I am not advocating for doing everything perfectly from the start but I think scientists who write code should at least try to learn some basic software engineering skills.
I am talking about simple things like writing a function instead of copy/paste the same 500 lines of code 10 times, using STL containers instead of heap allocated arrays everywhere, meaningful variable names or (my favorite) using variables for important numbers instead of putting the number 8 everywhere into the code so you don't know if 7 is "8-1" or really is just 7.
Some of the scientists I worked with somehow took pride in not following any coding practices because they were scientists, not mere coders.
I believe in this particular startup I worked for these practices contributed to the eventual failure since we had 100s of thousands of lines of prototype code that crashed left and right and nobody knew exactly what it's doing because there was almost no documentation.
I think a little respect for engineering practices is a good thing for scientists and I don't think it impedes their creativity.
A lot of scientists are ether self taught coders, or taught by other self taught scientists and so are ignorant of bad practices to avoid. Developers with experience should be helping educate the scientists in a positive way that the scientists don't feel like they are considered idiots by the developers. Respect is the key.
The problem with the startup you describe sounds more like a management problem. Hacked code that sometimes works is not code that should be used more than once and it should be rewritten before check in. If the scientists don't know how to do this then they need to be helped.
I am talking about simple things like writing a function instead of copy/paste the same 500 lines of code 10 times, using STL containers instead of heap allocated arrays everywhere, meaningful variable names or (my favorite) using variables for important numbers instead of putting the number 8 everywhere into the code so you don't know if 7 is "8-1" or really is just 7.
Some of the scientists I worked with somehow took pride in not following any coding practices because they were scientists, not mere coders.
I believe in this particular startup I worked for these practices contributed to the eventual failure since we had 100s of thousands of lines of prototype code that crashed left and right and nobody knew exactly what it's doing because there was almost no documentation.
I think a little respect for engineering practices is a good thing for scientists and I don't think it impedes their creativity.