Bitset with data type based upon value ranges

[Aaron Gray]

Hi,
I am looking for some template code to generate a type based upon a constant within value ranges, for specialization of a bitset class.

0...7 - unsigned char
8..15 - unsigned short
16...31 - unsigned int
32...63 - unsigned long
64...127 - unsigned long long
128... - unsigned long long [] - array

```
#include <cstddef>

template <size_t BITS>
class bitset {
public:
....
WORDTYPE data;
....
};
```

I am looking for solutions both with and without concepts for comparison, if its possible without and its simpler I will use that solution unless there is some real advantages to using concepts.

Many thanks in advance,

Aaron

[Aaron Gray]

On Saturday, 7 January 2023 at 00:14:35 UTC, Aaron Gray wrote:
> Hi,
> I am looking for some template code to generate a type based upon a constant within value ranges, for specialization of a bitset class.
>
> 0...7 - unsigned char
> 8..15 - unsigned short
> 16...31 - unsigned int
> 32...63 - unsigned long
> 64...127 - unsigned long long
> 128... - unsigned long long [] - array
>
> ```
> #include <cstddef>
>
> template <size_t BITS>
> class bitset {
> public:
> ....
> WORDTYPE data;
> ....
> };
> ```

ChatGTP has given me the following :-
```
#include <type_traits>

template <size_t BITS>
class bitset {
public:

static_assert(BITS > 0, "Number of bits must be positive");

using WORDTYPE = typename std::conditional<BITS <= 8, unsigned char,
typename std::conditional<BITS <= 15, unsigned short,
typename std::conditional<BITS <= 31, unsigned int,
typename std::conditional<BITS <= 63, unsigned long,
typename std::conditional<BITS <= 127, unsigned long long,
unsigned long long[]>::type>::type>::type>::type;

WORDTYPE data;
// ...
};
```

A good start, but withh two bugs, a missing extra '>::type' and the array really wants to be a whole new template specialization with WORDTYPE :-

unsigned long long[(BITS+127)/128]

> I am looking for solutions both with and without concepts for comparison, if its possible without and its simpler I will use that solution unless there is some real advantages to using concepts.

I am thinking concepts may be necessary ?

If you would like to help on the journey it would be well appreciated :)

Aaron

[Mut...@dastardlyhq.com]

Do you own coursework.

[Öö Tiib]

On Saturday, 7 January 2023 at 02:14:35 UTC+2, aaron...@gmail.com wrote:
> Hi,
> I am looking for some template code to generate a type based upon a constant within value ranges, for specialization of a bitset class.
>

But it is trivial with std::conditional in <type_traits>. Also the example
in <https://en.cppreference.com/w/cpp/types/conditional> is almost what
you asked for. So what is the problem?

Otherwise you can freely get innumerable implementations of bitset.
The std::bitset, QBitArray, boost::dynamic_bitset, std::vector<bool> are more popular.
So you should be swimming over your head lake of code ... not be asking for more
of it.

[Aaron Gray]

On Saturday, 7 January 2023 at 16:49:46 UTC, Öö Tiib wrote:

I am trying to put together a very efficient implementation to deal with fixed number of bits for dealing with lookahead sets for closure FIRST set guards for recursive decent parsers and a recursive descent parser generator. So want to use my own specific implmentation.

Aaron

[Aaron Gray]

No I am an open source software engineer.

Aaron

[Öö Tiib]

The std::conditional runs compile time so you can not get more efficient than
that.

[Chris M. Thomasson]

On 1/6/2023 5:48 PM, Aaron Gray wrote:
> On Saturday, 7 January 2023 at 00:14:35 UTC, Aaron Gray wrote:
>> Hi,
>> I am looking for some template code to generate a type based upon a constant within value ranges, for specialization of a bitset class.
>>
>> 0...7 - unsigned char
>> 8..15 - unsigned short
>> 16...31 - unsigned int
>> 32...63 - unsigned long
>> 64...127 - unsigned long long
>> 128... - unsigned long long [] - array
>>
>> ```
>> #include <cstddef>
>>
>> template <size_t BITS>
>> class bitset {
>> public:
>> ....
>> WORDTYPE data;
>> ....
>> };
>> ```
>
> ChatGTP has given me the following :-

[...]

Ouch. ChatGTP seems to train itself when somebody has to correct its
generated code that it likely lifted from somebody else. Also, is the
correction correct itself?

[Aaron Gray]

yep, its the types that matter :)

Aaron

[Aaron Gray]

Yep, big ouch, it stung me, and wasted time.

[Mut...@dastardlyhq.com]

What did you expect? ANNs are at their core simply statistical algorithms.
You'd probably get a similar result with a hidden markov model if it was given
enough input though admittedly it couldn't parse your question.