1. Rob, How?
Hello,
Another question i have, for my own programming purposes, how did you
implement the sequence idea in C/C++?
i've attempted my own linked list class and a template class but none are
stable enough to implement in a large program.
Could you help me?
Ian.
________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com
2. Re: Rob, How?
- Posted by Robert Craig <rds at ATTCANADA.NET>
Aug 18, 2000
-
Last edited Aug 19, 2000
Ian writes:
> Another question i have, for my own programming purposes,
> how did you implement the sequence idea in C/C++?
> i've attempted my own linked list class and a template class
> but none are stable enough to implement in a large program.
If you look at the sieve C code that I released yesterday
you'll get some idea.
I originally (pre-1.0) implemented sequences as linked lists.
The performance on random subscript operations was pretty
poor, so I made them doubly-linked, and I kept a pointer in each
sequence to the most recently referenced element.
That helped a lot when you looped over a consecutive
series of elements (as you often do). If you accessed
x[25] followed by x[26], x[27], ... you could quickly locate
the next element without starting at the beginning of the list
and stepping through 25 elements. Eventually though
I just implemented sequences as arrays. That made
subscripting really fast, but things like append() became
slower and had to be dealt with in a smarter way.
More importantly, getting rid of all those links
saved a ton of memory.
Euphoria is implemented in plain C.
I don't use any C++ features.
Regards,
Rob Craig
Rapid Deployment Software
http://www.RapidEuphoria.com
3. Re: Rob, How?
No Solution wrote:
> Another question i have, for my own programming
> purposes, how did you implement the sequence idea
> in C/C++?
In the odd chance that Robert doesn't want to give away any trade secrets,
you might want to take a look at Peuphoria, Pete's implementation of
Euphoria. You can find it at:
http://www.harborside.com/~xseal/euphoria/
-- David Cuny
4. Re: Rob, How?
>No Solution wrote:
>
> > Another question i have, for my own programming
> > purposes, how did you implement the sequence idea
> > in C/C++?
>
>In the odd chance that Robert doesn't want to give away any trade secrets,
>you might want to take a look at Peuphoria, Pete's implementation of
>Euphoria. You can find it at:
>
> http://www.harborside.com/~xseal/euphoria/
>
>-- David Cuny
Ah the hell.
Sequence are easy as crap to implement!
Sequences are just arrays!
If you're not so memory-hungry, but wanna have a fast implementation of
sequences (as fast or faster than Rob's) then do the following;
#define Slice 1
#define Integer 2
#define Double 3
#define Float 4
typedef struct
{
int * data;
char * mask;
} sequence;
// now to initialise a sequence, do this
sequence myseq;
myseq.data = (int*) malloc(sizeof(int)*10);
myseq.mask = (char*) malloc(10);
float f;
int i;
double d;
f = 50.1;
i = 70;
d = 500964.485;
myseq.data[1] = &f;
myseq.data[2] = &i;
myseq.data[3] = &d;
myseq.mask[1] = Float;
myseq.mask[2] = Integer;
myseq.mask[3] = Double.
// now you can extract the data in a sequence by getting the value
// at the integer address, and you can use the character mask to
// eventualy check the type of a given element. A Slice is just a
// pointer to another sequence, wich serves as a sub-sequence.
Above method sucks white ass sometimes. Allthough using realloc() you can
resize any part of the sequence pretty fast.
This method might not be that practical at times, but it's a hell of a lot
better than some of the ones I saw. It's fast, memory efficient, flexible,
and small.
The method I use is one simmiliar to the one above, except that I can
initialise whole sequences in one time without fetching the address of a
variable i'm puttin in there.
Sequences are no big secret, their a big joke for C programmers.
A C programmer can't use them in their code because they're not practical
for them, that's why you don't see C programs with sequences in them.
Atoms, though, are another story.
They *are* hard to implement because there is no way of doing fast math with
them. Everytime an Atom operation occurs the Atom must be checked to see if
it did not overflow. Atoms can be 4 or 8 bit integers or floats. And that's
what makes them a pain in the ass.
Mike The Spike
________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com
