search

JacquesCarette / Drasil

Generate all the things (focusing on research software)
https://jacquescarette.github.io/Drasil
BSD 2-Clause "Simplified" License
143 stars 26 forks source link
code-generation documentation-generator drasil-framework dsl haskell haskell-language knowledge scientific

readme

Drasil Build Status DOI

Generate all the Things! Visit our website to see all of the currently generated artifacts.

Drasil Tree

Table of Contents

  1. What is Drasil?
  2. Quick Start
  3. Building Specific Examples
  4. Running the Example(s)
  5. Finding / Building the Haddock Documentation
  6. Repository Contents

What is Drasil?

For well understood domains, building software ought to be a matter of engineering, based on solid scientific foundations. The ultimate test of "well understood" is being able to teach the domain language to a computer. Drasil is a framework for generating all of the software artifacts for (well understood) research software, from the natural knowledge base of the domain.

We take advantage of the inherent duplication of knowledge present in software artifacts (code, specification, tests, etc). More precisely, we capture the information present in software artifacts so that the particular view of that information in the artifacts can be reproduced by Drasil. For example, the equation F = ma will look different when rendered in documentation and in Java or Python, although it will mean the same thing. In this way, we obtain traceability: we know the exact relationship between information in the specification document and in the code and, furthermore, we know that they are coherent by construction.

Drasil is based on a combination of the following ideas:

  1. domain knowledge changes very slowly
  2. software in well-understood domains can be programmatically assembled from existing knowledge
  3. the various artifacts that make up software are different views on the same knowledge
  4. the most important information in crafting software are the design decisions and their rationale
  5. a lot of software lives for a very long time (10+ years, often as long as 40 years), which needs a different approach

To better understand the requirements for Drasil, we follow an example-driven approach, based on a set of case studies. This is akin to test-driven engineering but at the system level. The currently generated examples serve as a good introduction to what we mean.

We wrote a position paper detailing our original ideas - but this is getting somewhat obsolete now. You can also take a look at a poster. For more information on the details of Drasil, please see the Drasil Wiki. A collection of Drasil-related papers can be found here. To contribute to this project, visit the Contributor's Guide.

Quick Start

If you are on Windows, we recommend you use Cygwin. If you already have Git Bash installed, you can use that instead; you will just need to download util-linux-ng, which includes various system utilities (one of our scripts uses rev), and add its bin/ to your PATH. make is required as well and can be installed following these steps; on MacOS, you may need to install XCode to get that. Most Linux installs have it by default. You may also need to install git.

  1. Ensure you have Stack installed (if you have the Haskell Platform, you should already have Stack).
  2. If using Windows, you might need to add an exclusion to Windows Security for your bin/ folder where your stack.exe is to prevent Windows Security from blocking or even deleting the executable.
    • This can be done by going to Start > Settings > Update & Security > Windows Security > Virus & threat protection > Manage settings (under Virus & threat protection settings) > Add or remove exclusions (under Exclusions), then selecting the bin/ folder with your stack.exe
    • If Windows Security deletes the executable, simply reinstall it
    • This issue was encountered in Windows 10
  3. Run stack setup while in ./code/
    • Remember to change your working directory to ./code/ first
    • Use cd to change working directory, pwd to print your current working directory
    • Refer to File Directory for further help regarding file directory commands
    • e.g. ./Users/.../GitHub/Drasil/code (on MacOS)
  4. Use the basic make command to build Drasil and run all examples.
    • Run make help for a list of available commands.
    • Warning: this entire process takes around 10-15 minutes to complete (MacOS estimate)
  5. You can find the generated output in the build folder that appears in the ./code/ folder. Each example will have its own subdirectory.

For more information, please visit the New Workspace Setup Wiki.

Building Specific Examples

Simply run: make argument to build and run the corresponding example, where argument is detailed below:

Argument Example
gamephysics_diff 2D Rigid Body Physics Library
swhs_diff Solar Water Heating System with Phase Change Material
glassbr_diff Glass-BR
hghc_diff HGHC Toy Example
ssp_diff Slope Stability Analysis
swhsnopcm_diff Minimal SWHS Example, with PCM Removed
projectile_diff Projectile Motion Analysis
pdcontroller_diff Proportional Derivative Controller
dblpend_diff Double Pendulum
sglpend_diff Single Pendulum

For more commands related to Drasil, use make help or check out the Makefile documentation.

Running the Example(s)

Please note that if make has been used, the Software Requirements Specification (SRS) documents are already generated automatically and can be found in build. Automated testing can be done on these examples.

After building, you can run the examples by using stack exec NAME where NAME is detailed below:

NAME Example
gamephysics 2D Rigid Body Physics Library
swhs Solar Water Heating System with PCM (SWHS)
glassbr Glass-BR
hghc HGHC toy example
ssp Slope Stability Analysis (SSP)
swhsnopcm SWHS without PCM (SWHSNoPCM)
projectile Projectile motion analysis
pdcontroller Proportional Derivative Controller
dblpend Double Pendulum
sglpend Single Pendulum

This runs the examples manually from the .stack-work folder after building, and the generated docs will appear in this folder (i.e. in the SRS folders). Due to this placement, these generated versions will not be subject to automated tests. The tex files are generated, but they are not automatically compiled. To compile the tex files, use the generated Makefile (in the same folder as the tex file).

Finding / Building the Haddock Documentation

You can run make docs from the ./code/ folder to build the documentation. Note: this process can take about 10-12 minutes (MacOS estimate).

See the README in ./code/ for more information.

Citation

Please use our provided CITATION.cff file for our preferred citation information. GitHub provides an export of it in BibTeX and APA if needed.

For Any Developers

Please note that we only add to our CITATION.cff/preferred citation file on a consensual basis. If you would like your name added, please submit a PR with your information added to the CITATION.cff file.

Summary of Folder Structure and File Contents

code

doc

notes

Papers

People

Presentations

WindowsFix

.gitattributes

.gitignore

CITATION.cff

LICENSE

README.md