A Mathematician looks at D

[FG]

On 2013-02-19 01:28, Nick Sabalausky wrote:
>>> I can do that just as easily without a REPL.
>>
>> With a much reduced interactivity and more slowly.
>>
>
> Slightly so. I wouldn't say "much".
>
> But of course, I'm not saying that a REPL wouldn't be nice to have.
> Just saying that "edit, re-compile/run" really isn't all *that* bad.

If you're processing a significant data set in an exploratory way, with many 
intermediate calculations, then Python with gluplot or whatever beats the 
productivity of edit-compile-run. As another example of interactive vs. 
recompiled approach, I'm fine with using TeX for writing - it is quite 
comfortable even without seeing final formatting when writing (or thanks to 
that) - but making graphics in it using PGF/TikZ instead of an interactive 
illustration program becomes rather painful and is the example of much reduced 
interactivity and speed.

Edit-compile-run way would be quite good (assuming a very fast compiler) if 
processing blocks could be added incrementally and state from the previous runs 
could be _easily_ restored in the next one. Then the difference between a 
scripting language and a compiled one would be blurred.

Coming back to the TeX example. It's a fast typesetting system - can compile 
some books in under a second (up to a few seconds if using XeLaTeX and TTF 
fonts). Yet it's way to slow for testing many small changes or generating 
documents like invoices en masse. It would make a great invoicing system if it 
was possible to save the state of it using an empty document and all necessary 
packages loaded and then only restore that and add the body on each run.