Re: Bsearch -- RPG400-L


Hi Steve,

[SNIP]

    C                   Eval      Elem = (( P_match - Template)
    C                                    / %Size(MyTemplate)) + 1

[SNIP]

Won't "Template" in the above code always be *NULL? You're not searchingTemplate, you're searching Items. Wouldn't the following make a lot moresense?


                     eval  elem = (( p_match - %addr(Items) )
                                / %size(myTemplate) ) + 1

Think about what the code is trying to accomplish. It wants to know which(numeric) element was found, but all it has received from bsearch() is thespot in memory where the matching element is stored.


So how can it derive the element number from that?

Think about how the array is stored. If Items(1) (the first element ofthe Items array) is stored in memory at address 12345, and each element is25 bytes long, then Items(2) is stored at 12370. (12345 + 25).

Items(3) is at 12395, and so on.

So, if bsearch() returns the address of the 3rd element, it'll return12395... that'll be the number stored in the P_match variable. To figureout the element number, you can subtract the start (12345) from thep_Match variable to get the number of bytes between them. (in thisexample, 50.) Then you can divide that by the size of each element tofigure out how many elements that number of bytes represents. (50 / 25 =2). Then add 1 to get the element number (2 + 1 = 3, the 3rd element)

It doesn't make sense to subtract "TEMPLATE" from p_Match because youweren't searching TEMPLATE, you were searching ITEMS. What you reallywanted to do was subtract the starting address of the ITEMS array from theaddress of the elemnt that it found, divide by the size of each element,and then add 1.


and that's how you come up with this equasion:

                     eval  elem = (( p_match - %addr(Items) )
                                / %size(myTemplate) ) + 1

Since "Items" is a direct copy of "myTemplate" they should be the samesize. So the following code is equivalent to the above code:


                     eval  elem = (( p_match - %addr(Items) )
                                / %size(Items) ) + 1

So why use %size(myTemplate)?

Because, I always forgot how %size() works when you use it with an array.To my way of thinking, %size(Items) should be the whole array, whereas%size(Items(2)) should give you the size of one element. After all,%addr(Items) gives you the address of the whole array, whereas%addr(Items(2)) gives you the address of the second element, why should%size() work differently? I don't know, but it does.

In fact, %size(Items) is the size of ONE ELEMENT of the array, and youneed to use %size(Items:*ALL) to get the size of the whole array. I thinkthat's confusing.

Consequently, I'll typically use %size(myTemplate) since myTemplate isn'tan array, and I don't have to remember how %size() works!