Smalltalk › Squeak › Squeak - Dev

#<-

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

18 messages Options

Christoph Thiede

#<-

Hi all,

Vivide has the following small Object extension:

<- anObject

^ anObject -> self

I find it quite useful for reasons of symmetry.

Killer feature:

#(one two three) withIndexCollect: #<-

How would you think about integrating it into Trunk? :-)

Best,

Christoph

Carpe Squeak!

Jakob Reschke

Re: #<-

Thiede, Christoph <[hidden email]> schrieb am So., 15. Dez. 2019, 12:14:

Killer feature:

#(one two three) withIndexCollect: #<-

I find this hard to read. You have to wrap your head around it to understand what happens.

In which situations is <- more practical than ->? Can you give more examples?

Jakob Reschke

Re: #<-

Sometimes I think one should separate Smalltalk the programming language from Smalltalk as a scripting language (in Vivide or your spreadsheets, Christoph). I find such brevity cool for the latter but I despise it for the former.

Jakob Reschke <[hidden email]> schrieb am So., 15. Dez. 2019, 12:36:

Thiede, Christoph <[hidden email]> schrieb am So., 15. Dez. 2019, 12:14:

Killer feature:

#(one two three) withIndexCollect: #<-

I find this hard to read. You have to wrap your head around it to understand what happens.

In which situations is <- more practical than ->? Can you give more examples?

Stéphane Rollandin

Re: #<-

In reply to this post by Christoph Thiede

> <- anObject
> ^ anObject -> self

Please don't. The <- selector can be useful in other contexts (I use in
muO for example, in very specific classes).

Stef

David T. Lewis

Re: #<-

In reply to this post by Christoph Thiede

On Sat, Dec 14, 2019 at 07:59:43PM +0000, Thiede, Christoph wrote:

> Hi all,
>
>
> Vivide has the following small Object extension:
>
>
> <- anObject
> ^ anObject -> self
>
>
> I find it quite useful for reasons of symmetry.
>
>
> Killer feature:
>
> #(one two three) withIndexCollect: #<-
>
>
> How would you think about integrating it into Trunk? :-)
>

From the perspective of reading and cemprehension, I find it
easier to read this:

#(one two three) withIndexCollect: [ :e :i | i -> e ]

Why? Because I already understand what is meant by #-> and
I can directly read the expression without needing to check
the meaning of #<-. So for me it is slightly more verbose,
but faster to read and easier to understand.

Also, while it is easy to add something to Object, it is
extremely difficult to take it out later. Nice features like
this tend to accumulate over time, and eventually make the
system harder to understand. Have a look at Morph for a
notorious example :-)

Dave

timrowledge

Re: #<-

In reply to this post by Jakob Reschke

> On 2019-12-15, at 3:38 AM, Jakob Reschke <[hidden email]> wrote:
>
> Sometimes I think one should separate Smalltalk the programming language from Smalltalk as a scripting language (in Vivide or your spreadsheets, Christoph). I find such brevity cool for the latter but I despise it for the former.

Also +1. What next -
1 value: 4 value: 'two'
being handled by converting the 1 to #one, sending #isKindOf: to the second arg and then coercively converting the first argument to a compatible class via an indirection dictionary provided by a database lookup from the receiver?

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Useful random insult:- Ought to have a warning label on his forehead.

Christoph Thiede

Re: #<-

Hi, thanks for your arguments!

I think that #<- would be symmetric to #->, such as #< and #>, etc. ...

Another example would be [Dictionary newFrom: {1 <- #one. 2 <- #two. 3 <- #three}], but I agree, readability is quite subjective here.

From my point of view, Stef, implementing #<- with a completely different meaning would be confusing. I guess this depends on your specific domain.

@Jakob:

> Sometimes I think one should separate Smalltalk the programming language from Smalltalk as a scripting language (in Vivide or your spreadsheets, Christoph). I find such brevity cool for the latter but I despise it for the former.

Interesting statement! From a scripting perspective, it would be great to have a maximum number of convenient utils. However, I find it always confusing when you start building a solution via scripting and then need to rephrase your code because scripting syntax is not available for real programs. This compromises the idea of the One Uniform Smalltalk.

Best,

Christoph

Von: Squeak-dev <[hidden email]> im Auftrag von tim Rowledge <[hidden email]>
Gesendet: Sonntag, 15. Dezember 2019 19:59:19
An: The general-purpose Squeak developers list
Betreff: Re: [squeak-dev] #<-

Carpe Squeak!

Stéphane Rollandin

Re: #<-

> From my point of view, Stef, implementing #<- with a completely
> different meaning would be confusing.

This is really a personal preference. From my own point of view, I do
not see why a formal symmetry in a method selector should imply a
symmetry in the method semantics.

For example, would you implement #\ in Number, so that 2\3 is 3/2 ?

Anyway my actual point is that what you propose here is that #<- would
be implemented in Object, making it unavailable to all classes in the
system.

This represents a very important cost for no new semantics, just
syntactic sugar.

