IBM Mainframe Forum

hi,
i have some confused about TSO Free command operands file and dataset, are they the same function,whichever i use is ok ? or what condition to use ?
i ran some test in rexx,

1,fisrt example, one file allocate to one ddname, so at then end of code, i use "free fi(myindd)" or "free da(CUSTOMER.MASTER)" , it works fine, resubmit wouldn't be error.


2, second example, two file alloc to one ddname,
if i use "free da(CUSTOMER.MASTER,COPY.MASTER)" , after display two files,it will say:
"DATA SET userid.CUSTOMER.MASTER NOT FREED, MEMBER OF CONCATENATION
DATA SET userid.COPY.MASTER NOT FREED, MEMBER OF CONCATENATION"
resubmit will be error.
if i use "free fi(myindd)" , i will work fine.


3, third exaple, one file allocate to two ddname, and i use "free fi(myindd,mydd) " or "free da(CUSTOMER.MASTER)" both works fine,resubmit wouldn't be error.
another question, i use "old" operand, it should be exclusive use, i think the second allocate should fail, but it doesn't .why?

when a command argument consists of a list
<generally> all the element list are considered as atomic unrelated entities

the rules for the free command just follow that convention
when freeing by dataset using a dataset list each dataset is processed atomically with no concern for the other datasets in the list

so when checking for a possible concatenation it just check for a YES/NO
checking for an overall congruence would be asking too much

using for the free the <list> approach is just a shorthand to avoid multiple free

In this sequence -

ALLOCATE FILE(A) DATASET(B C) OLD
FREE DATASET(B)

What I think happens is the system goes through your allocations looking for all allocations of dataset B and tries to free them. It can't free "file" A for the reason stated by the system. There is no "fix;" the system is working as it is designed; it won't break the concatenation because you are not saying very explicitly you want it to break the concatenation. This sequence works because you are saying you want to free the entire concatenation, not just part of the concatenation.

ALLOCATE FILE(A) DATASET(B C)
FREE FILE(A)

In this sequence -

ALLOCATE FILE(A) DATASET(B) OLD
ALLOCATE FILE(A) DATASET(B) OLD

the system effectively ignores the second ALLOCATE command because it deduces you've already done the allocation. It will reject the second ALLOCATE command in this sequence.

ALLOCATE FILE(A) DATASET(B)
ALLOCATE FILE(A) DATASET(C)

because "file" A is already used. You can do this sequence -

ALLOCATE FILE(A) DATASET(B)
ALLOCATE FILE(A) DATASET(C) REUS

In the second ALLOCATE command the system unconditionally does what amounts to -

FREE FILE(A)

(and it will not complain if "file" A is not actually allocated) before it tries the actual allocation.

Many shops wrote commands like this -

CONCAT (A B C)
DECONCAT A

to use in this sequence -

ALLOCATE FILE(A) DATASET(D)
ALLOCATE FILE(B) DATASET(E)
ALLOCATE FILE(C) DATASET(F)
CONCAT (A B C)
...
DECONCAT A
FREE FILE(A B C)

before IBM changed the ALLOCATE command to do an implied concatenation, which was not possible to do back in the days of MVT. This CONCAT command was relatively easy to write using either IKJDAIR dynamic allocation before MVS or MVS dynamic allocation. Notice that ALLOCATE FILE(A) DATASET(B C) doe this under the covers: ALLOCATE FILE(A) DATASET(B); ALLOCATE DATASET(C) (and remember the "file" name);CONCAT (A "file" allocated for dataset C).

thanks for reply!