NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/comp/sys/mac/programm/14692

Internet Message Format | 1992-08-29 | 3.6 KB

Path: sparky!uunet!haven.umd.edu!darwin.sura.net!zaphod.mps.ohio-state.edu!uwm.edu!ogicse!das-news.harvard.edu!cantaloupe.srv.cs.cmu.edu!crabapple.srv.cs.cmu.edu!andrew.cmu.edu!sr0o+
From: [e-mail] (Steven Ritter)
Newsgroups: comp.sys.mac.programmer
Subject: Re: incrementation differences/THINK C 4.0 vs. 5.0
Message-ID: <[e-mail]>
Date: 28 Aug 92 10:48:01 GMT
Article-I.D.: andrew.webbHVm00WBM842pMu
References: <[e-mail]> <[e-mail]> <[e-mail]>
    <[e-mail]>
Organization: Doctoral student, Psychology, Carnegie Mellon, Pittsburgh, PA
Lines: 67
In-Reply-To: <[e-mail]>

[e-mail] (Roy Smith) writes:

>[e-mail] (Steve Dorner) writes:
>>It does *not* know where the missing ";" should go.  It knows where
>>something is missing.  An '=' would be a perfectly good choice, too.
>
>    I've been thinking about this for a day or so.  In the fragment
>"a=b=c d=e;" the blank could be replaced with just about anything and make
>a syntactically correct statement:
>
>    Any alphanumeric would turn "c d" into a variable name.
>
>    Any binary arithmentic operator (+, -, *, /)would be fine.
>
>    Any relational operator (==, !=, <, >, <=, >=) would be fine.
>
>    Any boolean operator (||, &&) would be fine.
>
>    Any binary logical operator (&, |, etc) would be fine.
>
>    Any assignment operator (=, =+, =-, =*, =/) would be fine.
>
>    Deleting the space completely would give you the variable "cd".
>
>    A comma would be fine, giving you two sequential assignments.
>
>    All things considered, the a-priori probability that you need to
>replace the space by a semicolon is pretty slim.

This is an AI problem, and not a very difficult one. In principle, the
compiler can know as much as you know. Suppose you were modifying
someone else's code and came across the line:
a=b=c d=e;
What would you do?

Most of the above suggestions are illegal. You can't add a binary
arithmetic operator (you'd get c+d=e). Same with a boolean, relational
or logical operator or any of the arithmatic assignment operators (=+,
etc.).

You could add an alpha character (or delete the space) to turn
it into a variable name. Unless there is a variable in scope (and of the
right type) of the form c*d, this is highly unlikely to be the right answer.

Your remaining choices are semi-colon, comma and =. In this case,
semi-colon and comma are semantically identical, so that's not much of a
problem. Deciding between = and the others is slightly tricky. You can
resolve this, in part, by seeing which variables have been initialized.
If only one of the five variables has been initialized, the answer is
almost certainly = (otherwise, you'd be assigning something to an
uninitialized variable). Similarly, if both c and e (for example) have
just been initialized, you're almost certainly missing a semi-colon.

The "if (a=b) case is much trickier, but I think we could do a pretty
good job on this as well. There are some idioms [like if
(fp=fopen(...))] that are clearly intended. For something like "if
(a=b)" I'd check whether 'a' was unassigned before the 'if' and is used
only in the body of that conditional (at least, up to a later
assignment). If so, the assignment is probably intended.

The point of all this is that, in the vase majority of cases, a compiler
with a little intelligence could guess right most of the time. Whether
the occasional errors would be worth it (after all, it's just finding
typos) is a separate question. Natural language grammar checkers guess
right most of the time as well but I find them more annoying than useful.

Steve