Cross Reference index problem (AREV Specific)
At 18 FEB 1999 10:41:14PM M. Braun at Blades wrote:
We are using Arev Ver 3.1, Novell 4.10 with the proper NLM. In a field called E_CREF we combine values from maybe six different fields of the same record using @ANS=@record:@VM: @record, etc. For example, we combine employee first name, last name, social security number, and date of hire. We then convert all separators /,_,-,@FM,@TM, etc. to spaces " ". Then we create a cross reference index on this E_CREF field. This way, a user can find an employee based on first name, last name, date of hire, or social security number by searching this one field. (This is used with many different tables, and it is not feasible to search each field separately.)
The problem is that when searching for "FRANKLYN" only "FRANK" and "FRANKLIN" are returned. When searching for "FRANK" I do get the correct number of hits, "FRANKLYN" not included in the hits. I have copied a subset (about 200 records) of the table to a temporary table. After indexing the E_CREF field with a cross reference index every thing works fine. It is only with the original much larger table that indexing does not work even after deleting and recreating the index field.
Any suggestions? Thanks in advanced.
At 19 FEB 1999 01:46PM Victor Engel wrote:
It looks to me like the field width of your field is too small. If this is the case, you need to expand it to the size of the largest item being indexed and then re-index the file.
At 24 FEB 1999 10:18PM akaplan@sprezzatura.com - [url=http://www.sprezzatura.com]Sprezzatura, Inc.[/url] wrote:
Do you mean the length of the key for IDX_SETS?
akaplan@sprezzatura.com
At 25 FEB 1999 12:32AM Victor Engel wrote:
The field being indexed was a multivalued symbolic consisting of a composite of various other fields. I was saying that the field width of this symbolic should be the max of the field widths of the other fields, which, presumably are each sized appropriately.
At 25 FEB 1999 02:40PM akaplan@sprezzatura.com - [url=http://www.sprezzatura.com]Sprezzatura, Inc.[/url] wrote:
Don't see how this makes a difference, except for display purposes though. All I know that relies on length info is the key length for IDX_SETS, though it would make sense the data should be sized as well. Never knew it to be gospel though.
akaplan@sprezzatura.com
At 26 FEB 1999 09:41AM Victor Engel wrote:
Maybe I'm wrong in my assumptions, but I thought that the sort portion of the build process used a flat version of the field for efficiency. Is this incorrect?