Why are scientific programming languages so weird? [closed]

Question

It seems to me that programming languages meant for use in science and engineering are consistently weird compared to general-purpose languages. Some examples off the top of my head:

• In Matlab, each function has to be placed in a separate file
• In R, <- is the assigment operator, as opposed to = in almost every other language
• Matlab, R, Julia and others are all 1-indexed
• Matlab uses % for comments, and not the standard # or //

Of course, these languages all have several design features that actually make them easier to use for scientific applications, such as more natural matrix notation. Still, they all inexplicably make all these bizarre choices which don't make anything easier and could easily have been avoided if the language designers had just chosen to do what 99% of other languages do. Is the reason vendor lock-in? A lack of contact with the wider software development community? Something else?

I read this thread and didn't find the explanations satisfactory. Just because R were designed as a scientific language doesn't mean it had to completely ignore conventions and use <- instead of =.

Answer 1

• There are different conventions. Conventions in mathematics, logic, and applied sciences and conventions in IT. The first ones are far older.
• The scientific languages are made to make the life of THEIR users more convenient. The user is seen as a scientist, who can realize some algorithm from time to time or to check some theory, without the need to learn something really new. So, the languages for scientists MUST be made up to non - IT standards. Because they are not meant for the use of IT people. They are up to OTHER standards and that is good because of the target auditory. Because the good SW UI, and language is SW UI, must be done based on needs of user, not of the coder.
• Our IT standards are industry standards. IT is industry. Science is not industry. Scientists are proud of it. And they would reluctantly take anything from our practice into theirs. And they don't like standards at all. And nobody likes foreign standards. So, if somebody will make a scientific language that will look up to IT standards, it would be hardly selling well, because of the dislike of the target auditory, even if it were objectively more convenient.

---

And even if we'll judge only according to IT standards... Sorry, what standards do you mean? Have you tried to write a prog in APL or SNOBOL? These two language are, IMHO, the MOST powerful in appropriate fields (counting and strings). But the syntax is something VERY strange (and effective) Reading a line of APL code could take days. On the other hand, such line is a serious piece of SW. You'd return to Mathlab with tears of relief.

As for "=", many people have problems to be accustomed that it is not equality, but assignment. BTW, in Pascal it IS equality and assignment is ":=".

And you really think that == for equality is more natural? On the contrary, mixing = and == is the MOST common error in C programming, it happens very often even in contemporary IDEs, with their automatic control.

About indexing from 1 - it is the only natural one. When you were a child, you had learned poems and songs, where you counted: one, two, three... And not 0,1,2... In school math we studied that the counting starts from 1, and that 0 doesn't belong to natural/counting numbers. Only with the definition of functions non-natural indices come. After all, the 0 was invented many thousands of years after our ancestor raised a finger up.

0-start was more simple to realize and immediately got into IT practice after C appearance. But in Fortran, the first language, the 1-indexing is used. The same with other languages of the pre-industrial epoch.

And yes, I had read Dyjkstra's article on naturality of the 0-based counting. And totally disagree with his argumentation. It is natural for musicians ony. And even 0 enthusiasts that create the C and Java compilers, count the lines of the code STARTING FROM 1!

Answer 2

Indexing from 1 is not weird, it is completely normal and expected except for programmers, because they've been conditioned to expect 0-based counting by C (which was conditioned from the properties of processor architecture).

Comments are denoted in many, many many different ways in different languages; there is no standard way, every language chooses a symbol or digraph that isn't already taken.

Assignment is likewise a strange and incomprehensible concept, except for programmers; most people couldn't care less whether it's = or := or <-, they struggle to understand the meaning (and for them, it is in fact better not to use =, because this emphasizes that assignment is not equality - the most common hurdle for non-programmers to understand code).

In short, programming languages intended for people other than professional programmers look different because the people who use them most want it that way.

Answer 3

There are three problems:

1. You are unaware of certain traditions, and the good reasons for certain choices.
2. You put too much emphasis on syntax, too little on semantics.
3. Engineers and scientists have no experience in language design, leading to questionable syntax.

Now to your specific points:

• I don't know Matlab, so I can't comment on the requirements of file organization. Note that Java wants you to use one file per public class.

• In R, = can be used as an assignment operator as well. Note that it needs multiple assignment operators <- and <<- to deal with its concept of scoping (<<- assigns to a symbol in an outside scope instead of creating a new symbol inside a function). The arrows can be used in the other direction too, potentially making cleaner code: complex_calculation() -> x.

• 1-based indexing is the standard in mathematics, which is what Matlab's and R's users are more comfortable with than C. Julia follows Matlab in order to have a better learning curve.

• % for comments is also used in TeX/LaTeX. The # is only a convention from Unix scripting languages, and their descendants.

You also ignore that “real” programming languages have many weird parts. Why doesn't Scheme use =? Instead:

(define foo 5)

Why does C use * for dereferencing, when obviously a caret ^x is more common in other traditions?

Answer 4

I guess it depends on your exposure to other languages. Off the top of my head:

• C/C++ have separate source files (.c/.cpp & .h)
• The -> characters are used in C# for lambda expressions
• Old versions of VB used 1 as a default index (although this could be changed with Option Base)