Re: simplifying foreign key/RI checks - Mailing list pgsql-hackers

Hi,

+       for (i = 0; i < riinfo->nkeys; i++)
+       {
+           Oid     eq_opr = eq_oprs[i];
+           Oid     typeid = RIAttType(fk_rel, riinfo->fk_attnums[i]);
+           RI_CompareHashEntry *entry = ri_HashCompareOp(eq_opr, typeid);
+
+           if (pk_nulls[i] != 'n' && OidIsValid(entry->cast_func_finfo.fn_oid))

It seems the pk_nulls[i] != 'n' check can be lifted ahead of the assignment to the three local variables. That way, ri_HashCompareOp wouldn't be called when pk_nulls[i] == 'n'.

+           case TM_Updated:
+               if (IsolationUsesXactSnapshot())

...

+           case TM_Deleted:
+               if (IsolationUsesXactSnapshot())

It seems the handling for TM_Updated and TM_Deleted is the same. The cases for these two values can be put next to each other (saving one block of code).

Cheers