I recently encountered a String Space Corrupt problem in a database sorting
program using QB45. I found a way around the problem, but I'm still looking for
answers and/or comments.

This is what I did and my solution:

First I started QB with the /ah option. This allows large arrays using far
pointers.
Here is some psudo code:

qty = 10000
DIM array(qty) as STRING * 16
for a = 1 to qty
    array(a) = ( info to be sorted along with the value of a )
next
do
    if condition exists then
         SWAP array(a), array(b)
    end if
loop until fully sorted
OPEN "file" for output as #1
for c = 1 to qty    
    print #1, value of a from  array(c)
next
close #1

As soon as it hit the SWAP command it either crashed the computer or a gave a
string space corrupt message and dumped QB45. I did the same thing on 2
computers. It worked better with smaller arrays. It also helped if I inserted
dummy delay loops. These solutions were unacceptable. It seems that QB45 can't
shift data in and out of array elements in large arrays very fast without
getting caught up somehow.
To solve the problem I had the program sort pointers instead of array elements.
It workes great! All I need are the sorted pointers. Here is an idea of what I
did:

qty = 10000
DIM array(qty) as STRING * 16
DIM ptr(qty)
for a = 1 to qty
    array(a) = ( info to be sorted)
    ptr(a) = a
next
do
    if condition comparing array(ptr(a))   array(ptr(b))    exists then
         SWAP ptr(a), ptr(b)
    end if
loop until fully sorted
OPEN "file" for output as #1
for c = 1 to qty    
    print #1,ptr(c)
next
close #1

Anyone ever have a similar problem? Any other solutions/comments?

Bob Radil

So where's the '$DYNAMIC metacommand, that array is Static and
limted to 64 kb.

<snip>

<snip>

  array(0) = array(a)
  array(a) = array(b)
  array(b) = array(0)

TTfn,

Craig___

I'm not sure if this is pertinent, but I had a similar problem.  The
use of the /ah allows "dynamic" arrays to be larger and up to 64K so I
had to use the "REDIM" instead of the "DIM" to make sure that my
arrays were dynamic.  Also, I had to use the "REM $DYNAMIC"
metacommand.

Give it a try, it might help.

Thanks for your input,

Bob

If you are sure that all the arrays are dynamic then the "String Space
Corrupt" probably refers to normal strings.  The swap command is
setting up a new temporary string to do the swap.  Try putting the FRE
command at strategic places in your program to check out what your
normal strings are doing.  It may not be the arrays.

If you are getting the corruption in the IDE and not in the compiled
program, you might have to remove some of your REM statements.  I
found that they will use some of your string space.

Barry

TTfn,

Craig___

Let me just add that I had no problem DIMming the array to 160K. There are 2
ways to do this. Both require the  /ah option. One way is to use the "$DYNAMIC
meta command. The other, which I used is to use a variable in the DIM command.
This creates an array that uses far pointers and goes outside of the BASIC data
segment.

Example:
qty = 10000
DIM array$(qty) as string * 16

This won't work:
DIM array$(10000)as string * 16

The problem was sorting the elements of the array.
I don't recall DIM or REDIM making a differance, but I can try REDIM again. It
still seems to be a shortcoming of QB45. If it is then others should have the
same problem.

Upon giving it more thought, my solution of SWAPping 16 bit integers used to
point to the array elements should be much faster than SWAPping 16 byte
strings.

Thanks for the comments,
Bob Radil