Benji Smith wrote:
Andrei Alexandrescu wrote:
Daniel Keep wrote:
Andrei Alexandrescu wrote:
Michel Fortin wrote:
On 2009-08-09 11:10:48 -0400, Andrei Alexandrescu
<[email protected]> said:
It's also arguable that all functions in std.string should take
const(char)[]. Or, you know, const(T)[], since D supports encodings
other than UTF-8, despite what std.string leads you to believe.
Yah, I think they should all be parameterized so they can work with
various character widths and even encodings.
But shouldn't they work with *ranges* in general, a string being only
a specific case?
That's true as well! In my dreams, me and the famous actress... oh
wait,
wrong dream. In my dreams, I eliminate std.string and put all of its
algorithms, properly generalized, in std.algorithm, to work on more
than
just arrays, and more than just characters.
Andrei
How do you define 'tolower' on non-characters?
That and others would remain specific for characters. I do help to be
able to abstract functions such as e.g. strip().
Andrei
How would you generalize the string functions into ordinary array
functions while still taking into account the different character types?
For example...
dchar needle = 'f';
char[] haystack = "abcdefg";
auto index = haystack.indexOf(needle);
That code is roughly equivalent to this code for generalized arrays,
which seems reasonable enough...
float needle = 2.0;
double[] haystack = [ 1.0, 2.0, 3.0 ];
auto index = haystack.indexOf(needle);
...since "float" is implicitly castable to "double".
But the string example has weird monkey-business going on under the
covers, since dchar is wider than char, and therefore a single dchar
element might consume multiple slots within the char[] array.
Are there any analogous examples of that behavior with other types,
where you'd search for a single element striding multiple indexes within
an array of narrower values?
--benji
Great question. I plan to add a range ByCharacter and a function
byCharacter that spans a string one dchar at a time. Actually I have the
implementation already, haven't checked it in yet. That's a
bidirectional range. In that approach, if you write:
char[] haystack = "abcdefg";
auto pos = find(haystack, needle);
the search is done with the usual array semantics, but if you say:
char[] haystack = "abcdefg";
auto pos = find(byCharacter(haystack), needle);
then a dchar-at-a-time search is performed.
This idea is raw, so improvements are welcome.
Andrei
