[squeak-dev] Object>>printOn: refined.
marcel.taeumel at hpi.de
Thu Jun 4 08:53:20 UTC 2020
> You +1'd the idea of sending #asOop from #printOn:.
Nope. I +1'd the idea to be applied to #defaultLabelForInspector because of that "super" issue.
Am 03.06.2020 21:57:01 schrieb Chris Muller <ma.chris.m at gmail.com>:
It seems there are a couple of different dimensions to consider. You +1'd the idea of sending #asOop from #printOn:. I'm definitely a -1 on that, because it would force everywhere that doesn't want id to override printOn:, and they wouldn't even be able to call "super"! And even if they DO want id printing, overriding #asOop means they're forced to mix concatenation with streaming, which isn't a good idea. To stay with streaming, you would at least want a,#printOopOn:, but that name is too domain-specific. (#asOop is already a common selector in persistence frameworks, I think).
A generic, supplementary printing method that works on a Stream avoids those issues, and can be either a no-op in Object, OR, Object>>#printOn: could check if it #respondsTo: it, allowing the supplementary method to be provided on Object by external frameworks without dirtying the Kernel package.
Over the years, I've found having a separate, String-based "id" representation (that is not always it's oop / oid), and not any of its other attributes of its standard #printString, to be a useful extrusion of behavior for disparate domain hierarchies to inherit rather than each implement their own.
On Wed, Jun 3, 2020 at 3:24 AM Marcel Taeumel <marcel.taeumel at hpi.de [mailto:marcel.taeumel at hpi.de]> wrote:
> Seeing just their identifications removes a level of formality that seems to foster a better UI "connection" to the objects.
Domain-specific objects can provide a good-enough identification in their #printOn: specialization. Maybe the class name is part of it. Maybe not. If you develop a database, for example, you surely have another notion of identity. Meaning, other than Object >> #identityhash or #oop. That other notion of (compact) identity should then be part of an object's print-string that is part of the database.
Yet, nothing one can foresee, when thinking about Object >> #printOn:. There, you have only the meta-object protocol at hand: class name, identity hash, maybe ref counter? :-D Having a new subclass of Object, what should that #printOn: look like out-of-the-box? The class name is fine. So you can actually see that your new class' instances are populating the world. Next step should be to implement a custom #printOn: in that new class.
Am 03.06.2020 05:16:31 schrieb Chris Muller <asqueaker at gmail.com [mailto:asqueaker at gmail.com]>:
Of all the attributes that objects print, I find the "type" (class) to be one of the least interesting. It's usually already obvious in its contextual usage. Indeed, it is the "identity" that I, too, am interested in seeing printed. My solution since 2006 has been an override of #printOn: that adds a one-line dispatch to #printIdentificationOn:. I then take care in my #printIdentificationOn: implementations to keep the printing as terse as it can be, and sans any line endings, to serve this seemingly recurring use-case I have of wanting a "short version" of an object's string. For example, when printing the elements of a collection. Having entire domain hierarchies exclude the type entirely from their printString's has been a fantastic experience. Seeing just their identifications removes a level of formality that seems to foster a better UI "connection" to the objects.
On Tue, Jun 2, 2020 at 7:10 AM Trygve Reenskaug <trygver at ifi.uio.no [mailto:trygver at ifi.uio.no]> wrote:
I find it frustrating to open 3 inspectors on different objects, all of them titled 'aString' (or whatever),
IMO, it is much better to open them on the 3 objects:  aString,  a String, [4567 a String.
The numbers in square brackets stand for the objects oop, actually its identityHash. They can be a 7-digit numbers; much too long for my short-time memory to hold many of them. I therefore truncate the number to 4 digits, accepting that I may, in rare cases, get 2 objects with the same identifier.
I'm running 'Squeak5.3'.
"Append to the argument, aStream, a sequence of characters that identifies the receiver."
" The previous version identified the class, not the instance "
" This new version identifies the instance with its oop. "
" I arbitrarily truncate the oop to 4 digits to simplify reading. "
| title |
title := self class name.
nextPutAll: '[' , (self asOop printString truncateTo: 4) , ']' ;
nextPutAll: (title first isVowel ifTrue: ['an '] ifFalse: ['a ']);
The essence of object orientation is that objects collaborate to achieve a goal.
Trygve Reenskaug mailto: trygver at ifi.uio.no [mailto:%20trygver at ifi.uio.no]
Morgedalsvn. 5A http://folk.uio.no/trygver/ [http://folk.uio.no/trygver/]
N-0378 Oslo http://fullOO.info [http://fullOO.info]
Norway Tel: (+47) 468 58 625
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Squeak-dev