Request for new function "Array Scan"

[Fred Oxenby]

I have not seen request for this.
If it is already on the wish-list, please tick it once more.

I frequently need to scan arrays in 'reverse' direction to ensure that
my first hit is the 'latest insert' in the array

ARRAY SCAN REVERSE A()....
or
ARRAY SCAN A(Ubound(A)),FOR -200,......


------------------
Fred
mailto:[email protected][email protected]</A>
http://www.oxenby.se

[Borje Hagsten]

Fred, are you a mind reader? Thought about the exact same thing
only a couple of days ago. I fully support this wish..

------------------

[Florent Heyworth]

Hi guys

You could SORT DESCENDING on a TAGARRAY and then do the ARRAY
SCAN...

As in:

#COMPILE EXE

FUNCTION PBMAIN() AS LONG
    DIM i AS LONG
    DIM a(0:26) AS STRING
    DIM b(0:26) AS LONG
    
    FOR i = 1 TO 27
        a(i-1) = CHR$(i+47) + "Hello there" + CHR$(i+47)
        b(i-1) = i
    NEXT
    
    ARRAY SORT a(0), TAGARRAY b(), DESCEND

    ARRAY SCAN a(), = "JHello thereJ", TO i
    PRINT a(i-1)
    
    WAITKEY$
END FUNCTION

But I agree it would be a useful short cut...

Cheers

Florent

------------------

[Borje Hagsten]

Yes, if the array can be manipulated, but we usually want to
keep it intact, plus, in time critical code, it makes things a
lot slower, especially with large arrays. Backwards SCAN would
enable us to speed up many things quite a bit.

For purpose of quickly finding latest inserted item, one could use
a tagged array of Longs though, where you set UBOUND(tArray) to
matching array item each time one is inserted and then use,
ARRAY SCAN tArray(), = UBOUND(tArray) - 1

But if items only are added to an array, Array(UBOUND(Array) - 1)
will of course do the same thing much faster..


------------------

[Guest]

If you are trying to determine if "my first hit is the 'latest insert' in the array"
why not examine the last subscript and not do an array scan?

I know it must not be this simple but it appears to be.


------------------
Ron

[Fred Oxenby]

Florent, My need is unsorted array of UDT's with >100,000 elements
I have tried your way, but you have to create (or maintain) the
array of element-numbers and make two Array Sort on this, one before and one
after the Array Scan. Just one word -ItIsNotFastEnogh
----
There is a way to 'sort'. If you have a Udt like this:

Type MyUdt
 ElementNr    as long
 OtherElement as string 
 Date         as Double
 etc
End Type

When you add new elements you do it this way
-----
Dim Udt as MyUdt
  .... 
  NewRec& = Ubound(Array)+1
  Redim Preserve Array(NewRec&)
  ceRweN$ = Peek$(VarPtr(NewRec&),4)
  ceRweN$ = StrReverse$(ceRweN$)
  Poke$ VarPtr(Udt.ElementNr),ceRweN$
  Array(NewRec&) = Udt
     
  Array Sort Array(),From 1 To 3,DESCEND
  ....
  Now the Array element is in 'reverse' order

Just a thought
---------
Ron, I am scanning for the last inserted item matching the search-criteria,
not the last inserted item...


------------------
Fred
mailto:[email protected][email protected]</A>
http://www.oxenby.se



[This message has been edited by Fred Oxenby (edited January 08, 2001).]

[Lance Edmonds]

Great idea guys... I've passed it on to R&D.


------------------
Lance
PowerBASIC Support
mailto:[email protected][email protected]</A>

[Michael Mattias]

Ron, I am scanning for the last inserted item matching the search-criteria,
not the last inserted item...

A couple of things...

A. If you are looking for the last inserted item meeting a criteria, why bother sorting the array at all?
B. Doing a REDIM PRESERVE for *every* insert takes time; better perhaps to allocate new "block" of elements (maybe 100 at a time, using something like REDIM PRESERVE FooArray (1: UBOUND(FooArray) + 100) and keep trak of the number of valid elements.)
C. If "entry sequence" is important to you, it should be part your sort key.
D. If none of these methods floats your boat, I published a linked-list method of keeping a list of UDTs in sorted order by three keys at once. This code is public domain, available at the IMS web site. The December, 1998 issue of "BASICally Speaking" included MS-DOS code along with a detailed explanation of what I was doing in an article entitled, "Make Haste,Not Waste." The April, 1999 issue contained the migration of that code to PB/Win32 in an article entitled, "Dynamic Allocation and Linked Lists: Porting from PB/DOS to PB/CC"
Since the code is public domain it may be posted in the source code forum by anyone. The article texts are copyright; reprints are available from IMS at http://www.infoms.com, follow the links to Basically Speaking.

Bottom line: ARRAY SCAN BACKWARDS may be a neat feature, but just looking at you application I think you have a number of things to try to speed it up without waiting for a new feature.




------------------
Michael Mattias
Racine WI USA
[email protected]