This software is created to analyze the evolution of the Stackage repository, and its analysis is reported in the submitted paper:Stackage Repository: An Exploratory Study of its Evolution

Abstract:

Context. Package repositories for a programming language are increasingly common. A repository can keep a register of the evolution of its packages. In the programming language Haskell, with its defining characteristic monads, we can find the Stackage repository, which is a curated repository for stable Haskell packages in the Hackage repository. Despite the widespread use of Stackage in its industrial target, we are not aware of much empirical research about how this repository has evolved, including the use of monads.

Objective. This paper conducts empirical research about the evolution of Stackage considering monad packages through 22 Long-Term Support releases during the period 2014-2023. Focusing on five research questions, this evolution is analyzed in terms of packages with their dependencies and imports; including the most used monad packages. To the best of our knowledge, this is the first large-scale analysis of the evolution of the Stackage repository regarding packages used and monads.

Method. We define six research questions regarding the repository’s evolution, and analyze them on 51,716 packages (17.05 GB) spread over 22 releases. For each package, we parse its cabal file and source code to extract the data, which is analyzed in terms of dependencies and imports using Pandas scripts.

Results. From the methodology we get different findings. For example, there are packages that depend on other packages whose versions are not available in a particular release of Stackage; opening a potential stability issue. The mtl and transformers are on the top 10 packages most used/imported across releases of the Stackage evolution. We discussed these findings with Stackage maintainers, which allowed us to refine the research questions.

Conclusions. On the one hand, like previous studies, these results may be evidence of how developers use Haskell and give guidelines to Stackage maintainers. One of our proposals is to generate control over the categories and stability that developers assign to their packages. On the other hand, we recommend that Stackage designers take more care when verifying the versions of package dependencies.

Scripts

This repository contains scripts for different purposes:

• Define LTS releases to analyze. In the src folder, you can add the releases to analyze modifying the file lts_list.csv. As an example, you can find all releases up to the release 20-15.

• Download LTS releases. In src/, you should execute the Python script scrapy_lts.py using a Python interpreter (tested with Python 3.9.6). This process can take hours. Note: Please, install all Python libraries required. The downloaded package will be in the folder lts_downloaded/tar_package. For each release, a folder will be created.

• Install Haskell and create parsers. First, install ghcup to install the Glasgow Haskell Compiler. Second, install just to create the required parsers. Then, in src/parse, execute the following commands: just install-libs, just make-parse-cabal, just make-package-info-json, and just make-package-imports.

• Create DataFrames. Execute the script generate_dfs.py in src. The dataframes are stored in the folder data/dfs. For each release, a folder is created. As an example, you can find all dataframes up to the release 20-15.

• Create DataFrames Revised. This is an optional step. If you need to analyze considering the revision of each package, you can execute the same scripts (scrapy_lts.py and generate_dfs.py) using the argument --revised. For the revised versions, scrapy_lts.py will download the packages in the folder lts_download/revised_cabal, and generate_dfs.py will create the dataframes in the folder dfs_revised.

Research Questions and Answers

For the paper, we replied to a set of research questions (RQs):

• RQ1 Which packages are imported the most by other Stackage packages? Do these packages have unstable or incompatible dependencies according to their Stackage release?

• RQ2 What is the average number of (in)direct dependencies per package?

• RQ3 How frequently are packages updated?

• RQ4 RQ4 How have the selected monad packages evolved?

• RQ5 How has the use of the selected monad packages evolved?

• RQ6 How many packages that depend on the mtl and transformers packages are added to and removed from Stackage? How many packages that depended on these monad packages stopped their dependencies?

These RQs with their answers are available and can be tested online in the folder notebooks. The code associated with the research questions is implemented in Python using Jupiter.

Research

This exploratory study is developed by Felipe Ruiz, Nicolas Sepulveda, Paul Leger, Ismael Figueroa, and Nicolás Cardozo at Pragmatics Lab.