10 a.m.
Hi,
I'm wondering why sometimes there exist informative error messages but
it doesn't seem easy to catch them. Here's a concrete (and
real-world-inspired) use case:
? m = feval("haha")
haha: function not found
That's exactly the right error message! However, now I try to retrieve
that programmatically:
? catch m = feval("haha")
? eval errmsg($error)
Datenfehler # ( = data error)
This is much less useful. Could the precise message be pushed through to
the error? Hm, I guess there aren't that many formally encoded error
types...
And BTW, here's a different and a little ironic example:
? m = I(2,3,4)
I: zuviele Argumente # (= too many arguments)
? catch m = I(2,3,4)
I: zuviele Argumente
? eval errmsg($error)
Befehl hat zu wenige Argumente # (= command with too *few* arguments)
So there's a nice contradiction... In addition, I see that even the line
with 'catch' does print out the error, whereas I expected it to be
silent in that case, no?
(I also thought that maybe my German translation was simply wrong and
switched to English. I still got "Command has insufficient arguments"
for the errmsg.)
This is with the March 30th snapshot on Win10.
thanks
sven
9:23 p.m.
On Mon, Apr 3, 2023 at 6:00 AM Sven Schreiber
<sven.schreiber(a)fu-berlin.de> wrote:
I'm wondering why sometimes there exist informative error messages but it doesn't
seem easy to catch them. Here's a concrete (and real-world-inspired) use case:
? m = feval("haha")
haha: function not found
That's exactly the right error message! However, now I try to retrieve that
programmatically:
? catch m = feval("haha")
? eval errmsg($error)
Datenfehler # ( = data error)
This is much less useful. Could the precise message be pushed through to the error? Hm, I
guess there aren't that many formally encoded error types...
The trouble is that an error caught by "catch" is by construction not
an error. When we're responding to errmsg($error) all we have to show
is the "boilerplate" message associated with the error code that was
set, then removed by "catch". But there may be a way to improve on
that...
And BTW, here's a different and a little ironic example:
? m = I(2,3,4)
I: zuviele Argumente # (= too many arguments)
? catch m = I(2,3,4)
I: zuviele Argumente
? eval errmsg($error)
Befehl hat zu wenige Argumente # (= command with too *few* arguments)
OK, that was a case where the wrong internal error code, E_ARGS, was
set in a case of too many arguments; E_INVARG is now substituted, with
a generic message of "Invalid argument" (not very specific, but not
flat-out wrong).
Allin
9:23 a.m.
Am 03.04.2023 um 12:00 schrieb Sven Schreiber:
. In addition, I see that even the line with 'catch' does
print out
the error, whereas I expected it to be silent in that case, no?
BTW, it seems that sometimes it's silent, and sometimes not. After a
fresh restart of gretl, I enter "catch matrix m = I(2,3,4)" in the
console, and I see the message concerning "too many arguments". I can do
this repeatedly, so it's not a matter of newly defining m, it seems.
Then I run a sample script from a function package, go back to the
console, and execute "catch matrix m = I(2,3,4)" again - I silently get
a new prompt.
To be clear, the silent behavior is what I'd expect when using catch,
not the printed error message.
thanks
sven
668
days inactive
669
days old
2 comments
2 participants
participants (2)
-
Cottrell, Allin -
Sven Schreiber