$CASE/gen/ver_12/aareadme - RJL 990514 rev. 5/19, 6/3

Rev. 011018  bo RJL:
-------------------------------
Updated group to  bdelog, ready for sjaganat's project in 01s522,
which will add changes that RJL manually edited into  
	~lechner/fac5/bde2alpha_rjl/bde/new_pr_util.2k0902.
$CASE/gen/ver_12 will be an incremental enhancement 
to the current chgen version in 
	genmerge/genv11/chgen/src/gen*.c .
and will be checked into the main branch of chgen in CVS.

Immediate Goal: enable chgen to generate unix/g++ code that is 
compilable and free of warnings from any .sch file..

Long-range goal: make chgen architecture more modular. 
better encapsulated, and closer to gencpp so an effective 
comparison can be made.

Note on: gencpp vs. chgen:
---------------------------
In 2kf522, the gencpp project by kmiu/mottesen modified ntansala's 
gencpp with similar but not identical fixes to make gencpp 
generated code pr_*.c and schema.h portable to two platforms:: 
	VC++/Wintel and g++/unixAlpha.

Their gencpp generated pr_*.c code which depended on -DUSE_STL
to either use external STL <list> containers 
or use earlier intrusive parent-child linked lists.
Unfortunately the code changes for external vs. internal lists 
are too large to effectively compare chgen vs. gencpp architecture.

MOttesen also began building genjava  (schema parser first)
in /usr/proj3/case/2kf522/mottesen/genjava (01s591 course)
with intrusive threaded lists for parent-child relations,
-----------------------------

RJL 011018
---------------------------------
1999 version:

This file contains proposed goals for BYoo's project.

Directory gen/ver_12 was created for zlu byoo and yxu to
complete their respective projects in 99s523.
However, there work is still under their directories below.

ZLu is studying the OLCArch XXcurr->tableSet/Getfield 
macros vs the pr_set_typ(XXcurr,table,field,value)  
macros for genlog/replay.  See $CASE/99s523/zlu.

BYoo is workng on partitioning chgen output
source and header files into table-type-specific 
files. See 99s523/byoo.

Both tasks are relevant to making chgen into gencpp
which creates class declarations, containers and 
methods from global struct declarations, tables and 
functions.

For Yicheng Xu's dbde/bdeReplay, see $CASE/99s523/yxu

======================================

Goals for BYoo genPartition project:
---------------------

Reverse the  control hierarchy in the pr_* and schema.h 
modules generated by $CASE/gen/ver_11   repository code
so that the  order in which table-specialized  functions 
are generated is sorted by table first then function name.

The collective main switch(table_type) will have small
cases  for each table type, that call the table-specific
methods earlier in the file.

The next step is to split the old files by table-type
into multiple files, one per table type,  with only
the switch(tabletype)  blocks remaining in schema.c
and pr*.c.

RJL email late May suggested the first  code splitting step
could be done one gen-created pr_*.c  file at a time.
(At least two should be done in 99su.)

After all files have their case block code split out  and retests
succeed, then code can be moved to table-specific files
with table_type XX prefixed to the name XX_ops.c or XX_schema.h

Reference: genv7 and pr_util structure chart diagrams.
(needs  updating  but is still a good guide).
THese diagrams in  $CASE/gen/ver_7.5orless/ver_7/doc/idraw
are  gzipped postscript subtype .idraw.
The  X11 idraw editor is on remus:/usr/local/bin/idraw
but not on  jupiter.

Naming  conventions: 
--------------------
(Any  changes require my  approval in advance)

Add  XX_schema.h and XX_ops.cc where XX is a table type.
If table  abbrev is 4 bytes not  2, then the same  4 chars
must be in  the prefix:  XXXX_ops.cc and XXXX_schema.h.
These chars are ALWAYS UPPER CASE not lower case.
Schema violatinos of this should be ERROR messages,
not warnings  from   GEN.

('schema' is generic  - gen should tailor this name
to  agree exactly with the schemaname.sch file that gen parses
before generating code for any test case.)