Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Stefan Monnier]

> +A CONDITION-NAME of t applies to any error symbol.

All error symbols are caught by `error` already.  Those extra conditions
caught by `t` are non-error conditions (the only one I know of is
`quit`, BTW).


        Stefan

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Noam Postavsky]

On 4 September 2018 at 22:46, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
>> +A CONDITION-NAME of t applies to any error symbol.
>
> All error symbols are caught by `error` already.  Those extra conditions
> caught by `t` are non-error conditions

Hmm, right. The rest of the docstring talks only about errors though.
Should we change the whole thing? While technically more accurate,
saying "Regain control when a signal is signaled" sounds a bit funny.
Otherwise, fixing just that sentence:

 A handler is applicable to an error if CONDITION-NAME is one of the
-error's condition names.  A CONDITION-NAME of t applies to any error
-symbol.  If an error happens, the first applicable handler is run.
+error's condition names.  A CONDITION-NAME of t applies to any condition
+name.  If an error happens, the first applicable handler is run.

> (the only one I know of is `quit`, BTW).

That, and possible silly code which does something like:

   (signal 'not-an-error-symbol "oops")

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Stefan Monnier]

> Hmm, right. The rest of the docstring talks only about errors though.
> Should we change the whole thing? While technically more accurate,
> saying "Regain control when a signal is signaled" sounds a bit funny.

I think the solution normally used is that "signal" is only used for the
action, while the event-object is called "condition".  I.e. 
"Regain control when a condition is signaled".


        Stefan

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Stefan Monnier]

>> (the only one I know of is `quit`, BTW).
> That, and possible silly code which does something like:
>    (signal 'not-an-error-symbol "oops")

I know there can hypothetically be others, but what I was saying above
is that in practice I've only ever seen `quit`.


        Stefan

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Noam Postavsky]

On 5 September 2018 at 11:31, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
>> Hmm, right. The rest of the docstring talks only about errors though.
>> Should we change the whole thing? While technically more accurate,
>> saying "Regain control when a signal is signaled" sounds a bit funny.
>
> I think the solution normally used is that "signal" is only used for the
> action, while the event-object is called "condition".  I.e.
> "Regain control when a condition is signaled".

So how about this:

 DEFUN ("condition-case", Fcondition_case, Scondition_case, 2, UNEVALLED, 0,
-       doc: /* Regain control when an error is signaled.
-Executes BODYFORM and returns its value if no error happens.
+       doc: /* Regain control when a condition is signaled.
+Executes BODYFORM and returns its value if nothing is signaled.
 Each element of HANDLERS looks like (CONDITION-NAME BODY...)
 where the BODY is made of Lisp expressions.

-A handler is applicable to an error if CONDITION-NAME is one of the
-error's condition names.  A CONDITION-NAME of t applies to any error
-symbol.  If an error happens, the first applicable handler is run.
+A handler is applicable to a condition if CONDITION-NAME is one of the
+condition's names.  A CONDITION-NAME of t applies to any condition
+name.  If a condition is signaled, the first applicable handler is run.

 The car of a handler may be a list of condition names [...]

-When a handler handles an error, control returns to the `condition-case'
+When a handler handles a condition, control returns to the `condition-case'
 and it executes the handler's BODY...
 with VAR bound to (ERROR-SYMBOL . SIGNAL-DATA) from the error.

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Stefan Monnier]

>> I think the solution normally used is that "signal" is only used for the
>> action, while the event-object is called "condition".  I.e.
>> "Regain control when a condition is signaled".
>
> So how about this:

Looks good to me,


        Stefan

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Davis Herring]

>   with VAR bound to (ERROR-SYMBOL . SIGNAL-DATA) from the error.

You left one "ERROR" here.

Meanwhile, it might be good to leave the word "error" somewhere in the 
text, just so no one uninitiated reads it and decides it _isn't_ for 
error handling.

Davis

-- 
This product is sold by volume, not by mass.  If it appears too dense or 
too sparse, it is because mass-energy conversion has occurred during 
shipping.

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Eli Zaretskii]

> From: Noam Postavsky <npostavs@gmail.com>
> Date: Thu, 6 Sep 2018 21:01:31 -0400
> Cc: Emacs developers <emacs-devel@gnu.org>
> 
>  DEFUN ("condition-case", Fcondition_case, Scondition_case, 2, UNEVALLED, 0,
> -       doc: /* Regain control when an error is signaled.
> -Executes BODYFORM and returns its value if no error happens.
> +       doc: /* Regain control when a condition is signaled.
> +Executes BODYFORM and returns its value if nothing is signaled.
>  Each element of HANDLERS looks like (CONDITION-NAME BODY...)
>  where the BODY is made of Lisp expressions.
> 
> -A handler is applicable to an error if CONDITION-NAME is one of the
> -error's condition names.  A CONDITION-NAME of t applies to any error
> -symbol.  If an error happens, the first applicable handler is run.
> +A handler is applicable to a condition if CONDITION-NAME is one of the
> +condition's names.  A CONDITION-NAME of t applies to any condition
   ^^^^^^^^^^^
This should be "conditions'", I think, as it's plural.

