Difference between revisions of "2019 Workshop:Software Engineering for Heliophysics"

From CedarWiki
Jump to: navigation, search
(Tutorials)
(Tutorials)
Line 34: Line 34:
 
* Mobile / Web app development for crowd-sourced science: Aurorasaurus (E. MacDonald 14:30-14:45)
 
* Mobile / Web app development for crowd-sourced science: Aurorasaurus (E. MacDonald 14:30-14:45)
 
* Instrumentation software engineering -- keeping the data flowing (open slot, please contact us 14:45-15:00)
 
* Instrumentation software engineering -- keeping the data flowing (open slot, please contact us 14:45-15:00)
* Intro to concurrent / parallel execution: examples in Python (asyncio, ProcessPool, ThreadPool) (Hirsch 15:00-15:10)
+
* Intro to concurrent / parallel programming: examples in Python (asyncio, ProcessPool, ThreadPool) (Hirsch 15:00-15:10)
 
* Room discussion: what topics did we miss? What should we do more of? Should we have side sessions this year? (15:10-15:30)
 
* Room discussion: what topics did we miss? What should we do more of? Should we have side sessions this year? (15:10-15:30)
  

Revision as of 10:06, 6 June 2019


Location: Monday, 13:30-15:30, Zia/Eldorado

Conveners: Michael Hirsch, Matthew Zettergren, Guy Grubbs

Description

This session targets those developing software and doing analysis, modeling or data collection every day (students and early-mid career) as well as more senior scientists interested in the best trends and techniques from industry as applied to heliophysics. We will discuss intermediate to advanced geospace software engineering at a level accessible and useful to all.

Scope includes coding languages commonly used in heliophysics, including: C++, Fortran, Matlab, Python

Use cases we address include:

  • scripting languages to analyze large data sets quickly
  • model developers reduce the time spent tutoring new users in building / modifying / using the model
  • reduce debugging effort by adding automated self-tests
  • ensure code will be usable on most current computing platforms and easily adaptable to future systems

We intend that most participants will be able to apply industry best-practices to their own work tonight, if not in the workshop itself.

Tutorials

Please let the organizers know if you have any additions or interest in speaking on any of these areas. Feel free to bump Hirsch from a spot if you would like to give a talk on that topic!


  • Intro: How to do X with Y--what aspects of software engineering are most important to heliophysics community (Hirsch 13:30-13:35)
  • Software "version control": working effectively in large and small teams with Git and GitHub (Hirsch 13:35-13:45)
  • Continuous Test / Integration: catching and tracking problems before you know they exist (Hirsch 13:45-14:00)
  • Making heliophysics data archival and accessible via lightweight standard: HAPI (R. Weigel, 14:00-14:15)
  • Challenges in parallelizing OpenMPI physics model: Fortran 2018 design patterns and Python integration (Zettergren 14:15-14:30)
  • Mobile / Web app development for crowd-sourced science: Aurorasaurus (E. MacDonald 14:30-14:45)
  • Instrumentation software engineering -- keeping the data flowing (open slot, please contact us 14:45-15:00)
  • Intro to concurrent / parallel programming: examples in Python (asyncio, ProcessPool, ThreadPool) (Hirsch 15:00-15:10)
  • Room discussion: what topics did we miss? What should we do more of? Should we have side sessions this year? (15:10-15:30)

Other sessions of interest

Priority topics

We didn't get to cover all of these, but they're important in general for heliophysics research and in general

  • Modern / efficient coding practices (for any language)
    • What should be object-oriented vs. functionalized
    • use of linters / type checkers (Python: flake8, mypy; C++: clang-tidy)
    • deciding what language(s) are best for a project and the development team
  • Language transitions / interfaces
    • proprietary software ↔ open world (e.g. IDL → GDL → Python)
    • using open software in a closed / proprietary environment
    • Python for Matlab / IDL users
  • Distributing code more easily via:
    • code sharing sites (GitHub)
    • Build systems (Meson)
    • Package managers (Julia, Python)
    • proprietary (IDL) or less common languages
  • Version control (Git):
    • sharing and developing code effectively across diverse teams
    • sharing versioned big data files as part of a program/library
  • Build systems (CMake, Meson, Pip)
    • make it easy for users to get prereqs and build your code on any computer
    • Deploying a complex application anywhere, from Raspberry Pi to Windows/Mac laptop to CentOS HPC
  • Continuous test / integration (Travis-CI)
    • automatically run tests "in the cloud" on Linux, Mac, Windows for each code change
    • examples with Meson and CMake + compiled languages (C++, Fortran)
    • Seamless integration and testing of scripted languages (Matlab, Python) with compiled (C++, Fortran)
    • how to create tests in your preferred language
  • Asynchronous architectures: parallel and concurrent processing
    • Examples in Python and Fortran:


Special technology requests

tables so attendees can use laptops. WiFi.

Justification

ST #5: Fuse the Knowledge Base across Disciplines


  1. Good software engineering practices expedite reliable, repeatable science results and encourage diverse outside collaborator participation. Repeatable, traceable science analyses are better trusted and solidify community and stakeholder confidence in published results.
  2. Software sharing / collaboration sites like GitHub have matured and are widely used by the heliophysics community. We address minor changes in practice that reduce time to science closure.
  3. Progress is readily measured by semi-automated metrics such as:
    • quantity and diversity (institutional, geographical, participant) of code contributions and issues opened
    • an increase in the use of continuous integration facilities such as Travis-CI
    • increased use of public data sharing such as Zenodo

ST #6: Manage, Mine, Manipulate Geoscience Data and Models


  1. Increased scientific computing efficiencies are essential for computer-aided discovery of the growing petabytes of data collected. Extracting value from the decades of diverse existing data sources can be greatly aided by more efficient software engineering practices
  2. Effective use of software toolchains can be a significant force multiplier in avoiding mistakes and repeated or manual work.
  3. Progress might be measured by mining papers for citations / keywords used such as links to software repos used, which can themselves be mined for use of continuous integration tools, build system type and specific software libraries

Workshop Categories

  • Altitudes: all
  • Latitudes: global

Format of the Workshop

Tutorials (2 hours)

Estimated attendance

100

Workshop Summary

This is where the final summary workshop report will be.

Presentation Resources

Upload presentation and link to it here. We will also try to archive talks in Zenodo.

Upload Files Here

  • Add links to your presentations here, including agendas, that are uploaded above. Please add bullets to separate talks. See further information on how to upload a file and link to it.