Grammar related unclarities

[PatrickvanLogchem]

There seems to be no mention of the actual format these sql-*-grammar.txt files are expressed in, so : What format is being used in this project, exactly?

It looks a bit like an extended version of BNF, but not quite EBNF because of some deviations:

• the document format is not specified (having chapters, paragraphs, function headers, function texts, format headers, format texts, rules, notes, etc.)
• there are text blocks between rule declarations, like Syntax Rules in sql-2008 line 5935 and NOTE in sql-2011 line 137
• the non-standard way that repeats are declared, using { expression }...
• grouping syntax doesn't use standard ( expression ), but instead uses { expression } (without the 3 dots after the })
• the non-standard !! comment prefixes
• the lack of double-enquoted literals
• the use of single-character declarations which use otherwise special characters, like <less than operator> ::= <, <double quote> ::= ", <left bracket> ::= [, <left brace> ::={ and others

Also, there are a few irregularities in the declared rules themselves;

• some rules are declared twice (once with a comment, once as a declaration) in the same document, like <item number> in sql-92 and <JSON path context variable> in sql-2016
• unspecified case-sensitivity of literals, like in <reserved word> and keyword occurrences in many rules
• some rules are referred to, but not declared, like in sql-1999 : <assertion definition>,<old or new value alias>,<attribute definition>, <source data type>, <user-defined cast definition>, <SQL data change statement>, <target table> and in sql-2003 : <contextually typed row value specification>, etc.
• some rules have multiple spaces and other non-identifier characters in them, like in sql-1999 : <delete statement: positioned>

Could you please address these issues?

[PatrickvanLogchem]

The BNF format used here seems mostly described in https://sql-99.readthedocs.io/en/latest/chapters/01.html#how-to-read-sql-syntax - is that correct?