On Mon, 16 Feb 2026, Sven Schreiber wrote:
Am 16.02.2026 um 15:43 schrieb Cottrell, Allin:
> On Mon, Feb 16, 2026 at 5:13 AM Sven Schreiber
> <sven.schreiber(a)fu-berlin.de> wrote:
>> Am 16.02.2026 um 11:00 schrieb Riccardo (Jack) Lucchetti:
>>> Well, the variable "hey" could well be a string, in the context of
the
>>> ui-maker function, containing the value "hey". Just sayin'
>>>
>> But in the context of _(depends=hey) in the ui-maker function, the
>> unquoted symbol 'hey' is undefined. (It's an argument in a different
>> function.) In that sense it's even weirder that there's no error
message.
> I think Jack's point stands. In your ui-maker 'hey' is undefined, but
> it _could_ be defined: for instance, one might make the dependency
> parameter string conditional on some feature of the dataset.
I'm not disputing that it _could_ be defined, but it just isn't, in my case
(and most similar other cases, I'd argue) - at least not in the current
function scope. So what I don't understand, why no error?
It would be nice if errors of this sort could be caught at the stage
of creating a gfn file but that would not be very easy. The
alternative is to flag an error when the ui-maker is actually called
upon.
What we do at present is issue an error message on stderr. The
thinking behind that is that ui-maker functionality is a refinement
which is not essential to getting results from a function package,
so we'd rather not abort on a ui-maker failure. But a function
writer can/should pay attention to stderr when debugging a package.
Allin