Just ran into something that seems impossible (but it's on UV 9.6):

CT TFILE X ( ] = mv )
X
001 1]2
002 1]2]3

CT DICT TFILE ACORR
002 0
008 A;1:' ':2

LIST TFILE 'X' ACORR
TFILE......................... XX........

XX 1 1
2 2
2 3

Why would '2' repeat on line 3 when the A-corr ref is '1' and not '1R'
as on Pick? Is there an env variable that controls this? Any help
appreciated, including the address of the uv list!

Chandru Murthi

I haven't done UV in a while, but I would think...

Since the AMC is set to 0, it only does this conversion once based soley on
the correlative code. Since it says to take to contents of attr 1 and cat
to attr 2, and since there's no particular relationship sturcture for those
two attributes, it just repeats the last one it found in attr 1 with
whatevers left in attr 2.

Try

dict tfile acorr
002 1
008 A;1:' ':2


Mark Brown

"cmurthi" <xyzcmurthi (AT) quest (DOT) with.a.w.net> wrote

Don't think Pick (D3, mvbase etc) work this way? Because this was a
listing moved from D3 and it worked fine there (ie you would get " 3" on
line 3. IMSR it could never be otherwise because the default on DIcts is
no association, and mv's line up well unless you do a print limiter.

Chandru

Mark Brown wrote:

I'm sure you're right (i just tested it). You used to have to put an R in
the correlative to get it to repeat a single value against an array of
values.

What do you get if you change the data to have only 1 value in attribute 1;
does it repeat with every value in attr2?

Mark


"cmurthi" <xyzcmurthi (AT) quest (DOT) with.a.w.net> wrote

B 1 1
1 2
1 3

!#%&^^&#@ attr 1 of 'B' is single valued.

Chandru

Mark Brown wrote:

That's kind of what I thought you'd find. It is apparently repeating the
last value found with all subsequent values with that construct. What
happens if you change the attr # from 0 to 1?

Just looking for hints and consistancies.

Mark


"cmurthi" <xyzcmurthi (AT) quest (DOT) with.a.w.net> wrote

"Mark Brown" <mbrown (AT) drexelmgt (DOT) com> wrote in
news:rvljb.16048$ZH4.8171 (AT) twister (DOT) socal.rr.com:

And I wonder what would happen if:

CT TFILE X ( ] = mv )
X
001 1]2]
002 1]2]3

Mark T: Yes, of course if I put a null at the end it will work. that's
the solution we're trying to avoid...most good tools strip trailing vm's
before writing the item, and we're trying to not put 'em back.

Mark B wrote:>>That's kind of what I thought you'd find. It is
apparently repeating the last value found with all subsequent values
with that construct. What happens if you change the attr # from 0 to 1?

Nothing if 0->1 or 0->n. And that's not what I expect, tho' it's
consistent! The A correlative *must* use null when it runs out of mv's;
if it doesn't you'd *have* to put matching nulls or zeros in all related
fields. I'm absolutely sure (but then my absolutes are weakening with
age,) Pick does not repeat the last value unless you use an R (right,
frosty? anyone?)

I know why the R was put in- it was meant to re-use (hence REUSE in
BASIC) a *singlevalue* ie if you want to sum QTY (mv'd, attr4)* PRICE
(sv'd attr3) then A;S(3R*4) would work whereas A;S(3*4) would fail. As a
further aside, it's always bugged me that you couldn't use
A;S(N(QTY)R*N(PRICE)), but that's an old beef.


Chandru
ps what's the uv list again? i recently lost all my email history...

Mark Taylor wrote:

Comments in line, below.

cmurthi wrote:
It has been quite a while since I used D3 (only use unidata, now) but
I believe you are correct. I'll point out that unidata has a very nice way
of specifying which attributes are Single valued and which are Multi
Valued, and handles multi values better (IMO) than D3.

Shoulda brought that up when I was writing the A-corr compiler, then!
Would have been interesting to see if the extra code to support that
feature would force a use of a third (512-byte) frame. Kinda wish I had
a copy of acorr0 and acorr1... don't miss any of the other code I worked
on, but always thought those two frames were cool.

It has been written:
"Send an e-mail to: u2-users-subscribe (AT) oliver (DOT) com"
but I've never done that. Also, you might try:
http://www.oliver.com/lists/u2/index.html

-- frosty