Re: [Bug-apl] ⎕ES wrong message

[Juergen Sauermann]

Hi,

Kacper is right - it is in the standard.

But the standard merely says "Set the rest of event-message as appropriate".

The rest is rather APL2 specific. I could also argue that the current value of the parameter
is more relevant to the exception than the original value. I have put this into
README-4-compliance as mid-term issue because I believe the requirement is really bad.

/// Jürgen


On 07/14/2014 05:19 PM, Blake McBride wrote:

Dear Juergen,

Every place that I use ⎕ES I am being forced to treat the arguments as read-only.  I also had to re-order code to account for optional arguments. It makes for some ugly code.  I hope when you add lazy copying of arrays, the internal code to do this correctly won't be a problem.

Showing it as a locked function, IMO, is very important.  The point being that you are telling the user, don't worry about what is inside, it doesn't matter, you passed the wrong arguments.  If it didn't work this way, the user/programmer would think the function they called (the callee) had the problem.  It is telling them the problem is with the caller and not the callee.

As commented by Kacper, it is in the spec - section 11.5.7.

Thanks.

Blake

On Mon, Jul 14, 2014 at 9:15 AM, Juergen Sauermann <address@hidden> wrote:

Hi Blake,

I believe the ⎕ES related bugs reported recently should be fixed in SVN 372.

This one is not, though. The reason is that fixing it would have considerable performance
impacts (we would need a copy of every defined function argument only for the rare case that ⎕ES
is called),

I also find it somewhat arbitrary to show a non-locked function as locked in this very specific case.

Since ⎕ES is not ISO-standard anyhow, I take the freedom to trade performance for APL2 compatibility.

/// Jürgen

On 07/12/2014 06:34 PM, Blake McBride wrote:

      )CLEAR

CLEAR WS

      ∇test x

[1] x←'xx'

[2] ⎕ES 'SOME ERROR'

[3] ∇

      test '55'

SOME ERROR

      test 'xx'

      ^       ^

⎕ES is showing the modified and not the original argument passed to the function.  IBM APL 2 displays the original argument passed.

Thanks.

Blake