Performance Field Symbols Vs Work Area

Question

Hi,
I want to some operations (Validations and Calculations) form each entry of internal table.
Which is better way to access the data in internal table either using field symbol or using work area?
I heard that using filed symbols is performance issue. May I know in which cases we should not use field symbols?
Regards,
Kumar

Answer 1

I do not think, the runtime creates a pointer to each line of the itab.
The developer creates ONE fieldsymbol ( for a loop for example ) --> a pointer, which is iteratively assigned, pointing to only one line at one time.
With the usage itself it  seems to offer a benefit, which isn't really one, If You only read data.
People might say, You can omit a clear, if working with fieldsymbols.
But You have to add a check for sy-subrc, or You will suffer from a dump, which usually never happens, if working with work areas the same way, like fieldsymbols.
The real benefit only comes into play, if You want to modify the data in the internal table, because You are already acting on the line of the table (pointer to record n ), what allows You, to omit any modify to transfer changed values back to the internal table.
In terms of performance I must see it technically and I hope, someone can confirm/reject my statement, so that I (if I was wrong) could keep that in mind.
A fieldsymbol (lets talk about them in technical way, as a pointer) is in fact a variable, which holds an address of another variable( which itself also can hold an address and so on).
Pointers are either 4 byte or 8 byte. So, it does not matter, how much columns Your internal table has, technically the fieldsymbol is either 4 byte big or 8 byte big. And exactly this amount is used in the ram of the application server.
In fact pointers even do not have a type, as usual pointers are always integer.
A workarea is a linear, continous allocation of memory space, exacly matching the data-structure, which is used for declaration. So it is obvious, that this workarea uses much more memory of the ram, as long their elements exceed 4 / 8 bytes.
Still then it is not more performant, because changed data of a workarea has to be transferred to the internal table, (mostly, if one wants to modify), but with field-symbols this step is removed.
So in fact there are two technical reasons for using field-symbols.
One is, You save memory-space and the second is, You save the transferring.
In an even deeper way, one can say, that , by design, a cpu with its address bus is just intended to understand addresses best ( directly after 0 and 1 and hex ).