Hi Werner, I don't know what you are trying to achieve, but it looks like you want to do conversions. Have you considered this ? (Bag with: 1 with: 2) asArray. (Bag with: 1 with: 2) asOrderedCollection. These both work. I personally never noticed or used #newFrom: As for #as: some people consider it not the cleanest way to do things. But since the API is there, it should work I guess. Sven On 05 Mar 2014, at 16:28, Werner Kassens <wkassens@libello.com> wrote:
Hi,
1. i do this in pharo 2.0: b:=Bag newFrom: #(1 2). OrderedCollection newFrom: b."works nicely" Array newFrom: b."error" the error happens because ArrayedCollection>>newFrom: inherited by Array expects a SequenceableCollection that has 'at:' implemented. but the first line of ArrayedCollection>>newFrom: says: newFrom: aCollection it would be more clear,if it would say: newFrom: aSequenceableCollection
2. this also produces an error: b as: Array. because Object>>as: uses newFrom:. and since as: is a very general method, i'd guess that ArrayedCollection>>newFrom: should have an implementation that does not throw an exception. if changing a Bag to an OrderedCollection works, this should also work with an Array. or does this make much sense: <grin> (b as: OrderedCollection) as:Array "this way it works"
3. withAll: does essentially the same as newFrom: and has the same problem with the argument naming. but Array>>withAll: is much much faster than Array>>newFrom:. try: b:=Array new:100000. [Array withAll: b]timeToRun. [Array newFrom: b]timeToRun. perhaps one could differentiate the two methods in such a way, that withAll: is the occasionally faster method that for some collections only accepts aSequenceableCollection as argument and newFrom: accepts any aCollection. This way also Object>>as: would work as expected. if one does not want to change this in my opinion slightly inconsistent behaviour, then <grin> it would make sense to implement Array>>newFrom: aSequenceableCollection ^Array withAll: aSequenceableCollection. simply for speed reasons. and of course a consistent argument naming would be nice. werner