DROP ALL fails when no records have been selected (long)

[GCC Consulting]

> -----Original Message-----
> From: 
> filepro-list-bounces+gccconsulting=comcast.net at lists.celestial
> .com 
> [mailto:filepro-list-bounces+gccconsulting=comcast.net at lists.c
> elestial.com] On Behalf Of Barry Wiseman
> Sent: Wednesday, November 07, 2007 11:46 AM
> To: filepro-list at lists.celestial.com
> Subject: DROP ALL fails when no records have been selected (long)
> 
> This browse lookup problem has been driving me crazy. 
> 
> >From an input screen, my program accepts low/high value 
> ranges for each 
> >of
> several fields, all of which are subfields of a compound 
> index key in the target file.  (This is an "old" index, built 
> on one long field into which the several subkey values are assembled.)
> 
> I then run a browse lookup, and my browse processing must 
> validate that each of the subfields is within range, 
> otherwise DROP the record. I have initially assembled a pair 
> of compound keys consisting of lowest- and highest-possible 
> values in each subfield. My browse processing begins by 
> testing against these; if a compound key built from the 
> record under test does not fall between them, I issue a DROP 
> ALL BEFORE or DROP ALL AFTER as appropriate. 
> 
> The problem is that, if there are no records at all that 
> satisfy the search, the DROP ALL is performed (as running 
> with the debugger demonstrates), but has no effect, i.e., the 
> browse lookup continues to crank through all records to the 
> end of the file!
> 
> Here is a simplified piece of code that will demonstrate the 
> problem.  It can easily be adapted to any data you have 
> laying about.  In this case, it is looking for a range of 
> invoice numbers for a given account key.  Try it first with 
> values in qq and qx such that some records will fall within 
> that range; it works like a champ.  Then alter the range so 
> no matching records will be found. 
> The MSGBOX will show you the browse lookup is reading through 
> the entire file despite having performed a DROP ALL. 
> 
> ::end:
> @KEYQ:::
> ::qq(20,*)="ZUPNICK    362429999":
> ::qx(20,*)="ZUPNICK    34051    ":
> ::zz="(brw=12 xkey=X prc=BPRC fill=desc,top)":
> ::zz=zz { "[Key         Invoice]*4          *3":
> ::lookup vtr = drvtrans  k=qq   i=A -nl b=(zz):
> ::end:
> BPRC::qz=vtr(4)&vtr(3):
> ::msgbox qz:
> :qz lt qx:drop all after; end:
> :qz gt qq:drop all before; end:
> ::end:
> 
> I'm guessing the true meaning of the command is, "drop all 
> records before/after *the last record displayed in the browse 
> lookup*, not before/after the record currently being 
> considered.  Thus, if NO records are selected, DROP ALL 
> becomes a no-op.  Grrr!!
> 
> If my actual program were this trivial, certainly MLEN= would 
> be a better tool than DROP ALL, however, in my case I cannot 
> assume that the first subkey would narrow things down that much. 
> 
Barry,

Might I suggest that you break the drop into 2 checks
Test for first field value of index
:this_name le "name_field":drop all before;end
:this_name gt "name_field":drop all after;end
:starting_number lt number_field;drop all before;end
:ending_number gt number_field;drop all after;end

Since it appears that you are only testing the first 5 digits of the number
field, the use starting_number lt mid(number_field,"1","5").

This is a few more lines but should solve your problem.

Richard Kreiss
GCC Consulting