Improper script count in symbol - function signature?

[Grégory Vanuxem]

Hello,

I have two very simple domains, an anonymous function domain and a
named function domain, but one barfs with a strange error, see below.
Any idea what that means? The anonymous function domain "works" fine.
Not the other one. The code is almost the same, only the difference
AnonymousFunction//FunctionCalled(Symbol).

(1) -> )di op jfunc

There are 2 exposed functions called jfunc :
[1] String -> JLObjFunction from JLObjFunction

>> Error detected within library code:
Improper script count in symbol

(1) -> )sh JLObjFunction
JLObjFunction is a domain constructor.
Abbreviation for JLObjFunction is JOBFUNC
This constructor is exposed in this frame.
25 names for 36 operations in this domain.

>> Error detected within library code:
Improper script count in symbol

The code is at a very early stage but I wonder what this error means.
The file (domains) compiles fine.

Greg

PS : Just in case, attached are the two (very) simple domains, they
use Julia behind the scenes. I don't know if this file is compilable
in pure FriCAS.

[Waldek Hebisch]

On Mon, Feb 23, 2026 at 11:04:04PM +0100, Grégory Vanuxem wrote:
> Hello,
>
> I have two very simple domains, an anonymous function domain and a
> named function domain, but one barfs with a strange error, see below.
> Any idea what that means? The anonymous function domain "works" fine.
> Not the other one. The code is almost the same, only the difference
> AnonymousFunction//FunctionCalled(Symbol).
>
> (1) -> )di op jfunc
>
> There are 2 exposed functions called jfunc :
> [1] String -> JLObjFunction from JLObjFunction
>
> >> Error detected within library code:
> Improper script count in symbol

You have:

jfunc : FunctionCalled(Symbol) -> %

This causes trouble: argument to 'FunctionCalled' should be a symbol,
but you pass 'Symbol' which is a type (and not a symbol). Such
mismatch is a weakness in current FriCAS typechecker, checks that
arguments to types have correct type are delayed and some things
(like this one) go trough unchecked.

In your command display code assumes that you have correct type,
and seeinf 'FunctionCalled' tries to convert its argument to
OutputForm. So, FriCAS calls "coerce : Symbol -> OutputForm'.
But 'Symbol' is internally represented by Lisp list '(Symbol)'.
FriCAS symbol handling code assumes that this is a scripted
symbol, but then realizes that there are no scipts. So you
get appropriate error message.

BTW: 'FunctionCalled' and few similar things are part of interpeter
type inference/coercion machinery. Namely, 'FunctionCalled'
is internal representation of effectively untyped function.
This internal representation will give true function only after
coercion to appropriate function type (such coercions are done
behind the scenes by the interpreter). Normally such types should
not be used by algebra functions. You may have reasons to use this
type, but it is awkward to use from algebra: either you hardcode
name (and hope that you are not affected by troubles due to
types in signatures) or parameter to 'FunctionCalled' must be
a parameter to your domain. Normally FriCAS will not invent
parameters to domains, so you will need a lot of explicit
qualification when using your constructor. And you will have
to deal with internal representation.

--
Waldek Hebisch

[Grégory Vanuxem]

Hi,

So you're suggesting to not use FunctionCalled, just keeping
AnonymousFunction? Since I hold a reference in Julia to anonymous
functions, at first glance that seems not be a problem. In fact I just
wanted to have support of the two for completeness. I think I need
more thought about this. BTW thanks for your insight.

I will still test a parameter to my named function domain. Don't know
as of now how I will do it. This is something I'll be following up on.

Greg

> --
> You received this message because you are subscribed to the Google Groups "FriCAS - computer algebra system" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to fricas-devel...@googlegroups.com.
> To view this discussion visit https://groups.google.com/d/msgid/fricas-devel/aZ0cs7cSqZuhHwIV%40fricas.org.

[Grégory Vanuxem]

Hello,

Le mar. 24 févr. 2026 à 04:36, Waldek Hebisch <de...@fricas.org> a écrit :
>
> On Mon, Feb 23, 2026 at 11:04:04PM +0100, Grégory Vanuxem wrote:
> > Hello,
> >
> > I have two very simple domains, an anonymous function domain and a
> > named function domain, but one barfs with a strange error, see below.
> > Any idea what that means? The anonymous function domain "works" fine.
> > Not the other one. The code is almost the same, only the difference
> > AnonymousFunction//FunctionCalled(Symbol).
> >
> > (1) -> )di op jfunc
> >
> > There are 2 exposed functions called jfunc :
> > [1] String -> JLObjFunction from JLObjFunction
> >
> > >> Error detected within library code:
> > Improper script count in symbol
>
> You have:
>
> jfunc : FunctionCalled(Symbol) -> %
>
> This causes trouble: argument to 'FunctionCalled' should be a symbol,
> but you pass 'Symbol' which is a type (and not a symbol). Such
> mismatch is a weakness in current FriCAS typechecker, checks that
> arguments to types have correct type are delayed and some things
> (like this one) go trough unchecked.

Solved! Thanks. And shame on me. I've encountered this too when I have
to implement coercing an "indeterminate" in a Nemo univariate
polynomial domain (Variable(x)). Adding a Symbol domain parameter and
modifying the relevant domain signatures accordingly solved it. I just
need to look more at the 'coerce' function but the main point is
solved. Yes!

(1) -> f(x,y)== cos(x)*sin(y/7)
Type: Void
Time: 0 sec
(2) -> g:=jfunc(f)

(2) f (generic function with 1 method)
Type: JLObjFunction(f)
Time: 0 sec
(3) -> jlApply(g,jobject("3.2"),2)

(3) -0.2813622468686953
Type: JLObject
Time: 0.05 (EV) = 0.05 sec
(4) -> jlType %

(4) Float64
Type: Symbol
Time: 0.01 (OT) = 0.01 sec

Greg