Stef

Levente Uzonyi

Re: #<-

In reply to this post by Christoph Thiede

>
> Hi, thanks for your arguments!
>
>
> I think that #<- would be symmetric to #->, such as #< and #>, etc. ...
>
> Another example would be [Dictionary newFrom: {1 <- #one. 2 <- #two. 3 <- #three}], but I agree, readability is quite subjective here.
>
>
> From my point of view, Stef, implementing #<- with a completely different meaning would be confusing. I guess this depends on your specific domain.
>
>
> @Jakob:
>
> > Sometimes I think one should separate Smalltalk the programming language from Smalltalk as a scripting language (in Vivide or your spreadsheets, Christoph). I find such brevity cool for the latter but I despise it for the
> former.
>
>
> Interesting statement! From a scripting perspective, it would be great to have a maximum number of convenient utils. However, I find it always confusing when you start building a solution via scripting and then need to
> rephrase your code because scripting syntax is not available for real programs. This compromises the idea of the One Uniform Smalltalk.
>
> Best,
> Christoph
>
> _________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> Von: Squeak-dev <[hidden email]> im Auftrag von tim Rowledge <[hidden email]>
> Gesendet: Sonntag, 15. Dezember 2019 19:59:19
> An: The general-purpose Squeak developers list
> Betreff: Re: [squeak-dev] #<-
>
>
> > On 2019-12-15, at 3:38 AM, Jakob Reschke <[hidden email]> wrote:
> >
> > Sometimes I think one should separate Smalltalk the programming language from Smalltalk as a scripting language (in Vivide or your spreadsheets, Christoph). I find such brevity cool for the latter but I despise it for the
> former.
>
> Also +1. What next -
> 1 value: 4 value: 'two'
> being handled by converting the 1 to #one, sending #isKindOf: to the second arg and then coercively converting the first argument to a compatible class via an indirection dictionary provided by a database lookup from the
> receiver?
>
> tim
> --
> tim Rowledge; [hidden email]; http://www.rowledge.org/tim
> Useful random insult:- Ought to have a warning label on his forehead.
>
>
>
>
>

Christoph Thiede

Re: #<-

I use associations with #-> quite often - for example, with Dictionary class >> #newFrom::

Dictionary newFrom: {1 -> #one. 2 -> #two}

In Contrast to #at:put:, the above version checks that each key is unique.

Best,

Christoph

Von: Squeak-dev <[hidden email]> im Auftrag von Levente Uzonyi <[hidden email]>
Gesendet: Montag, 16. Dezember 2019 15:52:12
An: The general-purpose Squeak developers list
Betreff: Re: [squeak-dev] #<-

I think #-> was not a good idea. I personally use it fairly rarely, and
it's just a shorthand variant of Association>>key:value:.
It was probably more useful when brace arrays were not part of the
language, but I find the latter way superior for creating tuples.

Levente

On Mon, 16 Dec 2019, Thiede, Christoph wrote:

>
> Hi, thanks for your arguments!
>
>
> I think that #<- would be symmetric to #->, such as #< and #>, etc. ...
>
> Another example would be [Dictionary newFrom: {1 <- #one. 2 <- #two. 3 <- #three}], but I agree, readability is quite subjective here.
>
>
> From my point of view, Stef, implementing #<- with a completely different meaning would be confusing. I guess this depends on your specific domain.
>
>
> @Jakob:
>
> > Sometimes I think one should separate Smalltalk the programming language from Smalltalk as a scripting language (in Vivide or your spreadsheets, Christoph). I find such brevity cool for the latter but I despise it for the
> former.
>
>
> Interesting statement! From a scripting perspective, it would be great to have a maximum number of convenient utils. However, I find it always confusing when you start building a solution via scripting and then need to
> rephrase your code because scripting syntax is not available for real programs. This compromises the idea of the One Uniform Smalltalk.
>
> Best,
> Christoph
>
> _________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> Von: Squeak-dev <[hidden email]> im Auftrag von tim Rowledge <[hidden email]>
> Gesendet: Sonntag, 15. Dezember 2019 19:59:19
> An: The general-purpose Squeak developers list
> Betreff: Re: [squeak-dev] #<-
>
>
> > On 2019-12-15, at 3:38 AM, Jakob Reschke <[hidden email]> wrote:
> >
> > Sometimes I think one should separate Smalltalk the programming language from Smalltalk as a scripting language (in Vivide or your spreadsheets, Christoph). I find such brevity cool for the latter but I despise it for the
> former.
>
> Also +1. What next -
> 1 value: 4 value: 'two'
> being handled by converting the 1 to #one, sending #isKindOf: to the second arg and then coercively converting the first argument to a compatible class via an indirection dictionary provided by a database lookup from the
> receiver?
>
> tim
> --
> tim Rowledge; [hidden email]; http://www.rowledge.org/tim
> Useful random insult:- Ought to have a warning label on his forehead.
>
>
>
>
>

Carpe Squeak!

Jakob Reschke

Re: #<-

In reply to this post by Christoph Thiede

Thiede, Christoph <[hidden email]> schrieb am Mo., 16. Dez. 2019, 12:16:

@Jakob:

> Sometimes I think one should separate Smalltalk the programming language from Smalltalk as a scripting language (in Vivide or your spreadsheets, Christoph). I find such brevity cool for the latter but I despise it for the former.

