Re: Variable length fields

[d.bocian@xxxxxxxxxxx]

> David,
> 
> Did you by any chance reverse the figures for cases 1 and 2?  It makes no
> sense that the file is larger for the allocated length case _unless_ a
> substantial number of the fields had content that exceeded the allocated
> length.  Perhaps that was the case in your tests.
> 
> As has been noted earlier, VARLEN will only save storage when a substantial
> proportion of the records have content _within_ the allocated length.
> 
> Jon Paris
> Partner400
> www.Partner400.com

No, the data I wrote was accurate.  Each file has eight fields:
VLFLD1        10A
VLFLD2        20A
VLFLD3        30A
VLFLD4        40A
VLFLD5        50A
VLFLD6        60A
VLFLD7        70A
VLFLD8        80A

The first file has VARLEN specified, but no allocation size.  The second file 
has VARLEN(10) on each field.

The same 10,000 records are in each file.  The first 9900 records have the data 
'FLD1', 'FLD2', 'FLD3', 'FLD4', etc. respectively.  So realistically, I only 
fill four bytes of data in fields that range up to 80 bytes in size.

The final 100 records are filled to their maximum size; that is, with data of 
'1111111111', '22222222222222222222', '333333333333333333333333333333', etc. 
respectively.

I just now repeated the test, and got the same results that I posted previously.

So, my interpretation of these results is that if you going to specify an 
allocation size on the VARLEN parameter, it should be as accurate as possible.  
In my case, I allocated 10 bytes, but then only used 4 for the majority of 
records.  The size of the file with VARLEN(10) suffered as a result.

I hope I explained that well enough.

Thanks.

David Bocian