Skip to content

Latest commit

 

History

History
46 lines (34 loc) · 2.51 KB

README.md

File metadata and controls

46 lines (34 loc) · 2.51 KB
Raw

TSInfer and TSEvolve

TSInfer is a tool for automatically inferring TypeScript declaration files for JavaScript libraries.

TSEvolve reports differences between two versions of a library, by comparing the JavaScript implementations of the two versions.

Installation

The tools have been tested on 64 bit Windows and Linux, but should also work on Mac OS X.

  • Make sure node.js, Chrome, jdk 8, and Intellij IDEA are installed.
  • Download and unpack this repository
  • Run the below two commands in the unpacked repository
    • npm install
    • node node_modules/typescript/bin/tsc.js --module commonjs node_modules/ts-type-reader/src/*.ts

Usage

The tools do not have a command line interface, instead they are executed by opening the repository as an IntelliJ project, and writing the source code that does the intended task.

The file src\dk\webbies\tscreate\main\Main.java is the main starting point to start using TSInfer and TSEvolve, Main.java includes comments describing how to run TSInfer and TSEvolve

Experiments

In the main package alongside Main.java is also a couple of classes that run experiments, these are: PerformanceTest, PrecisionTest, TSCheckMissesCounter and UsefullnessTest. They are configured as an IntelliJ run configuration, and can thus be executed by selecting and running the appropriate run configuration in the top right corner of IntelliJ.

Mapping the tables from the paper to classes in the implementation:
Table 1: PerformanceTest for the left part, and PrecisionTest for the right part.
Table 2: TSCheckMissesCounter
Table 3: UsefulnessTest
Table 4: UsefulnessTest (comment in the part starting with Map<String, Pair<DeclarationType, DeclarationType>> toCompare)
Table 5/6: Manual by running TSEvolve (from Main).

Performance

These tools can be very hungry for RAM, for the biggest benchmarks it is recommended to have at least 16GB of RAM and an SSD, or 32GB of RAM. If your computer doesn't have that amount of RAM, it is recommended to only execute the smaller libraries.

Some benchmarks that should run fine on most computers are: async, Backbone, Hammer, Handlebars, jasmine, Knockout, moment, and underscore.