Selectors as variables?

> Bruce,

> How would you do SUnit without perform:?

With Pragmas.

>  
> --
> Travis Griggs
> Objologist
> "You A students, you'll be back soon teaching here with me. You B
> students, you'll actually go on to be real engineers. You C
> students, you'll go into management and tell the A and B students
> what to do." - My Fluid Dynamics Professor whom I have yet to disprove

>  

You'd still need to invoke code somehow :)

-Boris

--
+1.604.689.0322
DeepCove Labs Ltd.
4th floor 595 Howe Street
Vancouver, Canada V6C 2T5
http://tinyurl.com/r7uw4

[hidden email]

CONFIDENTIALITY NOTICE

This email is intended only for the persons named in the message
header. Unless otherwise indicated, it contains information that is
private and confidential. If you have received it in error, please
notify the sender and delete the entire message including any
attachments.

Thank you.
disprove
>
> >
>
>

You can call #valueWithReceiver:arguments: on the actual compiled
method that contained the pragma, thus avoiding a perform:.

Is it clearer or any easier? .. nah, it's probably about the same.

> You'd still need to invoke code somehow :)

> -Boris

"I put forward the use of perform'ed symbols for the character scanners.
You've got me wondering if now if these tables wouldn't be better of as
tables of blocks. Since a block invokation is faster than a symbol
perform and that's a speed critical area."

Is it actually faster?

today := Core.Date today.
b := [:date | date year].
Array
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today
perform: #year]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:
today]])

#(181 288)
#(157 207)
#(183 214)

Cheers!

-Boris

--
+1.604.689.0322
DeepCove Labs Ltd.
4th floor 595 Howe Street
Vancouver, Canada V6C 2T5
http://tinyurl.com/r7uw4

[hidden email]

CONFIDENTIALITY NOTICE

This email is intended only for the persons named in the message
header. Unless otherwise indicated, it contains information that is
private and confidential. If you have received it in error, please
notify the sender and delete the entire message including any
attachments.

Thank you.
with
> the
> distinction he makes.  I think his thoughts on this are more
refined
> than mine. I shall steal them forthwith.
>
>
> Well, this is quite a refinement. Especially, since the original
poster
> did not make it clear that he was actually doing something like what
Boris
> opposed. For all we know, he was making an inspector, where the
selectors
> of a given protocol we're queried and then passed to this method to be
> executed with additional behavior around them (there's another
example:
> everytime you "dive" in trippy...).
>
>
> Perhaps we can give this topic a more extensive flogging at
StS07?
>
>
> Yes please. Let's do it right after the yearly STIC meeting. :)
>
> So here's an example of a place I used them once upon a time. I used
them
> for a small stateMachine. Where each state transition was encoded as
as a
> message send of #currentState_nextState and performed. It worked
> remarkably well. And was much easier to deal with the former version
that
> went overboard trying to make an object for each state. Much easier to
> write unit tests for.
>
> I put forward the use of perform'ed symbols for the character
scanners.
> You've got me wondering if now if these tables wouldn't be better of
as
> tables of blocks. Since a block invokation is faster than a symbol
perform
> and that's a speed critical area.
>
> And as you noted earlier... just about any sort of "serialization"
> mechanism is *easier* to do with perform:ed Symbols (though Blocks are
not
> impossible).
>
> --
> Travis Griggs
> Objologist
> "Is success the potential of what could be, or the reality of what
is?"
>

--
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.13/726 - Release Date: 18/03/2007 15.34

--
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.15/728 - Release Date: 20/03/2007 8.07

question? how many engineers at Cincom Smalltalk actually as a norm code  
and test on a MAC?  Travis I really hope you have a Win box :)


On Wed, 21 Mar 2007 13:27:40 -0500, Travis Griggs <[hidden email]>  
wrote:



--
Charles A. Monteiro
http://wiki.nycsmalltalk.org
http://www.monteirosfusion.com
http://monteirofusion.blogspot.com

Code on a Mac? I thought that was a deployment-only platform?

