| Transaction Type | Behaviour |
|---|
| Deletes |
For previously deleted rows, the PK is added to the `outDEL` table
If the row no longer exists in the base table, the row is added to the
`errDS` table instead.
|
| Inserts |
Previously newly added rows are added to the `outADD` table UNLESS they
are present in the Base table.
In this case they are added to the
`errDS` table instead.
|
| Updates |
Previously modified rows are merged with base table values such that
only the individual cells that were _previously_ changed are re-applied.
Where the row contains cells that were not marked as having changed in
the prior transaction, the 'blanks' are filled with base table values in
the `outMOD` table.
If the row no longer exists on the base table, then the row is added to
the `errDS` table instead.
|
To illustrate the above with a diagram:
@dot
digraph {
rankdir="TB"
start[label="Transaction Type?" shape=Mdiamond]
del[label="Base Row Exists?" shape=rectangle]
add [label="Base Row Exists?" shape=rectangle]
mod [label="Base Row Exists?" shape=rectangle]
outmod [label="outMOD Table" shape=box3d]
outadd [label="outADD Table" shape=box3d]
outdel [label="outDEL Table" shape=box3d]
outerr [label="ErrDS Table" shape=box3d]
start -> del [label="Delete"]
start -> add [label="Insert"]
start -> mod [label="Update"]
del -> outdel [label="Yes"]
del -> outerr [label="No" color="Red" fontcolor="Red"]
add -> outadd [label="Yes"]
add -> outerr [label="No" color="Red" fontcolor="Red"]
mod -> outerr [label="Yes" color="Red" fontcolor="Red"]
mod -> outmod [label="No"]
}
@enddot
For examples of usage, check out the mp_stackdiffs.test.sas program.
@param [in] baselibds Base Table against which the changes will be applied,
in libref.dataset format.
@param [in] auditlibds Dataset with previously applied transactions, to be
re-applied. Use libref.dataset format.
DDL as follows: %mp_coretable(DIFFTABLE)
@param [in] key Space seperated list of key variables
@param [in] mdebug= Set to 1 to enable DEBUG messages and preserve outputs
@param [in] processed_dttm_var= (0) If a variable is being used to mark
the processed datetime, put the name of the variable here. It will NOT
be included in the staged dataset (the load process is expected to
provide this)
@param [out] errds= (work.errds) Output table containing problematic records.
The columns of this table are:
@li PK_VARS - Space separated list of primary key variable names
@li PK_VALS - Slash separted list of PK variable values
@li ERR_MSG - Explanation of why this record is problematic
@param [out] outmod= (work.outmod) Output table containing modified records
@param [out] outadd= (work.outadd) Output table containing additional records
@param [out] outdel= (work.outdel) Output table containing deleted records