Steps in cloning a bsubcyc flatfile loader for a single file to another file:


*. convert .l file
*. convert .y file
* create XX-parse.h and structs needed by frame
* add clear_ routine, call it from grammar
  Use strings or lists (ie. pointers) for each field.
  NULL pointer = undefined = NULL value at load time
* add loading semantics to XX-load.pc
* add above files to Makefile

To add a column to a table:

1. Unless its value is to be ignored, add a field to the
   XX_entry struct in XX_parse.h
2. Add its keyword to XX.l; if it has special syntax,
   add code for scanning it as well.
3. Add the token for the keyword to XX.y; also
   add the production(s) needed to recognize and save the attribute value.
4. Add field to compound_clear_entry in XX.y
5. Add the column to the SQL INSERT code in XX-load.pc;
   add the code to set its indicator variable as well,
   to handle case where attribute is missing.

         
TODO before release:


=============================================================================
Issues with flatfile data:
general:
* COMMON-NAME has XML-like fields embedded eg. <SUP>

compounds.dat:
* no SMILES

reactions.dat:
* missing DELTAG0, SPONTANEOUS 
* is EC-NUMBER official or proposed if OFFICIAL-EC? slot is missing?
* many LEFT and RIGHT compounds not in compounds.dat. Should they be added to warehouse?
* ignored DBLINKS

pathways.dat:
* Are rxns ordered?

proteins.dat:
* Semantics of DBLINKS attribute? Ex:
UNIQUE-ID - DIHYDROLIPOYL-GCVH
TYPES - RED-DIHYDROLIPOAMIDE
TYPES - Gcv-H
COMMON-NAME - dihydrolipoyl-GcvH-protein
DBLINKS - (ENTREZ P23884 NIL pkarp 3102853742)
DBLINKS - (SWISSPROT P23884 NIL pkarp 3102853742)
SYNONYMS - dihydrolipoylprotein
//
UNIQUE-ID - FERREDOXIN-MONOMER
TYPES - Ferredoxin
COMMON-NAME - reduced ferredoxin
DBLINKS - (SWISSPROT P50727 NIL NIL NIL NIL NIL)
GENE - BG11388
MODIFIED-FORM - OX-FERREDOXIN
SYNONYMS - Fer
//




Issues related to Pro C:

Host variables that correspond to column type NUMBER (as opposed to NUMBER(5))
may not be a floating-point C type. This is consistent with the Pro C documentation.
However, it seems like char* C types work fine. Since all warehouse numerics are
NUMBER, char* host variables are used for loading.


Loader issues/todos:

Entry table: include CreationDate, ModifiedDate