From lechner@cs.uml.edu  Tue Dec  7 03:21:13 2004
From: Bob Lechner <lechner@cs.uml.edu>
Subject: $PH/COOL-LCP/pr_set_keyNotes041207.txt   -  RJL041207
To: 04f522
RJLRef: $PH/COOL-LCP/lcp04s/AISetStateNotes.txt 
AISetStateNotes041207.txt

[Revised 050926: 
MLauzier showed in her 045522 project that pr_set_fkey 
could not be defined without a switch on table-type index,
because this is the only way to generate the special cases
that each table requires, due to its different fkey
data members, one of which must be identified from
a macro argument. This project was never completed. 
The current state STid field of each AI row remains 
a non-key field (c8, with is_key = 0).

I cannot vouch for the same pr_set_key function in
pr_accessors.c from 04f522 genv13 project.
If this exists, it may have the same problem 
(failure to update the XXid_pp and XXid_fcp pointers 
when XXid is updated).

For details, see  
$CASE/04s522/lcp04s/ajlopez/lcp/olc/AISetStateNotes.txt
(revised 050926)
($PH/COOL-LCP/lcp04s hyperlinks to $CASE/04s522/lcp04s/.)
]

---------------------------------------------
$PH/COOL-LCP/lcp04s/

$CASE/04s522/ajlopez/lcp/olc/AISetStateNotes.txt 
and $PH/COOL-LCP/pr_set_keyNotes041207.txt 
are identical notes on the differences between 
lcp/olc and lcp/pr_util forms of field updating. 

pr_util's pr_set_field does logging, and is more complex.
olc's TableSetField does no logging, but it should.
TableSetField is easier to generate and hence maintain. 

Olc's author August Reinig
and chgen's authors Craig and Stephen Smith 
made different assumptions about how ASCII fields 
are stored in table rows.
pr_set_keyNotes041207.txt also covers this subject.

This note is suggested to become part of the 
04f522 project to produce a valid pr_set_key macro.
It suggests that a new function (with logging),
called as AISetSTid(AIcurr, TRcurr->STid2) 
would be  a valid alternative way to set a new fkey 
STid in table AI. Logging and pr_link/unlink actions
still need to be added when fkeys are updated.

The special case AISetSTid should be generated 
by updating Reinig's SETROUTINE macro. Later, 
every XXSetFieldname function can be generated 
from the .msdat metadata file, 
by looping over all of its tables and fields
to pick up the required parameters.
E.g.,new function  AISetSTid can be generated by this 
macro call: 
	GENSET(hcg_key,AI,STid,hcg_key);
then called as 
	AISetSTid(AIcurr, value).

Currently this macro is defined in olc3common; here,
I expanded the FUNCTION2(..) macro which is obsolete,
and changed it so XXcurr is used not XXid:
-----------------------------------
#define SETROUTINE(field_type, object_name, object_abbrev, field, value_type) \
void P(object_name,Set,field)(hcg_key P(object_abbrev,curr),value_type value)
{                                                                       \
    pr_find(object_abbrev, P(object_abbrev,id), P(object_abbrev,id));   \
    //pr_find checks XXcurr->XXid  before searching for XXid - RJL041207\
    if (P(object_abbrev,curr)!= 0) {					\
    	P(object_abbrev,curr)->field = (field_type)value;               \
    else P(object_abbrev,curr)->field = (field_type)AINULL;//null value \
}
------------------------------------
The above definition should be modified and renamed GENSET.
In GENSET, object-abbrev AI replaces object_name ActiveInstance,
so the GENSET macro can eliminate object_name arg2 in SETROUTINE
Invoking the new macro as  
	GENSET(hcg_key, AI,STid,hcg_key); 
should expand to this new AISetSTid function :
-------------------
void AISetSTid(hcg_key AIid, hcg_key value)
{
	pr_find(AI, AIid, AIid); //checks AIcurr->AIid before searching
	if (AIcurr!= 0) {
		AIcurr->STid = (hcg_key)value; 
		//TBD: add pr_unlink nd pr_link actions here
	else {
		AIcurr->STid = (hcg_key) AINULL; //customized? (e.g AI-row #0)
		//TBD: issue warning or error message here;
	}
	//TBD: Add logging output here (from pr_util's pr_set_key macro)
}
----------------
With gencpp, either the AIcurr object pointer or an implicit argument 
'this' can apply to class-based Set and Get methods. Perhaps these become 
two over-loaded varieties of AI::AISetSTid?
-------------------------------------