Kotlin Inference Study Analysis & Replication Package

This replication package contains all data and scripts needed to reproduce results from the paper.

TL;DR: Download data-cache.zip and run make to regenerate tables/figures.

Raw data is very large, however, cached Parquet files are smaller. You are likely able to skip downloading data.zip and can instead use data-cache.zip. This should be sufficient to run make.

Required Packages

The scripts require Python 3, and were tested on 3.9. Earlier versions likely work.

See the requirements.txt file for a list of required packages. These may be installed by running:

pip install -r requirements.txt

If you need to download Boa jobs, the Boa API does require authentication. You may, however, view the jobs online: they are public.

File Organization

Files are organized in the following layout:

• data data files are kept here. This includes unprocessed boa output (.txt), and processed versions (.parquet, .txt)
• boa contains Boa queries and snippets used. A description of the output and the snippets is available in CODEBOOK.md
• analyses contains code used to produce figures, tables and other analyses. This is organized in a common directory which contains common utilities, and directories for both java and kotlin.
• figures and tables contains generated tables/figures.

Building figures/tables

The following make targets are available:

• kotlin/table-counts or java/table-counts: These produce Tables Ia and Ib, respectively.
• kotlin/rq-usage-2 or java/rq-usage-2: These produce Figures 4a and 4b, respectively
• kotlin/rq-usage or java/rq-usage: These produce Figures 5 and 6, respectively
• kotlin/rq-mutability: This produces Figure 8.
• kotlin/rq-rhs-types or java/rq-rhs-types: These produce Figures 7a and 7b, respectively.
• kotlin/rq-rhs-methodcalls or java/rq-rhs-methodcalls: These produce Tables IIa and IIb, respectively
• kotlin/rq-survival or java/rq-survival: These produce Figures 9a and 9b, respectively