OpenEuphoria

2. Re: Your thoughts on two item id methods?

• Posted by Michael Nelson <mikestar13 at sbcglobal.net> Oct 18, 2005
• 702 views

Al Getz wrote:

>Im looking at something that needs to tell the difference between two
>different classes of elements in a loop.  The two methods presented
>here both work, but they work somewhat different.  I'd like to hear
>any thoughts anyone has on these two methods, as to which one if any
>seems better and why...
>
>METHOD #1
>
>Atoms are of one class, while sequences are of the other class.
>Here, we know items of class 1 are atoms, while class 2 items are sequences.
>This way we can tell the difference between class 1 items and class 2 items...
>
>itemlist={ 1,2,3,{8,9},4,{6,7} } --atoms and sequences
>for k=1 to length(itemlist) do
>  item=itemlist[k]--atom or seq
>  if atom(item) then
>    DoClass1Stuff(item)
>  else
>    DoClass2Stuff(item)
>  end if
>end for
>
>
>METHOD #2
>
>Atoms will be in a sequence where the first element is zero, while
>the other class will always have a '1' as first element with the real
>data in elements 2 and above.  Thus, class 1 elements start with a zero
>and class 2 elements start with a 1...
>
>itemlist={ {0,1},{0,2},{0,3},{1,8,9},{0,4},{1,6,7} } --all sequences
>
>for k=1 to length(itemlist) do
>  item=itemlist[k]--always a sequence
>  if item[1] then
>    DoClass2Stuff(item)
>  else
>    DoClass1Stuff(item)
>  end if
>end for
>
>I'm not that worried about the increase in memory space taken by
>the second method.
>In each itemlist there will be approx the same number of class 1 items
>as there are class 2 items.
>

Method 1 uses less memory, which you don't care about, but it will also 
run faster. The atom() type check is extremely fast, probably in the 
ballpark of  an "if variable then" construct, though likely just a bit 
slower. But it is sure to be faster than "if variable[subscript] then". 
Also, "DoClassStuff(item)" may need a slice in method 2 than method 1 
won't require.

However, this is only workable if you will only need to distinguish atom 
and sequence elements--it does not generalize. Say at some point you 
need to quickly distinguish ordinary sequences from character 
strings--now you could prefix the strings with a 2, but method 1 would 
require you to check the whole element to see if it is a string.

-- Mike Nelson

4. Re: Your thoughts on two item id methods?

• Posted by Bernie Ryan <xotron at bluefrog.com> Oct 18, 2005
• 687 views

Bernie

My files in archive:
w32engin.ew mixedlib.e eu_engin.e win32eru.exw

Can be downloaded here:
http://www.rapideuphoria.com/cgi-bin/asearch.exu?dos=on&win=on&lnx=on&gen=on&keywords=bernie+ryan


Al:
  Or this:

itemlist={ 1,2,3,{8,9},4,{6,7} } --atoms and sequences

for k=1 to length(itemlist) do

  if atom(itemlist[k]) then DoClass1Stuff(itemlist[k]) end if
  DoClass2Stuff(itemlist[k])

end for

6. Re: Your thoughts on two item id methods?

• Posted by Al Getz <Xaxo at aol.com> Oct 19, 2005
• 716 views

Michael Nelson wrote:
> 
> 
> Al Getz wrote:
> 
> >Im looking at something that needs to tell the difference between two
> >different classes of elements in a loop.  The two methods presented
> >here both work, but they work somewhat different.  I'd like to hear
> >any thoughts anyone has on these two methods, as to which one if any
> >seems better and why...
> >
> >METHOD #1
> >
> >Atoms are of one class, while sequences are of the other class.
> >Here, we know items of class 1 are atoms, while class 2 items are sequences.
> >This way we can tell the difference between class 1 items and class 2
> >items...
> >
> >itemlist={ 1,2,3,{8,9},4,{6,7} } --atoms and sequences
> >for k=1 to length(itemlist) do
> >  item=itemlist[k]--atom or seq
> >  if atom(item) then
> >    DoClass1Stuff(item)
> >  else
> >    DoClass2Stuff(item)
> >  end if
> >end for
> >
> >
> >METHOD #2
> >
> >Atoms will be in a sequence where the first element is zero, while
> >the other class will always have a '1' as first element with the real
> >data in elements 2 and above.  Thus, class 1 elements start with a zero
> >and class 2 elements start with a 1...
> >
> >itemlist={ {0,1},{0,2},{0,3},{1,8,9},{0,4},{1,6,7} } --all sequences
> >
> >for k=1 to length(itemlist) do
> >  item=itemlist[k]--always a sequence
> >  if item[1] then
> >    DoClass2Stuff(item)
> >  else
> >    DoClass1Stuff(item)
> >  end if
> >end for
> >
> >I'm not that worried about the increase in memory space taken by
> >the second method.
> >In each itemlist there will be approx the same number of class 1 items
> >as there are class 2 items.
> >
> 
> Method 1 uses less memory, which you don't care about, but it will also 
> run faster. The atom() type check is extremely fast, probably in the 
> ballpark of  an "if variable then" construct, though likely just a bit 
> slower. But it is sure to be faster than "if variable[subscript] then". 
> Also, "DoClassStuff(item)" may need a slice in method 2 than method 1 
> won't require.
> 
> However, this is only workable if you will only need to distinguish atom 
> and sequence elements--it does not generalize. Say at some point you 
> need to quickly distinguish ordinary sequences from character 
> strings--now you could prefix the strings with a 2, but method 1 would 
> require you to check the whole element to see if it is a string.
> 
> -- Mike Nelson
> 
> 

