Talk:Compiler/Verifying syntax

From Rosetta Code

Incorrect BNF

The BNF given does not permit = to be directly followed by not. Go/Julia/Perl all get that wrong (better to fix task tho?).
The Go output also claims that "g not = h" is valid, whereas Perl/Julia and now Phix call that invalid. --Pete Lomax (talk) 18:57, 3 January 2020 (UTC)

I may not understand the specification given for BNF syntax. Doesn't the rule:

expr_level_4 = ["not"] expr_level_5 [('=' | '<') expr_level_5] ;

mean any expr_level_5 expression can be preceded by not? --Wherrera (talk) 19:09, 3 January 2020 (UTC)

First/left one yes, second/right one no. Consider expr_level_4 = ["not"] expr_level_5 [('=' | '<') ["not"] expr_level_5] ;
Ideally the not would be moved into its own private rule, rather than clog up the (partial) comparison rule anyway. --Pete Lomax (talk) 19:31, 3 January 2020 (UTC)
Presumably the "not" applies to the relation as a whole, so "not a = b" means "not ( a = b )" and not "( not a ) = b". --Tigerofdarkness (talk) 22:01, 5 January 2020 (UTC)
Not sure, I would say that is undefined. Putting "not" in a separate rule would give it an unambiguous precedence. --Pete Lomax (talk) 23:44, 6 January 2020 (UTC)

Note the square brackets [] around the [('=' | '<') expr_level_5]. This means that the expression

'2 < 3 < 4'

should not be legal (i.e. a non-associative violation). [] means 0 or 1. {} means 0 or more. --Tybalt89 (talk) 22:28, 5 January 2020 (UTC)

Well spotted. Please sign your posts, with --~~~~ thanks --Pete Lomax (talk) 14:14, 4 January 2020 (UTC)
I didn't know how to sign :( --Tybalt89 (talk) 22:28, 5 January 2020 (UTC)

This problem is in dire need of a really extensive test suite. --Tybalt89 (talk) 22:28, 5 January 2020 (UTC)

That should be part of the task spec. --Pete Lomax (talk) 14:14, 4 January 2020 (UTC)

Would it not be better if the EBNF included explicit placement of whitespace characters? As it stands, it seems to me one is left to one's own opinion as to where whitespace is allowed or required. (A separate lexical analysis stage would alleviate this difficulty, but we already have separated lex and parse tasks.) --Chemoelectric (talk) 18:40, 15 October 2022 (UTC)