:)

Dave

Charles A. Monteiro wrote:

it better be , Wall Street ain't on MACs :), I doubt if Main St is either,  
Geek Street , no doubt

btw, I have a G3 I have been meaning to put Linux on :)

On Thu, 22 Mar 2007 11:27:32 -0500, Dave Stevenson <[hidden email]>  
wrote:



--
Charles A. Monteiro
http://wiki.nycsmalltalk.org
http://www.monteirosfusion.com
http://monteirofusion.blogspot.com

-----Ursprüngliche Nachricht-----
Von: Mark Plas [mailto:[hidden email]]
Gesendet: Donnerstag, 22. März 2007 11:06
An: vwnc
Betreff: RE: Selectors as variables?

Here are my results. A remark though: There is a difference in performance when you declare the variables ‘b’ and ‘today’. The first measurements are without variable declarations and executed in a workspace with ‘auto declaration’ turned on. In the 2^nd example, the variables are declared which makes quite an improvement.

 

today := Core.Date today.

b := [:date | date year].

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today

perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:

today]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value:

today]])  

 

#(361 343 251)

#(437 343 247)

#(461 339 243)

 #(368 340 240)

 

|today b|

today := Core.Date today.

b := [:date | date year].

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today

perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:

today]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value:

today]])

 

 #(253 158 153)

 #(306 155 154)

  #(332 159 160)

  #(333 164 245)

 

---

From: Björn Eiderbäck [mailto:[hidden email]]
Sent: donderdag 22 maart 2007 10:51
To: vwnc
Subject: RE: Selectors as variables?

 

I made the following change

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today

perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:

today]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value:

today]])

 

i.e. added the last test where b is inlined.

And got the following results from three runs:

#(381 620 438)

#(380 632 452)

#(383 627 457)

 

From: Giorgio Ferraris [mailto:[hidden email]]
Sent: den 21 mars 2007 19:19
To: 'Travis Griggs'; 'vwnc'
Subject: R: Selectors as variables?

 

Hi,

VW741

Version Id:

            #[60 41 60 64 74 1 0 0 60 41 60 64]

Windows xp sp2,

Pentium M 2.13.mhz

 

My result are:

 

#(363 304)

#(416 313)

#(351 314)

 

Much slower (so, what  cpu are you testing on?) and… perform slower than block!!!!

 

What happens?

 

Ciao

 

Giorgio

 

---

Da: Travis Griggs [mailto:[hidden email]]
Inviato: mercoledì 21 marzo 2007 0.00
A: vwnc
Oggetto: Re: Selectors as variables?

 

 

On Mar 20, 2007, at 14:16, Boris Popov wrote:

 

"I put forward the use of perform'ed symbols for the character scanners.

You've got me wondering if now if these tables wouldn't be better of as

tables of blocks. Since a block invokation is faster than a symbol

perform and that's a speed critical area."

 

Is it actually faster?

 

today := Core.Date today.

b := [:date | date year].

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today

perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:

today]])

 

Welll.... uh... duh. :)

 

I ASSumed that since the Symbol>>value: trick is about 60% as fast as the single arg block in an enumeration, that... On my machine, I ran 10x the iterations and reduced it to yourself sends. It's a wash on this box. But yes, there is no reason to use the blocks instead of a perform:'ed symbol. And I would argue that since there's no speed gain, the symbol dispatch table is far more concise then initializing it with blocks. Thanks for validating the original example. :)

 

--

Travis Griggs

Objologist

What's next, Intel Processors branded with "Apple Outside" stickers?

