Wolfram bros seek code slinger posse for IBM and Google round-up The brothers behind the powerful Wolfram Alpha search platform and Wolfram programming language want code-slingers to take on IBM and Google. Stephen and Conrad Wolfram, US and European CEOs respectively of Wolfram Research, are trying to attract Java, Python and Ruby programmers to Wolfram-as-a-cloud. The goal is for those …
This post has been deleted by its author
Your ignorance of the Wolfram Language is showing.
Its not even in the same level as Assembler - one line of Wolfram code is potentially pages of assembler.
Perl is a little better comparison in that they are both effectively scripted/interpreted languages but it would lack the graphical and algorithmic power that the WL has out of the box. The WL is very very high level - hence the large number of functions - 5000 odd.
In terms of language comparison Java, Python and Erlang are all closer. This is as good as summary as anything :
http://stackoverflow.com/questions/4430998/mathematica-what-is-symbolic-programming
In current real world use cases outside of academia its used most often as a rapid prototyping language for complex Algorithms, scientific and quantative analysis work, or as a "gold standard" independent check for algorithms written in another language. Wolfram/Mathematica code can be compiled, but for pure speed implementations the final algo would be written in a C variant.
This link is a good speed comparison - albeit that the different construction styles that are optimal for each language vary so much that its almost impossible to do a "fair" comparison across different languages for anything but the most trivial examples. i.e. Procedural programming styles generally run like a dog in Mathematica.
http://www.stats.uwo.ca/faculty/mcleod/epubs/MatrixInverseTiming/default.htm
I haven't yet gotten to play with WL, but my experience with Mathematica has shown that it can solve problems in seconds that entire research teams sometimes spend months working on, I witnessed it first-hand when visiting a Stanford genetics research lab - they were working for years on this grotesque C++ framework for modeling something that I found a function that already did it in Mathematica.
When I asked why they didn't just use Mathematica, the reply was that they only used open source tools, so they had received funding to write their own solution. I couldn't believe how it was possible in an advanced learning institution to make such short-sighted and dogmatic decisions - instead of spending a couple grand on some licenses, they spent hundreds of thousands writing some very nasty code that nobody would ever use outside of their team.
I hope someday the Raspberry Pi with Wolfram Language comes standard for every schoolchild - I am an almost rabid proponent of open-source, but only a nutcase would have an objection to every kid having super-computer capabilities for $25.
@AC: As the author of scientific codes, the license of Mathematica makes it impossible to publish in a useful way i.e. you need to include Mathematica with your code.
Funded implementations of published algorithms are needed, because believe it or not we can't trust binary blobs.
A company could use it, academic lab not so.
Do you work for Wolfram by any chance, because otherwise I would say this is a mild troll....
P.
When I asked why they didn't just use Mathematica, the reply was that they only used open source tools, so they had received funding to write their own solution. I couldn't believe how it was possible in an advanced learning institution to make such short-sighted and dogmatic decisions
If your foundations are proprietary, you can have the rug whipped out from under you at any time at the whim of some 3rd party. Open Source may be the hard way at times; but the reasoning is anything but short-sighted.
Thats BS. Wolfram provide a free reader (CDF player) as a browser plug-in that does most things, a paid reader that can deploy as enterprise wide apps, and a cloud solution that can be set up for public access, not to mention various internally hosted solutions.
The number of algorithms that are significant enough to warrant a public implementation is tiny.
Mathematica and most similar packages all have the capability to re-construct their inbuilt functions and algorithms from first principles if you choose, and such funding could just as easily be spent writing a test script to verify that 2 closed source implementations meet a gold standard. You'd have to write the test plan for a public implementation anyway - this way to get to junk huge amounts of developer time re-invent the wheel for the nth time.
Whining about having it open source is just Not Invented Here Syndrome with a FOSS fig leaf. Its a been well observed that Linus' law fails where the number of involved people is small, so the additional protections that result may not be realised.
Don't get me wrong MMA and Wolfram have their weaknesses - just the ones listed in the prior post are either inaccurate or ideological in nature - which aren't necessarily good reasons to commercial decisions upon. Such as employing x professional software developers at $100y,000 for the sake of a few $k in purchase and license fees to Wolfram (or any similar vendor).
Full disclosure - part time Mathematica developer in financial services.
As far as I can tell, I was neither inaccurate nor ideological. FOSS isn't a panacea for all of mankind's ills, but -from a long-term planning point of view- it has the dual advantages of distributed control and momentum.
