Getting started
This knowledge base is based on A Research Software Engineering Workflow for Computational Science and Engineering (Maric, Gläser, Lehr et al.; preprint available), a paper that originated from the work at the SFB1194 at TU Darmstadt.
See our section in literature.
The articles herein are equipped with tags to help find them as well as they are categorized and chapterized according to the sectioning of aforementioned paper, from which this figure is as well. It therefore is recommended to read this paper carefully and to try to follow this knowledge base when implementing these processes.
Minimum workflow
The minimum workflow contains the elements:
- version control
To keep an overview over your (software) versions and to collaborate with others a Version Control System (VCS) is a must in any (sofware) project.
Version control guide - build system
A build system handles software dependencies and makes your software platform-independent. Use e.g. cmake or wmake. Also see chapter 2.2 Build system - cross-linking (in minimum workflow flavour)
Connect your published evaluation data, code and papers with PID ´s and git tags .
Cross-linking guide - issue tracking (in minimum workflow flavour)
use the kanban system with tasks in the stati ‘To Do’, ‘In Progress’, ‘Done’
Full workflow
The full workflow adds different techniques to the minimum workflow, that can be implemented partially or over time:
- cross-linking (in full workflow flavour)
Connect all your data, your code and papers with PID´s and git tags.
Cross-linking guide - issue tracking (in full workflow flavour)
use a platform, e.g. Redmine, Youtrack or Jira -
continuous integration (CI)
The CI workflow proposed contains some of the other topics, like version control, build system and containerization. Basically this term describes triggering a pipeline of different jobs and evaluations to test the written code.
See Continuous integration on GitLab - containerization
Containers allow reproducibility across different sofware environments. They are immutable. Here Singularity is used. To read about it, see the literature.
Start working with: How to install singularity on a local machine and How to create a Definition file for a Singularity container - test-driven development (TDD)
The idea behind this is to first write your test (what requires setting the goals first) and then implement your code for as long as it takes to pass the tests.
Test driven development and code quality - automatic quantification, visualization and evaluation on HPC systems
Run your tests automatically as well as their post-processesing . Doing so on HPC clusters may impose special challenges, depending on the setup of your HPC.
Test quantification, visualization and evaluation on HPC clusters
Navigating within our Knowledge Base
The articles herein are equipped with tags to help find them as well as they are labeled with the chapters according to the sectioning of aforementioned paper, namely
- Version control,
- Build system,
- Cross-linking publication, software and datasets,
- Continuous Integration,
- Containerization,
- Test-driven development and code quality,
- Test quantification, visualization and evaluation on HPC-clusters.
We also have categories, whether a topic belongs to the
- minimum workflow or the
- full workflow.
Where to start
Continue with the guides e.g. version control with git.