On Mon, Jun 7, 2010 at 3:23 PM, Stéphane Ducasse <stephane.ducasse@inria.fr> wrote:

yes this one looks good. but

       copyFrom: 1 to: does not convey that you should pass the second argument starting reverse.

Well, in ruby it's a general convention that negative indices start counting from the right. A reasonable implementation would change all sequenceable collections such that they accept negative indices. In Phexample syntax: ((1 to: 5) at: -2) should = 4. The alternative is to always have two accessors, then of course non-negative: (1 to: 5) atFromEnd: 2) should = 4. and: ('hello' from: 2 toFromEnd: 2) should = 'ell' I clearly prefer the negative indices. Things you do all the time can be much shorter than things you don't do so often. Huffman-encode your programming :) Cheers, Niko -- http://scg.unibe.ch/staff/Schwarz twitter.com/nes1983 Tel: +41 076 235 8683

If you want a chance to your idea propose some code and some tests H E L L O 1 2 3 4 5 -5 -4 -3 -2 -1 hello from: 2 to: -2 -> ell ok 'hello' from: 2 to: -2 -> ello 'hello' from: 2 to: 0 -> ??? This is strange because I thought intuitively and wrong that it would be means complete from: 2 toFromEnd: 2 from: 2 to: -2 So do you have other usage of negative index in ruby? Because I'm not sure I like that? Stef

...

copyFrom: 1 to: does not convey that you should pass the second argument starting reverse.

Well, in ruby it's a general convention that negative indices start counting from the right. A reasonable implementation would change all sequenceable collections such that they accept negative indices.

In Phexample syntax:

((1 to: 5) at: -2) should = 4.

The alternative is to always have two accessors, then of course non-negative:

(1 to: 5) atFromEnd: 2) should = 4.

and:

('hello' from: 2 toFromEnd: 2) should = 'ell'

I clearly prefer the negative indices. Things you do all the time can be much shorter than things you don't do so often. Huffman-encode your programming :)

Cheers,

Niko

-- http://scg.unibe.ch/staff/Schwarz twitter.com/nes1983 Tel: +41 076 235 8683

_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

On Jun 7, 2010, at 3:38 PM, Damien Pollet wrote:

On Mon, Jun 7, 2010 at 15:23, Stéphane Ducasse <stephane.ducasse@inria.fr> wrote:

...

copyFrom: 1 to: does not convey that you should pass the second argument starting reverse.

Of course, since it's the sign that does that…

...

copyFrom: 3 toFromEnd: -2

then, why -2 if it's explicitly from the end ???

ah why not.

-- Damien Pollet type less, do more [ | ] http://people.untyped.org/damien.pollet

_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

On 6/7/2010 5:03 AM, Lukas Renggli wrote:

...

Am I the only one who finds the ruby protocol for accessing collections richer at times? In Smalltalk, if ary is an array, how do I get ary without its last element?

anArray allButLast

Or for other amounts, allButLast: x There's also allButFirst and allButFirst: x. A look at the protocol of SequenceableCollection would have revealed this in seconds. -- Ramon Leon http://onsmalltalk.com

On 6/7/2010 12:17 PM, Niko Schwarz wrote:

Well, I'm just arguing that in Ruby, the simple convention that negative indices start from left to right map to a zoo of different methods in Smalltalk.

You didn't comment on:

('hello' from: 2 toFromEnd: 2) should = 'ell'

Would that find your blessing?

niko

What's the goal, to be more like Ruby? If that's the case I'm sure there's a thousand things you could change, but it's not Ruby, it's Smalltalk. Ruby seems to have a style where you can send just about anything to a method and it tries to figure out what to do based on the type of the arg, they like magic and the smaller API; Smalltalk tends to just have a dozen different but similarly named methods that let you accomplish the same thing. I prefer that latter style, less magic and more smaller simpler well factored methods with a larger API and less implicit assumptions. Personally, I don't like a lot of unspoken assumptions like negative indices meaning something that isn't stated in the method name, so yes I like... ('hello' from: 2 toFromEnd: 2) should = 'ell' better than using negative indices. Not sure I like #from:toFromEnd: as the selector name, but I like the spirit of it, it makes it explicit. I'd probably like #copyFrom:fromEnd: better, but that's just personal preference as it goes better with #copyFrom:to: which would seem to be a related method in that protocol. I really don't want to learn 10 different ways to use #at:, I'd rather see 10 similarly named methods each doing it one way that look like they're all variations of a theme. When someone comes up with a new idea like #copyFrom:fromEnd:, it's easier to add that than it is to go back and patch #at: to do something new, it allows the system to grow over time yet remain stable for existing code. -- Ramon Leon http://onsmalltalk.com

On Mon, Jun 7, 2010 at 10:29 PM, Ramon Leon <ramon.leon@allresnet.com> wrote:

 Ruby seems to have a style where you can send just about anything to a method and it tries to figure out what to do based on the type of the arg, they like magic and the smaller API; Smalltalk tends to just have a dozen different but similarly named methods that let you accomplish the same thing.

I like that analysis.

  so yes I like...

('hello' from: 2 toFromEnd: 2) should = 'ell'

:)

I really don't want to learn 10 different ways to use #at:, I'd rather see 10 similarly named methods each doing it one way that look like they're all variations of a theme.

That's a valid stand point to take. I guess, at some point you just need to choose what your use cases are / what your audience is. You can get all the positions of c that differ from a, in Matlab: c(a!=c). In smalltalk, this could be something like: c at: (a differencesVectorWith: c). Cheers, Niko -- http://scg.unibe.ch/staff/Schwarz twitter.com/nes1983 Tel: +41 076 235 8683

nope this is a primitive and we do not have multiple dispatch Stef On Jun 7, 2010, at 9:30 PM, Hilaire Fernandes wrote:

Niko Schwarz a écrit :

...

Just to put something on the table, how about: (ary at: 1, -2)

Or #at: could accept an interval as argument

_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project