Hi,

is there a handy (and parseable) Open Source SQL-grammar around,
that can be fed into a parser generator without major problems?
I consider ISO 9075:2003 useless for this purpose and don't get
the point of this "standard".

I have (many) problems with the ISO grammar.
Here is one I spotted yesterday: Is the following string

"/* this is */ a strange comment */"

a valid SQL-Comment (according to SQL 2003)?
I think I have to say (reluctantly) yes:

Proof
Excerpts from ISO 9075-2:2003 (with rule numbers attached):

<comment> ::=
<simple comment>
<bracketed comment> ::=
<bracketed comment introducer> (2)
<bracketed comment contents>
<bracketed comment terminator>

<bracketed comment introducer> ::= /* (3)

<bracketed comment terminator> ::= */ (4)

<bracketed comment contents> ::=
!! See the Syntax Rules
[ { <comment character> | <separator> }... ] (5)

<comment character> ::=
<nonquote character> (6)
<nonquote character> ::= !! See the Syntax Rules. (7)

<solidus> ::= / (8)

<asterisk> ::= * (9)


Syntax rules: to (5) and (7)
10) Within a <bracketed comment contents>, any <solidus> immediately followed by an <asterisk> without
any intervening <separator> shall be considered to be the <bracketed comment introducer> of a <separator>
that is a <bracketed comment>.

11) A <nonquote character> is one of:
a) Any character of the source language character set other than a <quote>.
b) Any character other than a <quote> in the character set identified by the <character set specification>
or implied by “N”.

Derivation:
<comment>
--> <bracketed comment> (1)
--> <bracketed comment introducer> (2)
<bracketed comment contents>
<bracketed comment terminator>
--> "/*" (3,4,5)
[ { <comment character> | <separator> }... ]
"*/"
--> "/*" (6) und Syntax rules 10 und 11.
" this is */ a strange comment "
"*/"
--> "/* this is */ a strange comment */"
^
This <solidus> is not immediately followed by an <asterisk>
and it is a <nonquote-character>, too.

Did I (or the authors ouff this _standard_ (n.b.)) miss something I should have norticed?

Norbert

=== original message ===
from: Norbert_Paul
date: 2011-07-07 09:32

I'd say no. According to the grammer, "/* this is */" is the <bracket
comment> and what's following is not valid.
As soon as a token is found (in this case the first <bracket comment
terminator>) it is processed. Therefore the first "*/" ends the bracket
comment and the following characters are not part of the comment anymore.

Cheers,
Udo

Norbert_Paul wrote:

Did you try the BNF grammar files at
http://savage.net.au/SQL/index.html

If you happen to succeed could you please post a "how-to"?
I was looking for the same stuff some years ago (to implement a source
code beautifier) and i failed.

They just have the wording of a law code and you must be a lawyer (or
similar) to actually read (and understand) more than just a few
paragraphs ;-)

No.
If you don't remove the double quotes it's an (very strange) identifier
and if you remove them it should result in a syntax error because the
comment ends after the first "*/".

The standard defines nested comments, but not all DBMSes support them:
"/* this is a /* nested */ comment */"

You probably wouldn't consider this as a valid string:
'bla' blabla'

Dieter