GNU Octave: A high-level interactive language for numerical computations [pdf] • Hacker News Board

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#21

Post by ThenAsNow » Sat, Jan 19, 2019, 8:31 AM UTC

chrispeel wrote: post #17

wanderfowl wrote: post #7

As somebody in an academic field where many scripts exist for MATLAB (mostly digital signal processing focused), there are few projects I want to succeed more than Octave. It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in devel…

just use Julia. It's faster than Matlab, and a better language.

It will be very hard to displace Matlab out of certain very important niches. For example, across both academia and industry, Matlab is king in control systems design/analysis thanks to its toolboxes, Simulink, and (at least in industry), its autocoder. Matlab is used to generate the flight controls software on a lot of today's aircraft. In a number of other engineering disciplines, Mathworks has been astute in developing toolboxes that answer the needs of industry. There will probably have to bit quite a bit more paid development of Julia and its ecosystem to deal Matlab death blows in industry. I hope for it to happen sooner rather than later, as the 70s-era Matlab language and Mathworks' exorbitant license fees are holding things back.

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#22

Post by marmaduke » Sat, Jan 19, 2019, 8:32 AM UTC

wjn0 wrote: post #10

wanderfowl wrote: post #7

As somebody in an academic field where many scripts exist for MATLAB (mostly digital signal processing focused), there are few projects I want to succeed more than Octave. It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in devel…

> It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in developing nations. The future is here. As a research student in a developed nation, it's very annoying to deal with MATLAB projects. Our lab has a fairly strict no-new-code-w…

I know it’s not cheap but we have exactly one copy of MATLAB on our cluster to use their compiler and then run the compiled code on nodes without licenses.

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#23

Post by marmaduke » Sat, Jan 19, 2019, 8:42 AM UTC

wanderfowl wrote: post #7

As somebody in an academic field where many scripts exist for MATLAB (mostly digital signal processing focused), there are few projects I want to succeed more than Octave. It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in devel…

> Octave does to MATLAB what R has done to SPSS and SAS

It won’t because all those new algorithms aren’t written in Octave (for the most part), but MATLAB (or Python or Julia).

MATLAB is not a dinosaur: it has had line profiler, JIT compiler, 3D plotting, builtin scriptable debugger in the default installation, long before Python or Julia, who still struggle with these highly valuable tools. Now it has notebooks, and plenty of other new stuff inspired by recent new stuff in Python et al. On top of that there is a Python FFI so you can call out to a Python installation if need be.

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#24

Post by IshKebab » Sat, Jan 19, 2019, 9:49 AM UTC

wjn0 wrote: post #10

wanderfowl wrote: post #7

As somebody in an academic field where many scripts exist for MATLAB (mostly digital signal processing focused), there are few projects I want to succeed more than Octave. It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in devel…

> It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in developing nations. The future is here. As a research student in a developed nation, it's very annoying to deal with MATLAB projects. Our lab has a fairly strict no-new-code-w…

What do you use instead though? Octave is very slow, Julia is even slower and Numpy has an annoyingly verbose syntax because it has to be valid Python.

Additionally I've not found a plotting solution that is as robust and fast as MATLAB.

MATLAB seems to me to be one of those commercial programs that has no good FOSS alternatives. Maybe it's just too niche, like CAD and EDA.

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#25

Post by cbkeller » Sat, Jan 19, 2019, 10:01 AM UTC

IshKebab wrote: post #24

wjn0 wrote: post #10

Earlier quoted context omitted.

> It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in developing nations. The future is here. As a research student in a developed nation, it's very annoying to deal with MATLAB projects. Our lab has a fairly strict no-new-code-w…

What do you use instead though? Octave is very slow, Julia is even slower and Numpy has an annoyingly verbose syntax because it has to be valid Python. Additionally I've not found a plotting solution that is as robust and fast as MATLAB. MATLAB seems to me to be one of those commercial programs that has no good FOSS alternatives. Maybe it's just too niche, like CAD and EDA.

> Julia is even slower

Huh? This has definitely not been my experience -- most of the things I've ported over seem to run about twice as fast in Julia as in the original vectorized Matlab (and probably more like 10x faster than naive non-vectorized Matlab)..

