ignore-error?

ignore-error?

[Lars Ingebrigtsen]

It seems like people like writing `ignore-errors' a lot more than
`condition-case', and I think `ignore-errors' is the wrong thing to say
90% of the time.

So would it be a good idea to add a new convenience macro `ignore-error'
to entice people to not use `ignore-errors' so much?  It would take an
error condition as its parameter, so:

  (ignore-error 'file-error
    (directory-files dir))

which would simply expand to

  (condition-case _
      (directory-files dir)
    (file-error nil))

It indents nicely, too.  :-)

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: ignore-error?

[Stefan Monnier]

> It seems like people like writing `ignore-errors' a lot more than
> `condition-case', and I think `ignore-errors' is the wrong thing to say
> 90% of the time.

I agree, but I think in 99% of those cases, the author has no idea which
error might need to be caught.

> So would it be a good idea to add a new convenience macro `ignore-error'
> to entice people to not use `ignore-errors' so much?  It would take an
> error condition as its parameter, so:
>
>   (ignore-error 'file-error
>     (directory-files dir))

[ I think the argument shouldn't be quoted.  ]
More importantly, if you want me to use this, you'll have to improve
elisp-mode's completion so it DTRT for the error symbol, like it does in
condition-case.


        Stefan

Re: ignore-error?

[Lars Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> It seems like people like writing `ignore-errors' a lot more than
>> `condition-case', and I think `ignore-errors' is the wrong thing to say
>> 90% of the time.
>
> I agree, but I think in 99% of those cases, the author has no idea which
> error might need to be caught.

That's true.  If only Emacs Lisp had...  checked exceptions.  *runs
away*

*then returns*

But it sure would be nice if we had some way of telling at least some of
the errors that a function can raise... 

>> So would it be a good idea to add a new convenience macro `ignore-error'
>> to entice people to not use `ignore-errors' so much?  It would take an
>> error condition as its parameter, so:
>>
>>   (ignore-error 'file-error
>>     (directory-files dir))
>
> [ I think the argument shouldn't be quoted.  ]

True.

> More importantly, if you want me to use this, you'll have to improve
> elisp-mode's completion so it DTRT for the error symbol, like it does in
> condition-case.

Let's see...  It this stuff?

                       ((and (or 'condition-case 'condition-case-unless-debug)
                             (guard (save-excursion
                                      (ignore-errors
                                        (forward-sexp 2)
                                        (< (point) beg)))))
                        (list t obarray
                              :predicate (lambda (sym) (get sym 'error-conditions))))

Sure, that should be simple enough to cargo cult for the new macro, I
think.

The parameter should perhaps also be allowed to be a list of symbols:

(ignore-error (file-error some-other-error)
  (foo))

?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: ignore-error?

[Stefan Monnier]

> But it sure would be nice if we had some way of telling at least some of
> the errors that a function can raise...

We could decide that docstrings should mention which errors
are signal'd, indeed.  Ideally in a standardized format.

> Let's see...  It's this stuff?

Yup.

> The parameter should perhaps also be allowed to be a list of symbols:
>
> (ignore-error (file-error some-other-error)
>   (foo))

Even better,


        Stefan

Re: ignore-error?

[Lars Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> Let's see...  It's this stuff?
>
> Yup.
>
>> The parameter should perhaps also be allowed to be a list of symbols:
>>
>> (ignore-error (file-error some-other-error)
>>   (foo))
>
> Even better,

I have now added `ignore-error', and provided for completion...  but
with some caveats, as `elisp-completion-at-point' isn't...  er...  quite
right?  :-)

Right: If you say

(ignore-erC-M-i

then funpos will indicate that we're in a function position and complete
over functions (which is correct).

(ignore-error C-M-i

still has funpos indicating that we're in a function definition, which
doesn't seem correct?  But even if we fix that, then `end' will indicate
that we can't complete anything anyway.

(ignore-error argC-M-i

will have everything correct and complete to

(ignore-error args-out-of-range

Unfortunately, so will

(ignore-error foo
  argC-M-i

because we don't have a count at all where we are in the sexp we're
completing.  But on the other hand, the sexps are usually not closed
when we're completing anyway, so I guess that's a hard problem.

And, by the way

(condition-case nil
   (foo)
  (argC-M-i

completes well, but not

(condition-case nil
   (foo)
  ((argC-M-i

that is, where the condition is a list...

It seems like that completing function could, like, use more...  stuff.
:-)  But I guess it's a DWIM best effort kinda thing.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: ignore-error?

[Stefan Monnier]

> It seems like that completing function could, like, use more...  stuff.

Yes, if you have love to give it, it will be eternally thankful, and
I will also appreciate it,


        Stefan