Hi Mike,

I think i see what you're saying here...that method 2 is more
easily extensible than method 1, although method 1 is a little faster.
I guess this requires some thought: go with something a little faster
or with something that will be easier to update in the future perhaps.

Once the code is written there will be quite a few routines that depend
on the data structure so it's not an easy decision, but going with the
more extensible is sounding good right now just in case the program needs
that kind of update.  With method 1 atoms are limited in what they
can be too, which might not be that good of an idea.



Take care,
Al


Take care,
Al

And, good luck with your Euphoria programming!

My bumper sticker: "I brake for LED's"

8. Re: Your thoughts on two item id methods?

• Posted by Jason Gade <jaygade at yahoo.com> Oct 19, 2005
• 725 views

Al Getz wrote:
> 
> Hello there,
> 
> 
> Im looking at something that needs to tell the difference between two
> different classes of elements in a loop.  The two methods presented
> here both work, but they work somewhat different.  I'd like to hear
> any thoughts anyone has on these two methods, as to which one if any
> seems better and why...
> 
> METHOD #1
> 
> Atoms are of one class, while sequences are of the other class.
> Here, we know items of class 1 are atoms, while class 2 items are sequences.
> This way we can tell the difference between class 1 items and class 2 items...
> 
> itemlist={ 1,2,3,{8,9},4,{6,7} } --atoms and sequences
> for k=1 to length(itemlist) do
>   item=itemlist[k]--atom or seq
>   if atom(item) then
>     DoClass1Stuff(item)
>   else
>     DoClass2Stuff(item)
>   end if
> end for
> 
> 
> METHOD #2
> 
> Atoms will be in a sequence where the first element is zero, while
> the other class will always have a '1' as first element with the real
> data in elements 2 and above.  Thus, class 1 elements start with a zero
> and class 2 elements start with a 1...
> 
> itemlist={ {0,1},{0,2},{0,3},{1,8,9},{0,4},{1,6,7} } --all sequences
> 
> for k=1 to length(itemlist) do
>   item=itemlist[k]--always a sequence
>   if item[1] then
>     DoClass2Stuff(item)
>   else
>     DoClass1Stuff(item)
>   end if
> end for
> 
> I'm not that worried about the increase in memory space taken by
> the second method.
> In each itemlist there will be approx the same number of class 1 items
> as there are class 2 items.
> 
> Thanks for your time and thoughts...
> 
> 
> Take care,
> Al
> 
> And, good luck with your Euphoria programming!
> 
> My bumper sticker: "I brake for LED's"
> 

I would use the second method, because it is more extensible as has been pointed
out.

One thing that I noticed while thinking about strings and structures, is that
for any sequence except the empty sequence, you can actually acess two known
members: sequence[1] and sequence[$]. I was thinking to add type-data to
sequence[$] in sequences representing strings or structures. That way normal
elements can be accessed with sequence[1..$] in a more natural way.

Or embed a routine id as your marker value and for each loop iteration do 

dc1 = routine_id("DoClass1Stuff")
dc2 = routine_id("DoClass2Stuff")
itemlist={ {dc1,1},{dc1,2},{dc1,3},{dc2,8,9},{dc1,4},{dc2,6,7} } --all sequences

for k=1 to length(itemlist) do
  item=itemlist[k]--always a sequence
  call_proc(item[1], {item})  
end for

j.