The assert message can be an expression, so you could also separate out
the informative content from check_valid into another function and call it
to provide the second argument to assert.

On Tue, 26 Mar 2019 at 00:14, adrianv avm4@cornell.edu wrote:

nophead wrote

I can't think of a reason this would be useful.  I don't want to write:

assert(check_valid(parameter, big_long_list), generate_message(parameter,
parametername, big_long_list))

It means repeating code, which was what I wanted to avoid.  I suppose

let( result = check_valid(......) )
assert(result[0],result[1])

would be an option that gets the message entirely into the assert().

--
Sent from: http://forum.openscad.org/

Yes, or put the assert into check_valid() and rely on the the stack trace
to know where it is called from.

On Tue, 26 Mar 2019 at 09:52, adrianv avm4@cornell.edu wrote:

nophead wrote

If I do this then I need to use a dummy variable when I call assert, which
was basically the whole thing I was trying to work around.  If I could
define my own "function" that works like assert that would be the natural
way to do it, but assert is not a function and it works in a fashion that is
different from user definable commands.  If it's basically impossible to
avoid dummy variables then we should propose a dummy variable mechanism be
added to OpenSCAD.

--
Sent from: http://forum.openscad.org/

I don't understand where you need a dummy variable?

On Tue, 26 Mar 2019 at 12:03, adrianv avm4@cornell.edu wrote:

function check(x) = assert(x >= 0, str(x, " is not a square")) x;

function do_something(x) = sqrt(check(x));

echo(do_something(-4));

ERROR: Assertion '(x >= 0)' failed: "-4 is not a square" in file tests,
line 2

TRACE: called by 'check', in file tests, line 4.

TRACE: called by 'sqrt', in file tests, line 4.

TRACE: called by 'do_something', in file tests, line 6.

TRACE: called by 'echo', in file tests, line 6.

On Tue, 26 Mar 2019 at 12:05, nop head nop.head@gmail.com wrote:

Generally I think it's cryptic code and bad practice to mix the error
checking in with the computation, though to be honest, the example you have
below isn't terrible.

But in the case I was pondering, we are checking validity of user input, not
doing a computation.  The subsquent computation may be quite complex, and
there's no assignment that it makes sense to entangle the check with.  So
the code will end up looking like:

dummy = check_parameter(parameter, 'parameter',
["long","list","of","options"]),

somewhere in the let() statement.  I mean, I guess it could be

renamed_parameter = check_parameter(....),

and the check_parameter function returns the parameter value.  But being
compelled to rename all the parameters isn't ideal either.

nophead wrote

--
Sent from: http://forum.openscad.org/

It still works on Windows, so looks like a bug.

On Wed, 27 Mar 2019 at 18:49, corpsman corpsman@corpsman.de wrote:

On 27.03.19 20:06, nop head wrote:

Yes, but unfortunately it's a bit unclear what's wrong. Someone
tried to get it working again but there seems to be some driver
issues too -> https://github.com/openscad/openscad/pull/2466

ciao,
Torsten.