comp.lang.idl-pvwave archive: archive » What does an optimal scientific programming language/environment need?

Home » Public Forums » archive » What does an optimal scientific programming language/environment need?

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

What does an optimal scientific programming language/environment need? [message #36427]

Fri, 19 September 2003 11:29

grunes
Messages: 68
Registered: September 1993

Member

I'm working on creating an optimal scientific programming language and
environment. My hope is that people who use current environments have
specific things they love about it, that need to be included. For now
I'm trying to combine the best concepts from FORTRAN, BASIC, C, APL,
IDL, PV-WAVE, and possibly MATLAB.

I have been an applications programmer in support of scientific
research for 23 years, in radar and optical remote sensing, data
compression, and statistical image processing, and have also produced
operational software for ship-borne, airborne, and orbital platforms,
so I have some ideas of my own. But I want other people's input before
I begin, because I expect this to be a lot of work.

What I have in mind so far is:

1. Free or cheap software. Moderate price documentation ($20-50?).

2. Deliverable free and non-free executable products without special
permission or royalty. Unfortunately, this seems to be the single most
important criteria of modern scientific/engineering applications
development, from the sponsor's point of view.

3. Very rapid developement and testing. Requires extreme conciseness,
support of arrays, complex number, linear algebra, finite element and
numerical integration of functions and differential equations, and
little need for type and shape declarations. Can easily switch on
automatic detection of subscript checking, memory reference checking,
argument mismatches, fixed and floating point errors. FORTRAN style
adjustable array bounds (e.g., a(-3:4, 5:7)).

4. Very rapid learning. Most of the language and environment must be
summarizeable in a very few pages. Design must be consistent.
Documentation unambiguous. For those who like lots of words, examples
or homework problems, I could supply a larger tutorial manual.

5. 100% upwards compatibility with earlier versions of the same
language/environment, and between platforms.

6. Close enough to standard mathematical notation to be mostly
debuggable by inspection. But must have ASCII transliteration so
people can use there own editors if they don't like mine.

7. Can add the declaration statments that make efficiency possible.

8. A compiled mode that really is as fast as FORTRAN or C, if you add
those declarations. Compiler would produce 2nd level code for
compilation by g77 and gcc.

9. The interpreter and Executables must be rock solid stable. That
means very few heap dynamic memory allocation operations (malloc,
calloc, alloc, new, allocate). Support dyanmic stack allocation, and a
garbage collected heap area.

10. Screen and file graphics (plots, diagrams, images) must be very
easy to produce. Default style products must be publication quality.

11. Can call and be called from FORTRAN and C programs.

12. Must support two primary development platforms at this time:
Lintel and Wintel (PCs under Linux and Windows). (Nothing else is
economically viable any more.) For the moment, I will ignore multiple
CPU support, networking, and signal processing chips.

13. Be able to associate an area of virtual memory with an entire
file. Should handle raw bytesream files, as well as TIFF scientific
data sets. Since the indicated platforms only support 1-2 GB of user
virtual memory space, I will have to create a special FILE type that
allows 64 bit address manipulation.

14. Multi-precision calculations (e.g., 2, 4, 8, 16 byte and greater
floating point numbers), and exact (rational) calculations must be
easy to do.

15. No bugs. (I'll try.)

Please add your own criteria!

Previous Topic:	how to do *b++=100 as in C
Next Topic:	Error in IDL VM 6.0