Re: Tables with VARCHAR

IBM recommends allocate large enough to handle 90-95% of the data
https://www.ibm.com/support/knowledgecenter/ssw_ibm_i_72/rzajq/vardatatypes.htm

I believe that one 26 byte pointer is include for every record, regardless
of how many VARxxx/LOBxxx objects are present. This is inferred from the
MAX ROW SIZE documentation, "The maximum row buffer size is 32766 or, if a
VARCHAR, VARGRAPHIC, VARBINARY, LOB, or XML column is specified, 32740."

Yes, VARCHAR2(80) will always be full size in the record buffer presented
to your program.

Charles



On Thu, Apr 5, 2018 at 3:04 AM, Jevgeni Astanovski <jevgeniast@xxxxxxxxx>
wrote:

Thanks, Jon and Charles.

Suspected that there is no simple answer and I need to make content
analysis.
So first of all I've identified the 3 fields as potential candidates for
switching to VARCHAR.
They are of 80, 255 and 140 length.
Then I applied a query that Charles advised to smaller pieces of data -
just one month instead of 12 years.
Selected 2 one-month chunks, in the beginning and in the end of the period,
as data usage could change also.
Got numbers, sorted them and now I can easily find, that, for example for
CHAR(80) field, in 99% of records actual field length is 30 or less. This
should make sense to switch to VARCHAR2(80) ALLOCATE(30) I suppose?
By the way - are there any "rules of thumb" for this %? Should it be 80% or
99% for example?

I thought that I've indicated it in my question, however probably forgot -
to write that I'm using record-level access (_Rreadx) functions.
Do I understand correctly that for these functions ALLOCATE number has no
effect? I mean that if record has one field VARCHAR2(80), its size will be
ALWAYS 82 bytes in the returned buffer?

And I have not fully understood if this 16-bytes pointer to "overflow" area
is allocated to every record or only to those, where this "overflow"
happened in reality...
Will now make some experiments with the information obtained.

Thanks again,
J.A.




On Wed, Apr 4, 2018 at 7:12 PM, Charles Wilt <charles.wilt@xxxxxxxxx>
wrote:

Yep...it depends...

what's "pretty large char fields"? The entire record is only 650
bytes...so they can't be all that large to my way of thinking..

What's "almost empty"?

run a
select length(trim(MYFLD)), count(*) as cnt
from myfile
group by length(trim(MYFLD))
order by 2 desc

If the field is CHAR(100) and most of the records have 20 chars or

less...

Then you might be better off with VARCHAR(100) ALLOCATE (20)

That'd save 78 bytes for every record with less than 20 chars.

However, one additional point to consider. Are you using SQL or record
level access (RLA)?

With RLA, the system will do two disk I/Os for every record with more

than

20 chars. With SQL, those 2 I/Os will only be done if you ask for MYFLD

to

be returned.

(Thus the reason ALLOCATE(0) is not recommended, unless most records are
empty)

Charles


On Wed, Apr 4, 2018 at 8:30 AM, Jon Paris <jon.paris@xxxxxxxxxxxxxx>
wrote:

As is often the case - it depends.

When you define a field as Varchar you can specify both a "normal"

length

and a maximum length. Only the "normal" part is held with the main

record

and therefore if most of your records are within that length then the

file

size will reduce.

There is however an overhead for linking to the extended space - can't
recall how long it is 20 bytes is ringing a faint bell. This topic has
been discussed in the past on the Midrange list - I'd check it for you

but

have to run out.


Jon Paris

www.partner400.com
www.SystemiDeveloper.com

On Apr 4, 2018, at 10:05 AM, Jevgeni Astanovski <

jevgeniast@xxxxxxxxx>

wrote:

Hi,
Does anyone here know if switching from CHAR to VARCHAR gives any

positive

effect in terms of table size?
Explain the situation briefly.
I've got an application written on ILE/C that works with rather big

table

(historical data since 1996). Currently the table size is 260M

records

and

PF size 170GB.
The record size is app. 650 bytes and it has some pretty large char

fields,

that can be either empty or almost empty :-)

Technically I can make a new table with VARCHARs and rewrite my

program

to

use VARCHAR instead of CHAR, but shall it reduce the table size?

Somehow I suspect that it will only increase the table...

TIA and regards,
Jevgeni.
--
This is the Bare Metal Programming IBM i (AS/400 and iSeries)

(C400-L)

mailing list

To post a message email: C400-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: https://lists.midrange.com/mailman/listinfo/c400-l
or email: C400-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at https://archive.midrange.com/c400-l.

--
This is the Bare Metal Programming IBM i (AS/400 and iSeries) (C400-L)
mailing list
To post a message email: C400-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: https://lists.midrange.com/mailman/listinfo/c400-l
or email: C400-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at https://archive.midrange.com/c400-l.

--
This is the Bare Metal Programming IBM i (AS/400 and iSeries) (C400-L)
mailing list
To post a message email: C400-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: https://lists.midrange.com/mailman/listinfo/c400-l
or email: C400-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at https://archive.midrange.com/c400-l.

--
This is the Bare Metal Programming IBM i (AS/400 and iSeries) (C400-L)
mailing list
To post a message email: C400-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: https://lists.midrange.com/mailman/listinfo/c400-l
or email: C400-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at https://archive.midrange.com/c400-l.