 

 

 

--
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.13/726 - Release Date: 18/03/2007 15.34

 

--
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.15/728 - Release Date: 20/03/2007 8.07

Intel Core Duo 2 / 6600 @ 2.4Ghz,

 

2 Processor(s) Installed.

[01]: x86 Family 6 Model 15 Stepping 6 GenuineIntel ~2392 Mhz

[02]: x86 Family 6 Model 15 Stepping 6 GenuineIntel ~2392 Mhz

 

today := Core.Date today.

b := [:date | date year].

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value: today]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value: today]])

 

#(169 210 144)

#(175 218 144)

#(161 219 151)

 

| today b |

today := Core.Date today.

b := [:date | date year].

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value: today]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value: today]])

 

#(88 102 92)

#(92 105 92)

#(93 101 92)

-Boris

--
+1.604.689.0322
DeepCove Labs Ltd.
4th floor 595 Howe Street
Vancouver, Canada V6C 2T5
http://tinyurl.com/r7uw4

[hidden email]

CONFIDENTIALITY NOTICE

This email is intended only for the persons named in the message
header. Unless otherwise indicated, it contains information that is
private and confidential. If you have received it in error, please
notify the sender and delete the entire message including any
attachments.

Thank you.

---

From: Nowak, Helge [mailto:[hidden email]]
Sent: Thursday, March 22, 2007 10:18 AM
To: Mark Plas; vwnc
Subject: AW: Selectors as variables?

 

Pentium 4 M, 1.80 GHz, 256GB RAM, Windows 2000:

 

today := Core.Date today.
b := [:date | date year].
Array
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform: #year]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value: today]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value: today]])  

#(529 631 386)

#(430 543 359)
#(450 616 378)

|today b|
today := Core.Date today.
b := [:date | date year].
Array
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform: #year]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value: today]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value: today]])  

#(280 270 287)

#(295 309 291)
#(280 348 227)

-----Ursprüngliche Nachricht-----
Von: Mark Plas [mailto:[hidden email]]
Gesendet: Donnerstag, 22. März 2007 11:06
An: vwnc
Betreff: RE: Selectors as variables?

Here are my results. A remark though: There is a difference in performance when you declare the variables ‘b’ and ‘today’. The first measurements are without variable declarations and executed in a workspace with ‘auto declaration’ turned on. In the 2^nd example, the variables are declared which makes quite an improvement.

 

today := Core.Date today.

b := [:date | date year].

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today

perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:

today]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value:

today]])  

 

#(361 343 251)

#(437 343 247)

#(461 339 243)

 #(368 340 240)

 

|today b|

today := Core.Date today.

b := [:date | date year].

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today

perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:

today]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value:

today]])

 

 #(253 158 153)

 #(306 155 154)

  #(332 159 160)

  #(333 164 245)

 

---

From: Björn Eiderbäck [mailto:[hidden email]]
Sent: donderdag 22 maart 2007 10:51
To: vwnc
Subject: RE: Selectors as variables?

 

I made the following change

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today

perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:

today]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value:

today]])

 

i.e. added the last test where b is inlined.

And got the following results from three runs:

#(381 620 438)

#(380 632 452)

#(383 627 457)

 

From: Giorgio Ferraris [mailto:[hidden email]]
Sent: den 21 mars 2007 19:19
To: 'Travis Griggs'; 'vwnc'
Subject: R: Selectors as variables?

 

Hi,

VW741

Version Id:

            #[60 41 60 64 74 1 0 0 60 41 60 64]

Windows xp sp2,

Pentium M 2.13.mhz

 

My result are:

 

#(363 304)

#(416 313)

#(351 314)

 

Much slower (so, what  cpu are you testing on?) and… perform slower than block!!!!

 

What happens?

 

Ciao

 

Giorgio

 

---

Da: Travis Griggs [mailto:[hidden email]]
Inviato: mercoledì 21 marzo 2007 0.00
A: vwnc
Oggetto: Re: Selectors as variables?

 

 

On Mar 20, 2007, at 14:16, Boris Popov wrote:

 

"I put forward the use of perform'ed symbols for the character scanners.

