I presume you all know about Griggs work on the subject http://stackoverflow.com/questions/8204826/smalltalk-sort-a-collection-by-two-criteria http://objology.blogspot.com/2010/11/tag-sortfunctions.html
2014-03-12 11:23 GMT+01:00 carlo.t <snoob...@yahoo.ie>: > Hi Otto > > I think the main reason is to support the second usage scenario: > SequenceableCollection withAllSubclasses sortedAs: [:class | class > methodDict size]. > We could try and unify the sort: method to take only blocks while sortedAs: > a symbol but then at the same time we could just put everything into one > method and use cull: perhaps to handle all the scenarios. > > Personally I don't feel too comfortable with the code below, apart from the > performance penalties it also feels like code that is too 'clever' for > itself which obviously makes it complicated to understand and modify. I > wrote it mainly for people to voice opinions and get ideas flowing. I'd > have > 2 separate sort methods so that users can choose if they want the runtime > characteristics of either. > > Example which can handle all current scenarios i.e. symbol or block taking > 1 > or 2 arguments: > SequenceableCollection>>sort: aSortBlock > self > mergeSortFrom: 1 > to: self size > by: (self createSortBlock: aSortBlock) > > > SequenceableCollection>>createSortBlock: aSortBlockOrSymbol > ^ aSortBlockOrSymbol numArgs < 2 > ifTrue: [[ :a :b | (#(true false ) > includes: (aSortBlockOrSymbol > value: a)) > ifTrue: [ aSortBlockOrSymbol value: a] > ifFalse: [(aSortBlockOrSymbol value: a) < > (aSortBlockOrSymbol value: > b)]]] > ifFalse: [[:a :b | aSortBlockOrSymbol cull: a cull: b]] > > > BTW the previous versions had a bug where it fails when called on an empty > collection as the true/false check is done eagerly. > > Cheers > Carlo > > > > -- > View this message in context: > http://forum.world.st/Sort-by-property-tp4748500p4748757.html > Sent from the Pharo Smalltalk Developers mailing list archive at > Nabble.com. > >