Proprietary software is by definition controlled; and usually by a small group of people. There's less momentum there and that means that small changes can have a very large effect on the direction of a project. Now I love Wolfram Alpha to bits, and I doubt for a moment that the Wolfram brothers are going to go all Oasis and start trying to kill each other with guitars and throwing servers off the balcony; but centralised control is a risk if you're going to be throwing a lot of money in or if you're planning infrastructure.
So yes, I can absolutely see why planners would go open source where possible, and I also mentioned that it was not always the easy way. I was merely pointing out that using open source -as opposed to the proprietary solution proposed by the AC I replied to- might be a decision based on common sense and what was good for that particular group of people at that time; and not "short-sighted and dogmatic". That was as evangelical as I got; unlike the AC who seemed personally miffed that a group of people didn't do things his way.
I could have missed it but I have looked at their website a number of times and there does seem to be a complete lack of explanation understandable to those of us mere mortals who have not been immersed in the world of Mathematica for the last 20 years or so WHAT THE HELL IT IS. Also there is a disturbing lack of details as to what commercial basis it would be sold for as part of a platform we could use to build applications - and I can see that there are probably use cases for it for us (we write and sell high end engineering software). This is, of course, far from being the first bit of software where people are so intoxicated with their own cleverness and have been using / writing it for so long they forget that they have to explain it to people.
To some extent the (vast) library that Wolfram Language has looks to me like it could be viewed as a curated equivalent to CPAN and Ruby / Python equivalents - and I can see the value in a language having a consistent and documented set of libraries (and whilst many of the Ruby ones are good, I can see that there could be a benefit of someone with a big stick standing behind people saying 'no, you are going to do it THIS way, you are not a special snowflake who gets to choose'.
Thats a good comment - something that I've discussed with their reps a number of times. They are aware they have an image/comms problem outside of their core user base (academia and quants mostly), and the stuff El Reg is covering is some of their response to that.
Additionally Mathematica is hard to pidgeonhole due to the sheer breadth of functionality within it.
Basically its a programming language kernel that is very good with any problem that can be expressed mathematically or stastically, plus a set of UI's that allow to that language to be displayed with a variety of different paradigms ranging from Slideware, Word style docs, Fat client apps and various web enabled UI's.
Imagine the MS Office (including Access) suite with a focus on calculation, and visualisation instead of data and text, designed for Mathematical use cases over the last 20 years, and now going all cloudy and you're pretty much there. Everyone can now recoil in horror due to the mention of MS :)
To address commercial usage question. Their commercial terms are pretty reasonable in my experience - for instance 1 Premier license (a few £k) gives you unlimited distribution rights for apps within your organisation.
For commercial usage - there are a few examples out there - my favourite which is apparently 90% Mathematica/WL code :
http://emeraldcloudlab.com/how-it-works
I could have missed it but I have looked at their website a number of times and there does seem to be a complete lack of explanation understandable to those of us mere mortals who have not been immersed in the world of Mathematica for the last 20 years or so WHAT THE HELL IT IS.
Really? I thought the "Technologies" section of the website does a pretty good job of explaining the various components. The Wolfram Language Principles and Concepts page, for example, seems to lay it out concisely: it's a programming language that offers multiple paradigms and constructs (OO, functional, symbolic and numerical evaluation, etc), a very large library of algorithms and tools ("meta-algorithms") to help select the best algorithms for the problem domain, and support for many types of data (including e.g. natural language input).
It's a language designed for solving number-crunching problems. It emphasizes having many domain-specific features rather than notational purity, general abstractions, or simple language design. That makes it suited for a particular class of problem - I wouldn't suggest trying to write an OS or do simple business transaction processing in WL (though it could be very useful for analyzing those transactions). But ffor that class of problem it tries to offer as much as is possible, given the practical and theoretical limits of computation.
Of course it helps to understand what symbolic evaluation is, and have a decent background in algorithms, etc, to understand what something like WL can bring to the table.
The competitors to WL really aren't C++, Java, Ruby, or Python. They're languages designed for this kind of work, like Julia and R - but both Julia and (to a lesser extent) R are more like traditional general-purpose languages than WL. They're not kitchen-sink languages. For symbolic evaluation with Julia, for example, you need to use a package; it's not built into the core language.
But really the main ideas are not that elusive. Ever work with differential equations, or even simplify algebraic ones? WL and Mathematica (which is basically a fancy REPL for WL) can do that sort of thing. Ever take a math-heavy algorithm from an article in a CS journal and implement it? WL is designed to make that easier. Ever have a big data set (structured or not) and want to try grinding it through various sorts of statistical analyses or machine-learning models to see what sort of results you get? Again, WL provides the building blocks, plus assistance in fitting them together.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
