Why does the lexer allow relations that start with an underscore?

[stefjoosten]

What happened

I noticed that a relation name in Ampersand may start with an underscore. The compiler accepts the following code fragment:

RELATION _takes [Student*Course]
MEANING "A student takes a complete course"

The compiler lets me work with this underscored identifier without problems.

What I expected

I expected an error because the documentation states clearly that relation names start with a lower case letter.

Version of ampersand that was used

Ampersand-v4.1.4 [no git info], build time: 10-Mar-21 13:53:13 UTC

Steps to reproduce

1. Log into RAP4
2. compile an arbitrary script that has a relation with a name that starts with an underscore.

Analysis

I found this error by code inspection.
I traced it back to the lexer, in the function isIdStart:

It is used in the following code fragment in file lexer.hs:

mainLexer p cs@(c:s)
     | isIdStart c || isUpper c
         = let (name', p', s')    = scanIdent (addPos 1 p) s
               name''               = c:name'
               tokt   | iskw name'' = LexKeyword name''
                      | otherwise = if isIdStart c
                                    then LexVarId name''
                                    else LexConId name''
           in returnToken tokt p mainLexer p' s'

It is clear that the function isIdStart is deliberate. So I wonder why it was inserted. Maybe @DanielSchiavini knows?
In the meantime I suggest we treat this like an error and fix it. I propose the following change:

This change is ready for testing in commit 7c4d1ac

[DanielSchiavini]

@stefjoosten I think it was a wrong assumption on my side as most programming languages accept underlines.

[stefjoosten]

Thanks, @DanielSchiavini, we'll fix it.

[hanjoosten]

@stefjoosten , I fixed the readability of the error message that is thrown due to your latest change (02e04a0). Please fix this bug yourself, and make the ci pipeline happy 😃.