Hacker News Reader

Comment by knolan | Hacker News Reader

I guess it isn’t popular in these parts, but there is a simplicity to Matlab that’s hard to beat for lowly mechanical engineers. The documentation is also generally excellent.

Unfortunately most undergraduates get locked in early with free student versions and continue to use it in industry. Currently I’m trying to expose my students to both Matlab and Python/numpy in a matter of fact manner.

View on HN

I have found that in any situation that goes beyond one off research or experimentation into something that needs to be orchestrated oe deployed in production, you need to move to a language with a lot of bells and whistles libraries which are usually better accessed proceedurally rather than functionally. This is because at that stage, your users are no longer people who understand your code base or are even know programming necessarily.

You can do algorithm development in MATLAB or R or Julia or whatever and figure out ways to either translate the logic to Python/C++ later or setup some kind of IPC for production software engineers to call (platform/orchestration stuff is usually Python or Java)...

...OR you can just develop your algorithms in Python and keep code comprehension and platform integration much simpler for everyone

I'm a senior member of a small industry R&D team that's adjacent to a product development department. New engineering recruits have typically been exposed to Matlab. But they cover a pretty wide range in terms of their actual aptitude and interest in it. Only a fraction of people can really learn to program, and I don't think we understand why.

What I've noticed is that the ones who have gained some proficiency in programming can switch to Python in a jiffy. So I don't think it's a life or death choice.

In my experience, anyone can learn to program but only a fraction of people can learn enough or get good enough to move into senior and beyond roles. You can learn the language syntax, and have enough best practices (that you may or may not understand) jammed into your head, that you are still a net benefit to the team. But the folks who move into senior/staff/principal roles either have a natural aptitude for it, or spend a lot more than 40 hours a week on it, or both.

Indeed, what I'm guessing is that we're not talking about people who are destined to become software developers at all, since those folks tend to learn something other than Matlab. In my own case, I'm at a "principal" level for reasons other than programming, at least I hope. ;-)

The use of Python in my work group is largely as a problem solving and prototyping tool. The product development teams that we serve have their own software departments, with their own choices of languages and so forth.

Yeah, if your problem is mostly banging around matrices, Matlab feels really great (it is, after all, Matrix Laboratory). I prefer python for the most part, but it does feel a lot more verbose in comparison whenever you are in Matlabs wheelhouse.

Not to mention the IDE is actually very good, and you almost never have to deal with package management etc.

> Unfortunately most undergraduates get locked in early with free student versions and continue to use it in industry.

Why no octave? Or julia?

I'm afraid Julia will always be ignored in favour of Python. Python/C++ have a firm grip over its niche for years to come.

I think the question is "why Julia?" instead of "why not Julia?" Python is the de facto standard, for better or worse. If someone wants to unseat it, they need a compelling reason. You shouldn't put the onus on folks defaulting to the industry standard to justify their choice over a much less well-known competitor.

Depends on how much people love to write "Python" that is actually C with Python bindings.

There are many compelling reasons. Doesn't seem that way if Julia is just construed as a faster python with some Matlab.

Goes far beyond that

I hope you're just alternating between the two rather than making them double up on homework. As a student I would find that quite annoying.

Definitely not doubling up. Some parts of my modules are in Jupyter Notebooks, others inherited PowerPoint that I’m slowly phasing out. I’ve added various examples where I show how a particular equation might be plotted with Matlab and better understood.

Reply on news.ycombinator.com