Arrays in Rev (long)

[Richard Gaskin]

Troy Rollins wrote:
> 
> On Jul 12, 2004, at 10:10 PM, J. Landman Gay wrote:
> 
>> If speed doesn't matter, then base your storage methods on 
>> convenience. If speed does matter, use variables.
> 
> Alright. Noted. (Though Richard's article on stack properties suggested 
> otherwise.)

It depends on what you mean by "slow".

Yes, it takes almost six times longer to access a property than it does 
to access a local var. Run the benchmarking script below; on my modest 
single-processor G4 it gets:

Props: 62 ms
Vars:  12 ms

But step back from relative comparisons and look at absolutes: that 
script runs 10,000 iterations, which means a single property access 
takes only 0.0062 milliseconds.  In most real-world operations that's 
plenty fast enough. ;)

But then there's the question of what you might do with the values in a 
var and how you might access them.

The example in my script below uses only a single variable containing a 
single value.  This is by far the fastest use of vars, but not 
necessarily reflective of the types of things people commonly store in 
them, which may also include delimited data that must be parsed.

While using delimiters to parse data is fast, in my earlier benchmarks 
it's only faster than using array elements in one specific case: where 
you need to walk through the entire list sequentially.

In most real-world usages you're likely interested in accessing specific 
items; if you benchmark random-numbered array element access vs. getting 
a random-numbered line you'll find array accesses are about two or three 
times faster.  This is along the lines of what we might expect, as 
arrays are indexed for quick lookups while "get line 100 of tMyChunk" 
requires that the engine count 100 line delimiters to find your data.

Also, keep in mind that you can convert arrays into chunks and back 
again with the split and combine commands.  So if most of your code 
accesses single elements you can work with the data in an array, but if 
you have an operation in which you need to access all elements 
sequentially you can trim about 15-20% of execution time by converting 
it to a block of text for that routine.

Moral:  with Rev, "slow" is very relative.  Perhaps the only truly slow 
storage is field access, and even those benchmark faster in Rev that 
most other xTalks (though much slower than vars or props).

In brief, use whichever you're comfortable with and chances are your 
users will never know the difference. :)

-- 
  Richard Gaskin
  Fourth World Media Corporation
  ___________________________________________________
  Rev tools and more:  http://www.fourthworld.com/rev


-------------------
on mouseUp
   put 10000 into N
   --
   put the millisecs into s
   repeat n
     set the uTest of this stack to "hello world"
     get the uTest of this stack
   end repeat
   put the millisecs - s into s1
   --
   put the millisecs into s
   repeat n
     put "Hello World" into tMyVar
     get tMyVar
   end repeat
   put the millisecs - s into s2
   --
   put "Props: "&s1 &cr& "Vars: "& s2
end mouseUp