# 006 Internal

Theme: Testing for science, part 1

- introduction

- 10-15 Testing for Science
- Run as a two-person discussion
- https://github.com/ResearchSoftwareHour/rsh-notes/issues/35
- What makes a unit test a "good test"?
- test one thing or several things in one test?
- multiple asserts can lead to later asserts never to be tested
- size of the unit to test
- How to write good unit tests for more complicated functions than int fibonacci(int n)? For example, what if the return value of a function is random?
- demo below
- Radovan "how would you verify it as a human? how would you explain it to me over a telephone?"
- sometimes I test a function comparing with a noddy but "obviously correct" function I wrote earlier
- How to write unit tests for private class-member functions?
- How to test e.g. functions that only print out some summary on the terminal?
- maybe reformulate the function to be "pure" and return a "Maybe"
- or use logging
- one can also capture prints in Python
- How to test the failure of functions, e.g. error messages, warnings.

- 10 Richard: pytest
- quick demonstration
- simple example of a deterministic function
- celsius_to_farenheit
- is testing 0 and 100 enough?
- simple example of a probabilistic function
- testing bounds, e.g. -10 <= x <= 10
- fixture to seed the RNG
- Should the tests be in the same file?

=== halftime===
- 5-10 Radovan: github actions
- Deploy the previous demo to github-actions
- comparison with travis
- also mention GitLab CI, Appveyor, Circle CI

- 10 Radovan: Changing git history: `--amend`, `--fixup`, `--autosquash`
- Why might we want to fix up history?
- When *not* to fix up history?
- `--amend`
- `commit --fixup` and then `rebase -i --autosquash`

- 5 Richard: Blub paradox
- https://en.wikipedia.org/wiki/Paul_Graham_(programmer)#The_Blub_paradox
- Original source: http://www.paulgraham.com/avg.html
- Definition of the paradox
- When is it worth staying with the simpler solution?
- time investment
- number of contributors who can use it
- ecosystem
- When is it worth moving to more complex solutions?
- How do you learn complex solutions, if you just need to "get stuff done"?
- Examples:
- docker and containers
- moving from make to cmake
- moving from one language to another

- conclusion
- epilogue discussion
- Richard: commit messages starting with filename, opinions?

## Ideas

- Github actions: board game network finished
-
- board game networks example

- 10 What's a `#!` ?
- Turning any script to an executable
- #!python
- #!/usr/bin/env python
- #!make -f
- #!
- Next time: how to make a python scirpt suitable for this

- module -> script, __main__, argument parsing

- 5 minutes Radovan
- probably shows Cicero unless I get an inspiration

- blub paradox