On 2/4/12 10:05 PM, Manfred Nowak wrote:
> Jos van Uden wrote:
>
>> bool isPangram
>
> The presented code is not an acceptable example for the usage of the D
> programming language.
>
> 1)
> `indexOf( s, c)' has a worst case running time of O( `s.length').
> `indexOf' is called once for each `c' in the used member `alpha' of
> `Alphabet'.
> Therefore the runtime of the presented code is
>    O( `s.length' * `alpha.length')
> whereas
>    O( `s.length' + `alpha.length')
> is possible.
>
> 2)
> The optional third parameter of `indexOf' can be called with
> `CaseSensitive.no'. But that parameter is left untouched. Instead a
> check with `toUpper( c)' is used, thereby risking a further visitation
> of the whole string .
>
> 3)
> The use of the literal value `-1' stands out of the code and hints to a
> maldesign in phobos. This is because `-1' is not mnemonic.

Sensible arguments. Would you want to redo the example in better style?

Andrei

On 5-2-2012 5:05, Manfred Nowak wrote:

> 2)
> The optional third parameter of `indexOf' can be called with
> `CaseSensitive.no'. But that parameter is left untouched. Instead a
> check with `toUpper( c)' is used, thereby risking a further visitation
> of the whole string .

Using CaseSensitive.no is a lot slower

>> The optional third parameter of `indexOf' can be called with
>> `CaseSensitive.no'. But that parameter is left untouched. Instead a check with `toUpper( c)' is used, thereby risking a further visitation of the whole string .
>
> Using CaseSensitive.no is a lot slower

Hmmm then perhaps a whole re-write. I've gotten this done, although I'll need someone to tell me how well it does. Far as I can tell it's O(n). It's based loosely on the ascii version, except templatized. Hope it's not too ugly...



import std.stdio, std.string, std.traits, std.uni, std.conv;

string T_isPangram(string name, string set, string uset = "") {
	string x = "bool " ~ name ~ "(S)(S s) if (isSomeString!S) {\n";
	if(set.length > 32) {
		x ~= "\tulong bitset;\n";
	} else {
		x ~= "\tuint bitset;\n";
	}
	x ~= "\tforeach(dchar c; s) {\n\t\tswitch(toLower(c)) {\n";
	foreach(i, c; set) {
		x ~= "\t\t\tcase '" ~ c ~ "':";
		x ~= "\tbitset |= 1 << " ~ to!string(i) ~ "; break;\n";
	}

	int i = set.length;
	while(uset.length >= 6) {
		x ~= "\t\t\tcase '" ~ uset[0 .. 6] ~ "':";
		x ~= "\tbitset |= 1 << " ~ to!string(i) ~ "; break;\n";
		uset = uset[6 .. $];
		i++;
	}

	x ~= "\t\t\tdefault:\n\t\t}\n\t}\n\treturn bitset == (1 << " ~ to!string(i) ~ ") - 1;\n}";

	return x;
}


mixin(T_isPangram("EN_isPangram", "abcdefghijklmnopqrstuvwxyz"));
mixin(T_isPangram("DE_isPangram", "abcdefghijklmnopqrstuvwxyz", "\\u00DF\\u00e4\\u00f6\\u00dc"));
mixin(T_isPangram("SV_isPangram", "abcdefghijklmnopqrstuvwxyz", "\\u00e5\\u00e4\\u00f6"));

unittest {
	writeln(T_isPangram("TEST_isPangram", "ab", "\\u00DF\\u00e4"));	//example output below

	assert(!EN_isPangram("this doesn't cover everything"));
	assert(EN_isPangram("the quick brown fox jumps over the LAZY dog")); //to-lower check along with english
   assert(DE_isPangram("Falsches Üben von Xylophonmusik quält jeden größeren Zwerg"));
   assert(SV_isPangram("Yxskaftbud, ge vår wczonmö iqhjälp"w));
}

/*
bool TEST_isPangram(S)(S s) if (isSomeString!S) {
       uint bitset;
       foreach(dchar c; s) {
               switch(toLower(c)) {
                       case 'a':       bitset |= 1 << 0; break;
                       case 'b':       bitset |= 1 << 1; break;
                       case '\u00DF':  bitset |= 1 << 2; break;
                       case '\u00e4':  bitset |= 1 << 3; break;
                       default:
               }
       }
       return bitset == (1 << 4) - 1;
}
*/

> mixin(T_isPangram("DE_isPangram", "abcdefghijklmnopqrstuvwxyz", "\\u00DF\\u00e4\\u00f6\\u00dc"));

somehow this got mixed up with the earlier messed up one.. this is the correction for this line.

mixin(T_isPangram("DE_isPangram", "abcdefghijklmnopqrstuvwxyz", "\\u00DF\\u00e4\\u00f6\\u00fc"));

Am 06.02.2012, 07:38 Uhr, schrieb Era Scarecrow <rtcvb32@yahoo.com>:

>> mixin(T_isPangram("DE_isPangram", "abcdefghijklmnopqrstuvwxyz", "\\u00DF\\u00e4\\u00f6\\u00dc"));
>
> somehow this got mixed up with the earlier messed up one.. this is the correction for this line.
>
> mixin(T_isPangram("DE_isPangram", "abcdefghijklmnopqrstuvwxyz", "\\u00DF\\u00e4\\u00f6\\u00fc"));

Actually, yeah a bit Scarecrow like. For dlang or rosettacode I prefer short and easy to read code. The one example where the result is checked against 0b11_1111111_11111111_11111111 was really nice, since it expresses the bit set nature in a concise and readable fashion. But it also shows that you can put integer literals in dual notation and place _ to segment the number.

Andrei Alexandrescu wrote:

> Would you want to redo the example

No.

I do not see the usefullness of a random ten liner shown on the frontpage of a programming language.

If the goal is to attract the attention of a visitor to some feature of the language, that feature should be stated prominently and the example code should be shown on request.

At least such a goal is not obvious for this example.

-manfred