Is it necessary to use the standard function modules always?

Question

Hi All,
   Suppose i can retrieve the BUSINEES PARTNER ID from the table BUT000 with the help of 'BP_GUID' field through a select query statement. 
Now is it necessary to use a standard function module to retrieve the same data. what about the performance? which is the better solution. to implement our own code or to use the function module available.
Please someone clarify me about this issue.
Regards,
Sharry.

Answer 1

If you'd ask the question in general it's obvious that there is no good answer as it always would depend on the circumstance. However, as you've posted your question in the performance tuning forum, let's look at your question in this context first. I think it's fairly safe to say, that in almost 100% of the cases a standard function module will not be as fast as any well implemented and custom tailored code. If in doubt, just try reading the code of the function module you want to read; most likely you will find lots of unnecessary fluff in there, often they read data that you don't even need.
And as far as buffering is concerned, that can also be an issue. E.g. let's say you do a massive read of distinct business partners and you only need the data once for each partner. Obviously buffering is silly in this case and it might even slow down your program, because the implemented buffering is not necessarily well implemented. I've come across a few cases, where the buffer tables where standard tables with no efficient read access (so no sorted table or hash table) and once tons of data were buffered, it actually took a moment to look up if the record already was buffered (just to find it wasn't there yet).
Now most of the times this slowdown is not an issue as the lookup is still fast compared to the (usually) prevalent database accesses with their slower IO timing. But another factor to consider is memory consumption and why buffer data in memory (and waste it), if you can't use the buffer anyway? So when I do use standard function modules, I usually try to see what they retain in global memory, to better understand if I need to call some initialization/clearing/refresh function...
Note also that often you need to process lots of data, but don't find a standard function module for "mass access", so you need to read your object one by one. Obviously that's another good candidate for implementing your own queries, where you can process the data in packages.
So what might be the reason to use a standard function module? Well, often it's required because the logic is so complex that it would take you too much time to implement it (not to mention that we often simply don't know all the logic that's coded in there). Another advantage is that you might get independent of the data model. E.g. for IDoc interfaces SAP switched in some release the tables where the data is stored. So if you had coded your own queries to get IDoc data, they wouldn't have worked after this particular release upgrade. The standard function on the other hand, checks both the old and new tables, so by using it you would have been fine.
The one major caveat I'd give though is to always check whether the function module is released or at least documented to some degree. I.e. if it's not released, SAP won't support it and does not guarantee that the interface or logic stays the same. In my experience lots of functions I do want to use are not released (or documented), but I will still use them because they seem to be the best alternative in certain scenarios.
Cheers, harald