Conference azur::mcc

2066.0. "On commas" by TAEC::HARPER (John Harper, DTN 830 3647) Fri Jan 10 1992 10:53

    Following on from 2065.0, the general handling of commas in the FCL is
    distinctly user-unfriendly.  Again, I'm not suggesting that this is a
    "must fix" for V1.2, but it could make the product noticeably less
    daunting for the novice if it could get fixed some time.
    
    The whole business of commas is very confusing.  Why do some things
    need commas and others not?  Most of the time it's perfectly clear
    without them, e.g.:
    
    show recording node4 tenere circuit bna-0 partition counters in domain .blah
    
    couldn't possibly be interpreted other than as intended. The actual
    error message, "%MCC-E-LATIN1STRING_ERROR, error in Latin1String
    value", is neither friendly nor especially illuminating.
    
    Now, since I was involved in the architecture of NCL, I vaguely
    remember why the comma got invented, which was to do with some obscure
    cases of possible ambiguity. 99.9% of the time, however, there is
    no ambiguity and a mildly intelligent parser could sort it out.
    
    Thus, instead of giving an error, the parser could "imagine" a comma
    and see if it could do better.  Similarly for:
    
    MCC> show recording node4 tenere circuit bna-0, partition .........
    %MCC-W-QUALUNKNOWN, qualifier is unknown: PARTITION
    
    the parser could "imagine" the absence of the comma.  (If you wanted to
    be pedantic, you could have an information message saying what the
    "right" syntax was, to which the command had been translated).
    
    (There is a parallel here with programming languages and semicolons.
    Only one language that I know of, BCPL, had the wit to allow you to
    miss out semicolons in the 99% of cases where it was unambiguous
    anyway).
    
    	John

    What we currently do is suggest that people use the forms mode of the
    FCL to learn the syntax. The context sensitive help <pf2> helps with
    putting in the commas and building a syntactically correct command.
    Right now the problem with removing commas is that we need some form of
    delimiter between arguments weather it is a DCL \ or an ultrix - most
    languages have something. Unluckily NCL (the language DECmcc uses) is
    very english like except in its use of punctuation.
    
    If you ahve more suggestions in this area they would be appreciated. 
    
    As to the error messages, the problem with the ambiguities is that the
    parser tries all possibilities before returning an error. Unluckily the
    error we return is on the last thing we tried not the first. If you use
    forms mode you will find that PF2 also gives better messages.