Interesting statement! From a scripting perspective, it would be great to have a maximum number of convenient utils. However, I find it always confusing when you start building a solution via scripting and then need to rephrase your code because scripting syntax is not available for real programs. This compromises the idea of the One Uniform Smalltalk.

And introducing all the scripting shortcuts to the main language compromises the idea of the well readable source code ;-) whenever I look at shell script or Perl I am reminded that they are apparently not made to be read by humans.

For example, I again found your Dictionary newFrom: { 1 <- #one. 2 <- #two } very confusing. Keys usually come first in the Smalltalk "literature" (other code), now whaaat??

In my opinion, it is like allowing people to occasionally write English right-to-left if the author thinks it is appropriate.

timrowledge

Re: #<-

> On 2019-12-16, at 9:57 AM, Jakob Reschke <[hidden email]> wrote:
>

>
> And introducing all the scripting shortcuts to the main language compromises the idea of the well readable source code ;-) whenever I look at shell script or Perl I am reminded that they are apparently not made to be read by humans.

Exactly. If anyone goes down the path of making Smalltalk accept nonsense like that, or makes it look like APL etc, then I will reach down the internet and flick their earlobes until they scream uncle.

>
> For example, I again found your Dictionary newFrom: { 1 <- #one. 2 <- #two } very confusing. Keys usually come first in the Smalltalk "literature" (other code), now whaaat??
>
> In my opinion, it is like allowing people to occasionally write English right-to-left if the author thinks it is appropriate.
>
puY.

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Receptacle: a very playful welcome from a receptionist

Nicolas Cellier

Re: #<-

I don't buy #<- neither.

We don't expect symmetry here probably because keys and values have very asymetrical role.

Le lun. 16 déc. 2019 à 21:45, tim Rowledge <[hidden email]> a écrit :

> On 2019-12-16, at 9:57 AM, Jakob Reschke <[hidden email]> wrote:
>

>
> And introducing all the scripting shortcuts to the main language compromises the idea of the well readable source code ;-) whenever I look at shell script or Perl I am reminded that they are apparently not made to be read by humans.

Exactly. If anyone goes down the path of making Smalltalk accept nonsense like that, or makes it look like APL etc, then I will reach down the internet and flick their earlobes until they scream uncle.

>
> For example, I again found your Dictionary newFrom: { 1 <- #one. 2 <- #two } very confusing. Keys usually come first in the Smalltalk "literature" (other code), now whaaat??
>
> In my opinion, it is like allowing people to occasionally write English right-to-left if the author thinks it is appropriate.
>
puY.

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Receptacle: a very playful welcome from a receptionist

Squeak - Dev mailing list

Re: #<-

+1000

I can't believe we're having a discussion about that nonsense!

On 2019-12-16 16:08, Nicolas Cellier wrote:

> I don't buy #<- neither.
> We don't expect symmetry here probably because keys and values have
> very asymetrical role.

--
-----------------
Beno??t St-Jean
Yahoo! Messenger: bstjean
Twitter: @BenLeChialeux
Pinterest: benoitstjean
Instagram: Chef_Benito
IRC: lamneth
GitHub: bstjean
Blogue: endormitoire.wordpress.com
"A standpoint is an intellectual horizon of radius zero". (A. Einstein)

Christoph Thiede

Re: #<-

Sorry for wasting your time - I found it as always very interesting to hear your arguments to such design questions. I understand that the side effects of such a selector would probably outweigh the scripting convenience. Thanks for your patience!

Best,

Christoph

Von: Squeak-dev <[hidden email]> im Auftrag von Benoit St-Jean via Squeak-dev <[hidden email]>
Gesendet: Montag, 16. Dezember 2019 22:16:43
An: [hidden email]
Betreff: Re: [squeak-dev] #<-

Carpe Squeak!

David T. Lewis

Re: #<-

On Mon, Dec 16, 2019 at 10:15:28PM +0000, Thiede, Christoph wrote:

> Sorry for wasting your time - I found it as always very interesting to
> hear your arguments to such design questions. I understand that the
> side effects of such a selector would probably outweigh the scripting
> convenience. Thanks for your patience!
>
>
> Best,
>
> Christoph
>

Don't apologize, you are exploring and asking interesting questions.
The rest of us old guys can stop debating any time we want to ;-)

Dave

Stéphane Rollandin

Re: #<-

> Don't apologize, you are exploring and asking interesting questions.

Agreed. It's nice to have some fresh brainstorming at times...

Stef

marcel.taeumel

Re: #<-

This is how I make use of #<- in Vivide:

Why this effort?

1) The tool builder can concisely express read/write operations on method objects for a #text property.

2) Interactive views - such as text fields - can just call "model node text: newText" to trigger -- here -- the recompilation of that method.

Here is more: https://github.com/hpi-swa/vivide

Best,

Marcel

Am 17.12.2019 09:26:23 schrieb Stéphane Rollandin <[hidden email]>:
> Don't apologize, you are exploring and asking interesting questions.

Agreed. It's nice to have some fresh brainstorming at times...

Stef