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#26

Post by type0 » Sat, Jan 19, 2019, 10:20 AM UTC

joe_the_user wrote: post #20

wanderfowl wrote: post #7

As somebody in an academic field where many scripts exist for MATLAB (mostly digital signal processing focused), there are few projects I want to succeed more than Octave. It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in devel…

The thing with Matlab is that, just as software, it's "great-terrible". It has an amazing number of cool little tweaks that allow quick development and it has umpteen convenient algorithms and allows one to work with matrices in a very self-contained way. But all this niceness means really that the software is more in the realm of spreadsheets than standard programming languages. Maintenance is guaranteed to be a nig…

I cringe every time that some engineering friend mentions how they has been programming an application with MATLAB. And next time the problem arises this "application" is not applicable anymore. Some manager types do not even realize that there's such a thing as technical debt.

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#27

Post by pvitz » Sat, Jan 19, 2019, 10:26 AM UTC

joe_the_user wrote: post #20

wanderfowl wrote: post #7

As somebody in an academic field where many scripts exist for MATLAB (mostly digital signal processing focused), there are few projects I want to succeed more than Octave. It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in devel…

The thing with Matlab is that, just as software, it's "great-terrible". It has an amazing number of cool little tweaks that allow quick development and it has umpteen convenient algorithms and allows one to work with matrices in a very self-contained way. But all this niceness means really that the software is more in the realm of spreadsheets than standard programming languages. Maintenance is guaranteed to be a nig…

I have to use Matlab professionally in my organization for financial risk modelling and there are some larger legacy codebases. However, I wouldn't share your opinion about maintenance issues. Like with every other language, if the code is well organized, everything is fine.

Matlab integrates with SVN or GIT, you have got the functionality for unit-testing, proper debugging and profiling and namespaces through packages. OOP works also well although you will pay for it in performance. Additionally, you can easily use Java packages and interface with .NET and COM.

Compared to how R is used here, the whole Matlab environment feels much more complete and solid prototyping is certainly an advantage. Nevertheless, we feel the pain when it comes to datasets of 10-100 GB for which we don't want to set up a database. We haven't bought the Parallel toolbox and without this, processing is just very slow. Here R shines with e.g. data.table.

If I would rewrite the applications, I would however not do it in Matlab, because the audience of people looking at the code is more familiar with open source alternatives.

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#28

Post by type0 » Sat, Jan 19, 2019, 10:29 AM UTC

noufalibrahim wrote: post #18

wanderfowl wrote: post #7

As somebody in an academic field where many scripts exist for MATLAB (mostly digital signal processing focused), there are few projects I want to succeed more than Octave. It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in devel…

The over dependence of academia on MATLAB when (atleast partial) FOSS solutions existed is what prompted the setup of the FOSSEE project in India https://fossee.in/ . They've been conducting several outreach programs to move academics and students off expensive and proprietary (and therefore often pirated) software to free alternatives (mostly around the Python ecosystem). The project is quite successful from what I…

I applaud such efforts, the entrenchment of MATLAB in engineering departments is hard to shake, we need more exposure on open source tools especially in academic conferences.

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#29

Post by dongping » Sat, Jan 19, 2019, 11:10 AM UTC

kkylin wrote: post #9

sitkack wrote: post #2

What benefit does Octave have over the Python ecosystem? Esp as a clone of proprietary software? The only thing I can see is as a way of running Matlab codes in clusters on cloud w/o having to also run the license servers.

I work in an academic setting. If I had the choice I'd prefer Python (or even Julia), but frequently Matlab is what's appropriate in a lower-division STEM (non-CS) course because it's what many engineers are used to and teach or expect their students to know. I'd rather use Octave than Matlab in such settings, and encourage students to do the same. Also, Matlab is still much more common (at least in undergrad enginee…

For those who also use Matlab/Simulink for simulation, Modelica could be a nice alternative with an open implementation, which also has serious industrial application like controlling power plants.

Re: GNU Octave: A high-level interactive language for numerical computations [pdf]

#30

Post by std_throwawayay » Sat, Jan 19, 2019, 11:24 AM UTC

I have used Matlab and GNU/Octave in parallel for some years half a decade ago. Octave can do most things that I needed but was slower to run for many things. With Octave you didn't run into licensing problems as those floating licenses had their limits and sometimes I needed to run quite a few instances in parallel.

Now I use Python exclusively. The language is so much better and more sane and almost any library is available. I will never go back.

Today, I would recommend Matlab only in rare circumstances. Currently I cannot think of one use case that I had in the last three years where I would prefer Matlab to Python. Using the Matlab language with a portable Octave installation can be useful for teaching matrix/vector topics to people who have little programming experience and don't know Python. You can get results very quickly. The syntax feels most natural for manipulating matrices and everything seems to be a matrix.

Other than that I think the Matlab language and ecosystem is kind of a dead end. But it will provide profits for a few more decades due to sheer inertia.