Hi,
in GNU APL you can have lambdas without result (eg. {}) and they don't
start with *λ*←.
As Kacper has pointed out you can also have user-defined functions
starting with *λ*←.
I have no plans to prevent that, but I have also no idea where the
lambdas will bring us
in GNU APL. Right now {...} is simply a wrapper for ⎕FX. IMHO it is not
a good idea to
have two kinds of functions with different rules in one language, even
if some APL programmers
seem to like that.
I have added a function UserFunction::is_lambda() in SVN 156 that tells
if the function
was *initially* defined via '{ ... }'.
No idea what happens if you edit the function later on.
/// Jürgen
On 03/06/2014 03:53 AM, Elias Mårtenson wrote:
The Emacs mode allows the user to edit an existing function by
pressing C-c C-f, or typing ∇ followed by the name of the function.
This opens up the function editor in a separate window.
However, when the user creates a function using the dfn notation:
avg ← {(+/⍵)÷⍴⍵}
And then tried to edit it using ∇avg, the Emacs mode will run the
equivalent of ⎕CR 'avg' behind the scenes, and present the following
in the function editor:
∇λ←avg ⍵
λ←(+/⍵)÷⍴⍵
This will of course fail when the user tries to save the function.
First a question for Jürgen:
*Can I rely on the fact that if the first character in a function
definition is **λ**, does that guarantee that the function is a
dfn?* A consequence of this is, for example, that I am guaranteed that
the function consists of only one statement.
So, I have a few options here and I would like people's input on it:
* I can simply pop up an error message if an attempt is made to edit
a lambda function (then how should they be edited?)
* I open the function editor in a special "lambda mode", limiting
the function to a single statement
* I replace λ, ⍺ and ⍵ with proper variables, thus converting the
lambda function into a normal function when saving.
* Jürgen allows the use of λ, ⍺ and ⍵ as variables in normal
functions so they behave just like any other function, except with
unusual parameter names.
*What solution do you think should be used?* The current state is
really not ideal.
Regards,
Elias
