Post by Dmitry A. KazakovPost by Georg BauhausPost by Dmitry A. KazakovPost by Georg BauhausPost by Marius Amado-AlvesLove OT threads in this clever forum:-)
*many* keywords can get in the way of naming identifiers. Ada, for one,
has a bit too many keywords, with some good identifier candidates (when,
others...)
"when" and "others" are general abstractions. What kind
of programs will in effect be about general abstractions?
The kind of programs in Ada language, which uses exactly these words.
And these are? (If you could avoid the tautological void.
Be constructive! ;-)
All Ada programs that use the case-statement.
How now? First, you distinguish syntax from some "rest" of
the program, and then, when I do the same, the distinction is gone?
My question is: If "others" were moved from syntax to identifiers,
then what fraction, and what kind of programs will hugely profit
because they can now name things "others"? (In place of "foreigners",
or "remaining_parts", or "tail", or "<the-opposite>", or ...)
Post by Dmitry A. KazakovPost by Georg BauhausPost by Dmitry A. KazakovYour
argument is bogus. If a reserved word is good for the language then it is
good for a program in that language.
If a reserved word is good for the language, then it is a general
abstraction, such as "when", or "others", because is it applicable to
all kinds of programs regardless of concrete notions.
Rephrased:
(reserved(word) = good-for(C(identifiers))
=> general-abstraction(word)
No further conclusions other than a wish for a list of words
not recommended for use as identifiers either, such as "what".
Post by Dmitry A. Kazakov=> it is bad to use in all programs => it is bad for the language in which
all these programs are written. q.e.d.
IIUC, this is close to what I have said about the word "what" as
an identifier. "What" is bad in programs except maybe in programs
whose problem domain is the singleton word "what". It should not be
used as an identifier, IMHO.
Post by Dmitry A. KazakovSurely "when" and "at" are prefect names for a formal parameter of the type
Time.
I disagree. In my initial response, I made them "Leaving_When" and
so on, as these names are more meaningful. "at" is worthless without
context (at a place? at a time?). You could require context with
dot notation. Pkg.T? Pkg.Obj? Really? I cannot at the moment
think of a really good way to use "at" as an identifier in the body
of a function.
when := offset (when => at);
Seen from another angle, refactoring becomes a nightmare if
you have wish to reuse an Ada function in a PascAlgol-17 program,
by rewriting, say, if its profile uses names that are ubiquitously
reserved :
Index (From => ..., To => ..., );
And vice versa for the reverse.
I'd rather not have Ada fans tell me that they couldn't care
less, why don't I use Ada etc., when at the same time they are
witness to programs breaking because Ada didn't at least warn
against using the fairly general word "interface" when it was
around already. Clear an plain.
Java did better, insofar as "goto" is reserved even though it
is not part of the language's syntax.
Post by Dmitry A. KazakovAlso your argument to universality would also apply to the parameters named
"X", "Y", "Left", "Right" all over the LRM. It did not work there, but
suddenly does for "when"?
A word may work in an LRM. However, a program's problem domain is
not the LRM, in general.
Simon Wright has recently shown a nice way how to mend the situation
with X and Y for surface orientation, using Ada 2012 (substituting
Northings and Eastings in the profile, IIRC).
Post by Dmitry A. KazakovPost by Georg Bauhaus"What" as an identifier shouldn't be, IMHO,
reserved or not, so words like "what" not being reserved does
not count as counter-argument in my book. I'd rather have a
list of words not recommended to augment reserved words.
So either they must be reserved or your argument is wrong. The choice is
yours.
There is no MUST here. Programming is a human activity, not to
be HALTed: What counts is understandability of a program, which I think
is not to be had from some reduction to an ad-hoc rules base on
a formalism made for a different purpose (C(identifier)).
Post by Dmitry A. KazakovStatistically irrelevant.
I know I am not statistically relevant. Yet, for actual statistics,
if you see how programs are written in the low budget industry
that might change your mind. The industry includes Google.