package asak

  1. Overview
  2. Docs
Partitioning OCaml codes

Install

Dune Dependency

Authors

Maintainers

Sources

0.3.tar.gz
md5=a1ada3488624203e71dec6806e02fdab
sha512=b34bf3cb5471b7ec2fb32682481a3fb253a6fc80cbbb9c310756d94320524921ea70594b590c0e3d0b4c80321243a634014b84aed1be98065bbac5f9ec3c0d27

Description

Asak provides functions to parse, type-check and identify similar OCaml codes. These functions are then used to partition codes implementing the same function and help to detect code duplication.

Published: 25 Jul 2021

README

asak

Asak is an OCaml library that allows to identify similar OCaml codes.

Why ?

For teaching

The module Asak.Partition offers a function create that produces a partition of codes implementing the same function, where two codes are in the same class if they are syntactically "close".

For redundancy detection

The binary anzad can detect redundant definitions of an OCaml project built with dune and compare it with a database of previously analyzed projects.

To use it on a project with sources in src/, run:

dune build @check
anzad src/

Documentation

The documentation of the API is available here: https://nobrakal.github.io/asak/asak/.

A man page is available for the binary anzad.

How ?

The idea is to compare AST (Abstract Syntax Tree) of codes. However, the OCaml AST is too rich for our purpose (since, for example, match x with ... and function ... generate two different AST). We decided instead to use the Lambda language, an intermediate language in the OCaml compilation pipeline, where such syntactic sugar is optimized away.

To efficiently compare Lambda trees, we use the methodology of Chilowicz et al. which consist in hashing recursively trees.

We then compare hashes and provide a clustering of the closest functions.

But really, how ?

There is two cores:

  • Asak.Lambda_hash, that defines a function hash_lambda : config -> threshold -> Lambda.lambda -> hash which is hashing a Lambda expression capturing the shape of the AST, with respect to the given configuration. We recommend to first apply the functions from Asak.Lambda_normalization to normalize the Lambda expression.

  • Asak.Clustering, that defines a function cluster : ('a * Lambda_hash.hash) list -> ('a list) wtree list which is making a kind of complete-linkage clustering of a list of hashes. The output is a dendrogram where leaves are close in a tree if they are similar. It is guaranteed that two codes in the same tree share at least a sub-AST.

More details

A paper (in french) about asak was published in the proceedings of the JFLA (Journées Francophones des Langages Applicatifs) 2020. It can be found here: https://github.com/nobrakal/asak-paper/

The name

This tool is about making partitions. "Partition" is the word in french for "sheet music". Consequently, its name is about music: asak is the name of Tuareg's traditional songs accompanied by a monochord violin.

This monochord violin is called an anzad, which is the name of the binary client of asak.

License and copyright

Asak is released under the MIT license. The copyright is held by IRIF / OCaml Software Foundation.

Authors

Asak is developed by Alexandre Moine.

Dependencies (4)

  1. cppo build & >= "1.6.0"
  2. dune > "1.5"
  3. cmdliner >= "1.0.0"
  4. ocaml >= "4.05" & < "4.13"

Dev Dependencies (2)

  1. alcotest with-test & >= "0.8.0"
  2. odoc with-doc & >= "1.2.0"

Used by (2)

  1. learn-ocaml >= "0.12" & < "1.0.0"
  2. learn-ocaml-client < "1.0.0"

Conflicts

None

OCaml

Innovation. Community. Security.