IBM Mainframe Forum

by **xboss** » Sun Dec 23, 2012 11:48 pm

Hello all,
Last time I posted a topic "Count number of volume and cylinders of each DASD" which according to forum veteran was ambiguous topic ( I accept it). So I am starting new topic which I believe is straight forward (hope so

)
My requirement is to count each unique record and sum its associated value for example,
AAAA 20
BBBB 10
BBBB 4
BBBB 1600
BBBB 33
CCCC 5
CCCC 9
CCCC 12
DDDD 13
DDDD 14

If above is my input, my output should be

RECORD COUNT SUM
AAAA 1 20
BBBB 4 1647
CCCC 3 26
DDDD 2 27

I know this can be done easily with SORT (though I don't know the exact syntax) but I am finding a way to do it with REXX. So far, I am only able to count the unique number of records (which is 4 in this case but not each unique record)

Select all

count = 0
DO j = 1 to i-1
k = j + 1
if (serial.j == serial.k) then iterate
else count = count + 1
END
say count

I started with Akatsukami suggestion by sorting the records first, but no luck after than. Sorry for making this long, but any help would be really appreciated. Help this poor beginner who wants to learn.

by **enrico-sorichetti** » Mon Dec 24, 2012 1:13 am

You were already given a full working solution to achieve that result here
clist-rexx/topic8595-10.html

no need to sort anything ( if the number of <major> keys is reasonable )
You will only have to parse differently the inputs
and add one additional stem variable to hold the counts

by **enrico-sorichetti** » Mon Dec 24, 2012 1:37 am

but since I am in a good mood here is a snippet which does it

the code

Select all

#! /usr/bin/rexx

/* k refers to the key */
/* c refers to the count */
/* s refers to the sum */

lk = ""
lc = ""
ls = ""

Address HOSTEMU "EXECIO * DISKR <your input file/YOUR INPUT DDNAME> (stem REC. FINIS"

do i = 1 to rec.0

k = strip(word(rec.i,1))
z = translate(k,"__","-*")
s = strip(word(rec.i,2))

if symbol( "sum_key."z ) = "LIT" then do

interpret "sum_key."z" = k"
interpret "sum_cnt."z" = 1"
interpret "sum_sum."z" = s"

lk = lk "sum_key."z
lc = lc "sum_cnt."z
ls = ls "sum_sum."z

end
else do
interpret "sum_cnt."z" = sum_cnt."z" + 1"
interpret "sum_sum."z" = sum_sum."z" + s"
end

end

say left("key",8) right("count",8) right("sum",8)
say ""
do i = 1 to words(lk)
say left(value(word(lk,i)),8) right(value(word(lc,i)),8) right(value(word(ls,i)),8)
end

exit

::requires "hostemu" LIBRARY

the data

Select all

aaaa 10
bbbb 20
cccc 30
dddd 40
aaaa 100
bbbb 200
cccc 300
dddd 400
aaaa 1000
bbbb 2000
cccc 3000
dddd 4000
aaaa 10000
bbbb 20000
cccc 30000
dddd 40000
aaaa 100000
bbbb 200000
cccc 300000
dddd 400000
aaaa 1000000
bbbb 2000000
cccc 3000000
dddd 4000000
aaaa 1
bbbb 2
cccc 3
dddd 4

the result

Select all

key count sum

aaaa 7 1111111
bbbb 7 2222222
cccc 7 3333333
dddd 7 4444444

tested with object rexx on my MAC
the same logic will work for TSO ( instead of a PC file name use a ddname )

by **enrico-sorichetti** » Mon Dec 24, 2012 2:20 am

since it is Christmas time
here a better snippet that sorts also on ascending key

Select all

#! /usr/bin/rexx

/* k refers to the key */
/* c refers to the count */
/* s refers to the sum */

lk = ""
lc = ""
ls = ""