You've got me wondering if now if these tables wouldn't be better of as

tables of blocks. Since a block invokation is faster than a symbol

perform and that's a speed critical area."

 

Is it actually faster?

 

today := Core.Date today.

b := [:date | date year].

Array

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today

perform: #year]])

  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:

today]])

 

Welll.... uh... duh. :)

 

I ASSumed that since the Symbol>>value: trick is about 60% as fast as the single arg block in an enumeration, that... On my machine, I ran 10x the iterations and reduced it to yourself sends. It's a wash on this box. But yes, there is no reason to use the blocks instead of a perform:'ed symbol. And I would argue that since there's no speed gain, the symbol dispatch table is far more concise then initializing it with blocks. Thanks for validating the original example. :)

 

--

Travis Griggs

Objologist

What's next, Intel Processors branded with "Apple Outside" stickers?

 

 

 

--
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.13/726 - Release Date: 18/03/2007 15.34

 

--
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.15/728 - Release Date: 20/03/2007 8.07

--
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.15/728 - Release Date: 20/03/2007 8.07

--
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.15/728 - Release Date: 20/03/2007 8.07

Giorgio,

Can you try these as well please?

today := Core.Date today.
b := [:date | date year].
Array
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform: #year]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value: today]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value: today]])

 
#(169 210 144)
#(175 218 144)
#(161 219 151)


| today b |
today := Core.Date today.
b := [:date | date year].
Array
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform: #year]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value: today]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date year] value: today]])
 

#(88 102 92)
#(92 105 92)
#(93 101 92)

-Boris

--
+1.604.689.0322
DeepCove Labs Ltd.
4th floor 595 Howe Street
Vancouver, Canada V6C 2T5
http://tinyurl.com/r7uw4

[hidden email]

CONFIDENTIALITY NOTICE

This email is intended only for the persons named in the message
header. Unless otherwise indicated, it contains information that is
private and confidential. If you have received it in error, please
notify the sender and delete the entire message including any
attachments.

Thank you.

Yes, Shure, Boris

today := Core.Date today.
b := [:date | date year].
Array
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform:
#year]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:
today]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date
year] value: today]])
 #(337 319 224)
 #(399 315 224)
#(398 319 227)


| today b |
today := Core.Date today.
b := [:date | date year].
Array
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform:
#year]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:
today]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date
year] value: today]])  
 #(244 145 143)
#(292 151 143)
 #(329 147 142)


Still first is slower than second, opposite to your!
Ciao

Giorgio



-----Messaggio originale-----
Da: Boris Popov [mailto:[hidden email]]
Inviato: giovedì 22 marzo 2007 23.06
A: Giorgio Ferraris; Travis Griggs; vwnc
Oggetto: RE: R: Selectors as variables?

Giorgio,

Can you try these as well please?

today := Core.Date today.
b := [:date | date year].
Array
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform:
#year]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:
today]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date
year] value: today]])

 
#(169 210 144)
#(175 218 144)
#(161 219 151)


| today b |
today := Core.Date today.
b := [:date | date year].
Array
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [today perform:
#year]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [b value:
today]])
  with: (Core.Time millisecondsToRun: [10000000 timesRepeat: [ [:date | date
year] value: today]])
 

#(88 102 92)
#(92 105 92)
#(93 101 92)

-Boris

--
+1.604.689.0322
DeepCove Labs Ltd.
4th floor 595 Howe Street
Vancouver, Canada V6C 2T5
http://tinyurl.com/r7uw4

[hidden email]

CONFIDENTIALITY NOTICE

This email is intended only for the persons named in the message
header. Unless otherwise indicated, it contains information that is
private and confidential. If you have received it in error, please
notify the sender and delete the entire message including any
attachments.

Thank you.

--
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.15/728 - Release Date: 20/03/2007
8.07
 

--
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.15/728 - Release Date: 20/03/2007
8.07