Dear Victor,
yes, this looks nice. It is premature to add it to the set of builtins, i have not had much uses cases for this functionality. But as you say, it can be easily scripted (see below).... or provided as and optional package.
This functionality is even more enhanced with mixins and superclasses where each chained method would be passed to a mixin/superclass via [next].
one could pass in the list of methods as well per-object mixins, which could be added before and removed after the invocation. however, this would make invocation less elegant, and i doubt that there are many use cases for this.
It is not fully clear what's the best way to return the results or errors of the multiple method invocations. The small implementation below returns a list of results of the individual methods, and does nothing about errors.
we plan on the new web site a community space, where can place some implementation studies like this one.
Btw., the "opposite" problem is to call a single method with a set of argument vectors (the method is called repeatedly with each element of the set). There is an implementation of that in the new regression test using the tcl8.6 function apply (similar in length to the example below).
-gustaf neumann
================================= package require nx
nx::Object public method chain args { set methods [lrange $args 0 end-1] set arguments [lindex $args end] set result [list] foreach m $methods { lappend result [$m {*}$arguments] } return $result }
Class create MyClass { :method a {x y z} { return "a $x" } :method b {x y z} { return "b $y" } :create m1 }
puts [::m1 chain :a :b {1 2 3}]