> +name.  If a condition is signaled, the first applicable handler is run.

I find "condition is signaled" slightly confusing, though.  Why not
use "signal" instead?  A "signal" is not necessarily an error, so I
think the root cause for this change will be taken care of.  WDYT?

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Davis Herring]

>> +A handler is applicable to a condition if CONDITION-NAME is one of the
>> +condition's names.  A CONDITION-NAME of t applies to any condition
>     ^^^^^^^^^^^
> This should be "conditions'", I think, as it's plural.

No: the names are plural, but the condition (that has been signaled) is not.

Davis

-- 
This product is sold by volume, not by mass.  If it appears too dense or 
too sparse, it is because mass-energy conversion has occurred during 
shipping.

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Eli Zaretskii]

> Cc: monnier@iro.umontreal.ca, emacs-devel@gnu.org
> From: Davis Herring <herring@lanl.gov>
> Date: Fri, 7 Sep 2018 01:00:26 -0600
> 
> >> +A handler is applicable to a condition if CONDITION-NAME is one of the
> >> +condition's names.  A CONDITION-NAME of t applies to any condition
> >     ^^^^^^^^^^^
> > This should be "conditions'", I think, as it's plural.
> 
> No: the names are plural, but the condition (that has been signaled) is not.

But "condition's names" means that a single condition has several
names.  Is that what we want to say there?

RE: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Drew Adams]

> > +A handler is applicable to a condition if CONDITION-NAME is one of the
> > +condition's names.  A CONDITION-NAME of t applies to any condition
>    ^^^^^^^^^^^
> This should be "conditions'", I think, as it's plural.

No. "one of the condition names" (as in the original, but without "error").
Or "the name of one of the conditions".

RE: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Drew Adams]

> > >> +A handler is applicable to a condition if CONDITION-NAME is one of the
> > >> +condition's names.  A CONDITION-NAME of t applies to any condition
> > >     ^^^^^^^^^^^
> > > This should be "conditions'", I think, as it's plural.
> >
> > No: the names are plural, but the condition (that has been signaled) is not.
> 
> But "condition's names" means that a single condition has several
> names.  Is that what we want to say there?

No, it's not what we want to say here. But we also don't want
to say "one of the conditions names".

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Noam Postavsky]

On 6 September 2018 at 21:10, Davis Herring <herring@lanl.gov> wrote:
>>   with VAR bound to (ERROR-SYMBOL . SIGNAL-DATA) from the error.
>
>
> You left one "ERROR" here.
>
> Meanwhile, it might be good to leave the word "error" somewhere in the text,
> just so no one uninitiated reads it and decides it _isn't_ for error
> handling.

Yeah, hmm. Maybe it's not such a great idea to remove "error"
everywhere just to cover some obscure corner case. Just mentioning
that corner case specifically should be enough to get the idea across.

 A handler is applicable to an error if CONDITION-NAME is one of the
-error's condition names.  A CONDITION-NAME of t applies to any error
-symbol.  If an error happens, the first applicable handler is run.
+error's condition names.  Handlers may also apply when non-error
+symbols are signaled (e.g., `quit').  A CONDITION-NAME of t applies to
+any symbol, including non-error symbols.  If multiple handlers are
+applicable, only the first one runs.

Re: [Emacs-diffs] master 425c281 1/3: Allow t as a catch-all condition-case handler (Bug#24618)

[Noam Postavsky]

On 7 September 2018 at 22:16, Noam Postavsky <npostavs@gmail.com> wrote:
> On 6 September 2018 at 21:10, Davis Herring <herring@lanl.gov> wrote:

>> Meanwhile, it might be good to leave the word "error" somewhere in the text,
>> just so no one uninitiated reads it and decides it _isn't_ for error
>> handling.
>
> Yeah, hmm. Maybe it's not such a great idea to remove "error"
> everywhere just to cover some obscure corner case. Just mentioning
> that corner case specifically should be enough to get the idea across.

No further comments, so I pushed to master.

[1: 3a2ffa62a7]: 2018-09-16 20:25:45 -0400
  Clarify condition-case docstring
  https://git.savannah.gnu.org/cgit/emacs.git/commit/?id=3a2ffa62a7e491fda1083cbedb165a3e49fd21c6