measuring namespace imports

The more I write Smalltalk in VW the more I'm bothered with my domain
vocabulary being 'taken' by the Smalltalk namespace.
Such simple things as modeling a Message or a Model too often gives me
the jarring experience of the name being 'taken'.

So what imports do we actually use in code?
I started to play in new namespaces _without_ any imports and started to
import only specific names instead of using 'Smalltalk.*'.
This led to the surprising result that domain models often need less
than 5  specific imports from outside the Soops namespace hierarchy.

Here is a little analysis script that measures uses of non-Soops names
from methods inside Soops namespaces.
I only measure references from methods because references for the
purpose of subclassing do not use the import mechanisms.


---------------------
ns := Soops.
refsByNamespace := Dictionary new.
refsByPackage := Dictionary new.
ns enumerateMethods:
         [:class :sel :compiled |
         compiled literals do:
                 [:lit |
                 (lit isVariableBinding
                     ifTrue: [lit binding environment notNil]
                     ifFalse: [lit isBindingReference and: [lit isDefined]])
                         ifTrue:
                             [(lit binding absoluteName
asQualifiedReference path
                                 startsWith: ns absoluteReference path)
                                     ifFalse:
                                         [(refsByNamespace at: class
environment ifAbsentPut: [Set new])
                                             add: lit binding.
                                         (refsByPackage at: class
package ifAbsentPut: [Set new]) add: lit binding]]]].
refsByNamespace inspect.
refsByPackage inspect
---------------------

It turns out that in our code there are many packages that only use a
handful of names from the libraries.

We already have the option to set a 'default namespace' on a package (to
allow resolutions unrelated to the imports of a class we might be
extending).
My experiments suggest that we do not need a default namespace on
packages but rather that we need to go one level deeper and have a
'default imports' section so I can declare the handfull of names the
package references instead of all-of-cincom-smalltalk.



Thoughts?

R
-



_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc