On 21/10/2010 18:07, bearophile wrote:
Bruno Medeiros:
Rather, the big benefit of the statement above would be to reduce
certain wasteful discussions or comments that pop-up occasionally in
which someone proposes some "Pythonesque" change that might benefit
small programs but would be crap for medium/large ones.
I suggest to judge each proposed feature on its own, instead of refusing all the
"Pythonesque" changes together.
I agree, and never suggested or implied otherwise!
My reference to Python was a lighthearted one, I did not want to make a
generic statement about all Python features (that's why I put quotes
around "Pythonesque"). I am not even familiar with Python, other than
it's syntax, and in fact, the indentation-as-blocks was the example I
has in mind when I mentioned "Pythonesque".
A well designed tuple unpacking syntax will shorten D code and make it more
readable and more handy to write, so I think it's a positive change, both for
little and large D programs.
As every language feature tuples too may be abused: in large programs if many
of your functions/methods return tuples with five or six different anonymous
fields, your program will not be much readable.
Another "Pythonesque" example of change that I regard as positive for D
programs of all sizes are lazy/eager array/range comprehensions. If you don't abuse them,
they shorten code, make it more readable, and avoid the cluttering of brackets and
parentheses typical of lambdas with map+array/filter+array functions.
Bye,
bearophile
I've just recalled another "corollary" design goal, a strictening of the
previous one, that I would also like to see adopted:
* If changes are of limited use to medium/large scale projects, they
should also not be considered, even if they are not detrimental to
projects of such scale.
The reason for this is to save work in implementing languages tools
(compilers, IDEs, any kind of code analysis tools, etc.). It should be
obvious to everyone that the time of the programmers working such tools,
especially the compiler, is a very precious resource.
Indeed, Walter has already expressed his intention to not adopt features
that are of limited use, however, how useful a feature is is very
debatable and often not agreed upon. So what I am arguing for, in the
point above, is that usefulness should only be evaluated in the context
of medium/large projects.
I am not implying that the tuple features you mentioned above have
limited use like I described. I am not familiar with those changes at
the moment and won't comment on them now.
--
Bruno Medeiros - Software Engineer
