> Am 05.01.2015 um 14:43 schrieb Sebastian Sastre > <sebast...@flowingconcept.com>: > > one hugely typical case is having the model of an input that has either nil > because is pristine or an empty string and the app needs to guarantee some > default value that should not be nil or an empty string. > > Another frequent case is the response of some API that will typically answer > nil or an empty collection when something is not found and you want to > guarantee some value or model that should not be nil or an empty collection. > > About #thing being meaningless, sure, I’ve mentioned as general example. I > don’t see that every user of #thing has to use the ifNilOrEmpty:, only those > who care about guaranteeing that closure valued if none is found which is > expressed in the completely sensible form of receiving nil or an empty > collection :) > My point is that as long as you do not promise a certain type of object you will have to deal with the uncertainty what methods you can call on that object of uncertain type. By not using a check you just extend the life of this uncertainty a while longer (bad if the user of your code has to deal with it). Some has to deal with it if the object has to do something. And the earlier this uncertainty is removed the better it is. At least in my opinion.
Norbert > Thanks for giving it a thought > > > > >> On Jan 5, 2015, at 11:14 AM, Norbert Hartl <norb...@hartl.name >> <mailto:norb...@hartl.name>> wrote: >> >>> >>> Am 05.01.2015 um 14:01 schrieb Sebastian Sastre >>> <sebast...@flowingconcept.com <mailto:sebast...@flowingconcept.com>>: >>> >>> >>>> On Jan 5, 2015, at 10:38 AM, p...@highoctane.be >>>> <mailto:p...@highoctane.be> wrote: >>>> >>>> In business apps, the need for default values happen all the time, so the >>>> idiom has value (not sure for the message name though). >>> >>> Totally. In real apps, having to compare against uninitialized variable or >>> nil as response or empty string happens so often that having this method >>> makes it quite convenient (AKA lots of code becomes one-liners). >>> >>>> We could use >>>> >>>> x := [ self thing ] ifError: [ someDefault ] >>> >>> I understand you’re setting a similar, quite not like it example but in any >>> case this one raises and catches an exception and that sounds quite less >>> efficient if compared to return self (when object is not nil and is not an >>> empty collection/string) >>> >>>> for these purposes. Triggering errors is not too nice still. >>>> >>>> Now, what if self itself is nil or empty? >>>> >>>> BTW, isEmptyOrNil exists in the image for Collections and UndefinedObject. >>>> Empty has no meaning for Object, so why test against empty in the name? >>>> >>> Note that is not a testing method, it’s a conditional executor of the >>> closure. >>> The reason why was already mentioned, is to allow you to write this >>> one-liner convenience: >>> someVar := self thing ifNilOrEmpty: [blah] >>> >>> `self thing` could be an expensive process that returns something or nil or >>> an empty collection. If you get nil or empty as result then you would get >>> the block values resulting in having blah at someVar >>> >>> >>>> In the image, I see that we do have #default: anObject in several places. >>>> It seems to serve the same intent. >>>> >>>> What is the idiom for such things in Pharo? Importing idioms from other >>>> languages works but if we do have one already, we will introduce confusion. >>> >>> how can you do that one-liner without introducing ifNilOrEmpty: ? >>> >> What is #thing supposed to do? This whole problem looks like a typical >> javascript problem. You do anything and return anything and as all types are >> auto-coerced into their target type all expressions look like the same while >> meaning different things. >> It looks problematic to me to treat nil and empty collection the same. This >> might make sense in some business logic but not in general. In that move a >> method is added to Object using methods it cannot know of like #isEmpty. >> Object is no way more tied to Collection than it should be. >> Another problem is that #thing does return anything but nothing meaningful. >> So every user of #thing has to use the #ifNilOrEmpty: foo. This is probably >> something that needs to go into the class the implements #thing. Everything >> else is far from being an interface. >> Probably the solution to this is that #thing should return a concrete type >> object that can be used with its defined interface. So if having an >> one-liner is the ultimate goal one might need see the harm it produced on >> the way. >> >> my 2 cents, >> >> norbert >> >>>> >>>> >>>> Phil >>>> >>>> >>>> >>>> On Mon, Jan 5, 2015 at 1:19 PM, Tudor Girba <tu...@tudorgirba.com >>>> <mailto:tu...@tudorgirba.com>> wrote: >>>> This is not about taste. This is about not promoting the use of nil or >>>> dependency or the meaning of empty collection. >>>> >>>> A better way is to look at the upstream logic and modify that one so that >>>> it does not need to know about nil or empty. >>>> >>>> Cheers, >>>> Doru >>>> >>>> >>>> >>>> On Mon, Jan 5, 2015 at 1:17 PM, Sebastian Sastre >>>> <sebast...@flowingconcept.com <mailto:sebast...@flowingconcept.com>> wrote: >>>> taste is taste but would you care to illustrate your point with examples? >>>> I’m curious about it >>>> >>>> >>>> >>>> > On Jan 5, 2015, at 6:12 AM, stepharo <steph...@free.fr >>>> > <mailto:steph...@free.fr>> wrote: >>>> > >>>> > You summarise well the kind of code I do not like. >>>> > isNil everywhere and horrible tests. >>>> > >>>> > Stef >>>> > >>>> > >>>> > Le 4/1/15 23:27, Sebastian Sastre a écrit : >>>> >> Hi guys, >>>> >> >>>> >> I’ve started to use this little one: >>>> >> >>>> >> Object>>ifNilOrEmpty: aBlock >>>> >> >>>> >> self ifNil: [ ^ aBlock value ]. >>>> >> >>>> >> (self isCollection and: [ >>>> >> self isEmpty ]) ifTrue: [ ^ aBlock value ]. >>>> >> >>>> >> ^ self. >>>> >> >>>> >> >>>> >> It allows you to do the widely known JavaScript one-liner: >>>> >> >>>> >> var stuff = this.thing || ‘some default value for when this.thing is >>>> >> undefined, null or an empty string’. >>>> >> >>>> >> but in smalltalk in this way: >>>> >> >>>> >> stuff := self thing ifNilOrEmpty: [ ‘some default value for when self >>>> >> thing is nil or an empty string’ ] >>>> >> >>>> >> simple thing feels practical and nice :) >>>> >> >>>> >> >>>> >> >>>> > >>>> > >>>> >>>> >>>> >>>> >>>> >>>> -- >>>> www.tudorgirba.com <http://www.tudorgirba.com/> >>>> >>>> "Every thing has its own flow" >>>> >>>> >>>> >>>> -- >>>> --- >>>> Philippe Back >>>> Visible Performance Improvements >>>> Mob: +32(0) 478 650 140 | Fax: +32 (0) 70 408 027 >>>> Mail:p...@highoctane.be <mailto:mail%3ap...@highoctane.be> | Web: >>>> http://philippeback.eu <http://philippeback.eu/> >>>> Blog: http://philippeback.be <http://philippeback.be/> | Twitter: >>>> @philippeback >>>> Youtube: http://www.youtube.com/user/philippeback/videos >>>> <http://www.youtube.com/user/philippeback/videos> >>>> >>>> High Octane SPRL >>>> rue cour Boisacq 101 | 1301 Bierges | Belgium >>>> >>>> Pharo Consortium Member - http://consortium.pharo.org/ >>>> <http://consortium.pharo.org/> >>>> Featured on the Software Process and Measurement Cast - >>>> http://spamcast.libsyn.com <http://spamcast.libsyn.com/> >>>> Sparx Systems Enterprise Architect and Ability Engineering EADocX Value >>>> Added Reseller >
