency_fns(praezi_fn_closure, "praezi") and result will be dumped to out/ folder.

python3 -i analysis/evolysis-rustcg.py <timestamp>

Example

python3 -i analysis/evolysis-rustcg.py 2015-08

Analysis on a Static CDN

The Jupyter Notebook CDN Analysis.ipynb provide examples of how to load a CDN and perform descriptive statistics

Datasets

The call graph corpus and a statically generated CDN is available at Zenodo for download.