Address HOSTEMU "EXECIO * DISKR <...> (stem REC. FINIS"

do i = 1 to rec.0

k = strip(word(rec.i,1))
z = translate(k,"__","-*")
s = strip(word(rec.i,2))

if symbol( "key."z ) = "LIT" then do

interpret "key."z" = k"
interpret "cnt."z" = 1"
interpret "sum."z" = s"

lk = lk "key."z
lc = lc "cnt."z
ls = ls "sum."z

end
else do
interpret "cnt."z" = cnt."z" + 1"
interpret "sum."z" = sum."z" + s"
end

end

/* sort the keys */
lk = __sort(lk)

say left("key",8) right("count",8) right("sum",8)
say ""
do i = 1 to words(lk)
z = word(lk,i)
parse var z . "." k
say left(value(word(lk,i)),8) right(value("cnt."k),8) right(value("sum."k),8)
end

exit

/*- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
::routine __sort public

parse arg u
n = words(u)
a.0 = n
do i = 1 to n
parse var u a.i u
end

h = a.0
do while h > 1;
h = h % 2;
do i = 1 for a.0 - h
j = i; k = h + i;
do while a.k < a.j
t = a.j; a.j = a.k; a.k = t;
if h >= j then ,
leave;
j = j - h; k = k - h;
end
end
end

s = ""; l = ""
do i = 1 to n
c = a.i
if c \= l then ,
s = s c
l = c
end

return space(s)

::requires "hostemu" LIBRARY

the data

Select all

dddd 4000000
dddd 400000
dddd 40000
dddd 4000
dddd 400
dddd 40
dddd 4
cccc 3000000
cccc 300000
cccc 30000
cccc 3000
cccc 300
cccc 30
cccc 3
bbbb 2000000
bbbb 200000
bbbb 20000
bbbb 2000
bbbb 200
bbbb 20
bbbb 2
aaaa 1000000
aaaa 100000
aaaa 10000
aaaa 1000
aaaa 100
aaaa 10
aaaa 1

the result

Select all

key count sum

aaaa 7 1111111
bbbb 7 2222222
cccc 7 3333333
dddd 7 4444444

by **enrico-sorichetti** » Mon Dec 24, 2012 2:34 am

looks like You have more problems with the logic, rather than with the code

how would You approach the problem in, for example, COBOL ?

by **xboss** » Mon Dec 24, 2012 5:23 pm

Enrico, I have to learn both logic and syntax. Long way to go

Thank you for the solution. It took me a while to demystify it. Appreciate it. Merry X-mas.

by **enrico-sorichetti** » Mon Dec 24, 2012 7:58 pm

Enrico, I have to learn both logic

the logic is the same as finding duplicates in an ordered file

...
read the first record...
... initialize the key ( save the current key )
... initialize the count
... initialize the sum

iterate on all the rest
... if the current key is grater than the saved key
... ... print the values for the saved key
... ... REinitialize the key ( save the current key )
... ... REinitialize the count
... ... REinitialize the sum
... else if the current key is equal to the saved key
... ... sum 1 to the count
... ... sum the current value to the sum
... else ... record out of sequence
... ... tell so and terminate, or tell so and discard the record
... end if
end of iteration
print the values for the saved key
exit

and here is the snippet that does it

Select all

#! /usr/bin/rexx

/* k refers to the key */
/* c refers to the count */
/* s refers to the sum */

Address HOSTEMU "EXECIO * DISKR <file name/DDNAME> (stem REC. FINIS"

say left("key",8) right("count",8) right("sum",8)

k = strip(word(rec.1,1))
c = 1
s = strip(word(rec.1,2))

do i = 2 to rec.0

select
when ( strip(word(rec.i,1)) > k ) then do
say left(k,8) right(c,8) right(s,8)
k = strip(word(rec.i,1))
c = 1
s = strip(word(rec.i,2))
end

when ( strip(word(rec.i,1)) = k ) then do
c = c + 1
s = s + strip(word(rec.i,2))
end

otherwise do
say "discarding " strip(word(rec.i,1))
end

end

end

say left(k,8) right(c,8) right(s,8)

exit

::requires "hostemu" LIBRARY

written for ooRexx, but can be ported easily to TSO Rexx

IBM Mainframe Forum

count each unique record

count each unique record

Re: count each unique record

Re: count each unique record

Re: count each unique record

Re: count each unique record

Re: count each unique record

Re: count each unique record