On Mon, 2011-03-28 at 10:28 +0100, alex wrote:
> On 27 March 2011 18:37, evan.raskob [lists] <lists_at_lowfrequency.org> wrote:
> > Oh this is a big pet peeve of mine. (Error messages, that is)
> > Terrible, terrible. They are designed to tell humans about potential
> > mistakes, but written in a language more suited for a computer. They are
> > written in that passive-case, "engineering speak" that tried to sound
> > efficient but really talks around what it is trying to say. "Syntax error
> > could not find matching }" could easily have been "You have an open {
> > without a } to match it - perhaps one is missing somewhere around line 20?"
> > if written by a human with an understanding of written (human) language and,
> > even more important, compassion and base-level empathy for other lesser,
> > imperfect humans who might possibly make an error.
>
> Yes so true, it seems error messages and warnings (which are often
> errors too) are the last thing that compiler writers think about.
I asked a compiler programmer about this once and he told me this was
one of the hardest parts of compiler design. It's certainly not a case
of laziness - figuring out from the context of a parse error what you
were really trying to do is one of those Hard Problems.
It's been interesting following the progress of gcc over the years and
seeing how it slowly but steadily improves it's error reporting to infer
probable semantics from broken syntax.
cheers,
dave
Received on Mon Mar 28 2011 - 18:14:01 BST