Functional data science
Clone or download
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
analyze no RFrame warnings Oct 21, 2018
datasets Dataloader + CIFAR10 + ImageFolders + benchmarks (#31) Jan 13, 2019
dense-linear-algebra re-add stack.yamls Oct 21, 2018
dh-core-accelerate dep bounds for ghc8.2.2, rm subdir stack.yamls Oct 21, 2018
dh-core fix warnings in DecisionTrees (#36) Jan 14, 2019
.gitignore first Oct 15, 2018
.travis.yml fix travis Jan 3, 2019
README.md add dataset test stub Jan 3, 2019

README.md

DataHaskell/dh-core

Build Status

DataHaskell core project monorepo

Aims

This project aims to provide a native, end-to-end data science toolkit in Haskell. To achieve this, many types of experience are valuable; engineers, scientists, programmers, visualization experts, data journalists are all welcome to join the discussions and contribute. Not only this should be a working piece of software, but it should be intuitive and pleasant to use. All contributions, big or small, are very welcome and will be acknowledged.

Architecture

One single repository allows us to experiment with interfaces and move code around much more freely than many single-purpose repositories. Also, it makes it more convenient to track and visualize progress.

This is the directory structure of the project; the main project lives in the dh-core subdirectory:

dh-core/
  dh-core/              
  dh-core-accelerate/
  ....

Building

We use the stack build tool; in the dh-core subdirectory, run

$ stack build

and this will re-build the main project and the contributed packages.

More detailed developer information can be found in the wiki : https://github.com/DataHaskell/dh-core/wiki/Development

GHC and Stackage compatibility

Tested with :

  • Stackage LTS-11.22 (GHC 8.2.2)
  • Stackage LTS-12.13 (GHC 8.4.3)
  • Stackage LTS-13.0 (GHC 8.6.3)

Contributing

  1. Open an issue (https://github.com/DataHaskell/dh-core/issues) with a description of what you want to work on (if it's not already open)
  2. Assign or add yourself to the issue contributors
  3. Add code on the issue branch
  4. Add tests "
  5. Add yourself to the list of maintainers in the .cabal file (if you're not already)
  6. Send a pull request, referencing the issue
  7. Merge only after another contributor has reviewed and approved the PR

Contributed packages

A number of authors and maintainers agreed to move ownership of their repositories under the dh-core umbrella. In some cases, these packages were already published on Hackage and cannot simply disappear from there, nor can this new line of development break downstream packages.

For this reason, contributed packages will appear as subdirectories to the main dh-core project, and will need to retain their original .cabal file.

The stack tool can take care of multi-package projects; its packages stanza in the stack.yaml file has only its directory as a default, but can contain a list of paths to other Cabal projects; e.g. in our case it could look like:

packages:
- .
- analyze/
- datasets/

Packages that are listed on Hackage already must be added here as distinct sub-directories. Once the migration is complete (PRs merged etc.), add the project to this table :

Package Description Original author(s) First version after merge
analyze Data analysis and manipulation library Eric Conlon 0.2.0
datasets A collection of ready-to-use datasets Tom Nielsen 0.2.6
dense-linear-algebra Fast, native dense linear algebra primitives Brian O'Sullivan, Alexey Khudyakov 0.1.0 (a)

(a) : To be updated

NB: Remember to bump version numbers and change web links accordingly when moving in contributed packages.