Smalltalk › Squeak › Squeak - Dev

[squeak-dev] pragmas usage

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

10 messages Options

Sophie424

[squeak-dev] pragmas usage

I am looking for some good examples of using pragmas to learn how to use
them well. Found some good ones in Ramon's code, but few others in the 3.9
image. Looking for some more. Any pointers appreciated.

Thanks - Sophie

Lukas Renggli

Re: [squeak-dev] pragmas usage

> I am looking for some good examples of using pragmas to learn how to use
> them well. Found some good ones in Ramon's code, but few others in the 3.9
> image. Looking for some more. Any pointers appreciated.

<primitive:> and <primitive:module:> are (a bit special) examples and
they come with every image. If the Compiler/Parser encounters such
pragmas in the source code it dispatches the pragma to the Parser
class to create a primitive method.

In short the strategy is like this:

- You query for some Pragma using the class-side protocol of the Pragma class.

- You iterate over these objects and dispatch them to some
handler/interpreter object.

aCollectionOfPragma do: [ :each | each message sendTo: handler ]

HTH,
Lukas

--
Lukas Renggli
http://www.lukas-renggli.ch

Sophie424

[squeak-dev] Re: pragmas usage

"Lukas Renggli" <[hidden email]> wrote in message

> - You iterate over these objects and dispatch them to some
> handler/interpreter object.
>
> aCollectionOfPragma do: [ :each | each message sendTo: handler ]

Thanks!

Sophie424

[squeak-dev] Re: pragmas usage

In reply to this post by Lukas Renggli

"Lukas Renggli" <[hidden email]> wrote in message
> - You query for some Pragma using the class-side protocol of the Pragma
> class.

If pragmas were available when you started developing Magritte, I am curious
if and why you might have used them (or not)?

Sophie

Philippe Marschall

Re: [squeak-dev] Re: pragmas usage

2008/3/16, itsme213 <[hidden email]>:
> "Lukas Renggli" <[hidden email]> wrote in message
>
> > - You query for some Pragma using the class-side protocol of the Pragma
> > class.
>
>
> If pragmas were available when you started developing Magritte, I am curious
> if and why you might have used them (or not)?

MAPragmaBuilder

Cheers
Philippe

Lukas Renggli

Re: [squeak-dev] Re: pragmas usage

In reply to this post by Sophie424

> > - You query for some Pragma using the class-side protocol of the Pragma
> > class.
>
> If pragmas were available when you started developing Magritte, I am curious
> if and why you might have used them (or not)?

Certainly I would have used them. This could have avoided those ugly
naming conventions for the descriptions. The problem is that there is
still a lot of industrial code in Squeak 3.7 and 3.8 that depends on
Magritte.

As Philippe pointed out there was MAPragmaBuilder. I never used it in
Pier, due to backward compatibility concerns.

Lukas

--
Lukas Renggli
http://www.lukas-renggli.ch

Ramon Leon-5

RE: [squeak-dev] Re: pragmas usage

> > If pragmas were available when you started developing
> Magritte, I am
> > curious if and why you might have used them (or not)?
>
> Certainly I would have used them. This could have avoided
> those ugly naming conventions for the descriptions. The
> problem is that there is still a lot of industrial code in
> Squeak 3.7 and 3.8 that depends on Magritte.
>
> As Philippe pointed out there was MAPragmaBuilder. I never
> used it in Pier, due to backward compatibility concerns.
>
> Lukas

You'd have used them for pointing to descriptions to avoid naming
conventions, but I think what Sophie's getting at is what you think of
pragmas as a meta layer instead of descriptions to embed the metadata
directly into accessors.

Sophie, if you haven't already, I'd recommend you read his thesis on
Magritte, it'd probably answer many questions about why Magritte made
certain design choices. Pragmas are static and aren't nearly as flexible as
Magritte descriptors, which are real objects and don't face the limitations
of pragmas.

Ramon Leon
http://onsmalltalk.com

Sophie424

[squeak-dev] Re: Re: pragmas usage

"Ramon Leon" <[hidden email]> wrote in message

> Pragmas are static and aren't nearly as flexible as
> Magritte descriptors, which are real objects and don't face the
> limitations
> of pragmas.

Perhaps it need not be either-or? e.g. The pragma could be evaluated to
record the "static" descriptions on the classes as any kind of real objects,
including Magritte-like Descriptions:

Foo>>x
<setUpMeta: #bar>

PragmaInterpreter>>setUpMeta: aSymbol for: aClass
aClass recordMeta: aSymbol

And there could still be instance-level access from objects, which by
default just uses the "static" descriptions that PragmaInterpreter recorded
on the classes, but has a hook for dynamic (instance-specific) behavior:

Object>>getMeta: aSymbol
(self class getMeta: aSymbol)
overridenBy: (self dynamicGetMeta: aSymbol)

Just a thought.

- Sophie

Ramon Leon-5

RE: [squeak-dev] Re: Re: pragmas usage

> "Ramon Leon" <[hidden email]> wrote in message
>
> > Pragmas are static and aren't nearly as flexible as Magritte
> > descriptors, which are real objects and don't face the
> limitations of
> > pragmas.
>
> Perhaps it need not be either-or? e.g. The pragma could be
> evaluated to record the "static" descriptions on the classes
> as any kind of real objects, including Magritte-like Descriptions:
>
> Foo>>x
> <setUpMeta: #bar>
>
> PragmaInterpreter>>setUpMeta: aSymbol for: aClass
> aClass recordMeta: aSymbol
>
>
> And there could still be instance-level access from objects,
> which by default just uses the "static" descriptions that
> PragmaInterpreter recorded on the classes, but has a hook for
> dynamic (instance-specific) behavior:
>
> Object>>getMeta: aSymbol
> (self class getMeta: aSymbol)
> overridenBy: (self dynamicGetMeta: aSymbol)
>
> Just a thought.
>
> - Sophie

I'm sure you could, but I wouldn't, I prefer my metadata to be static. It's
a cross language approach that allows me to use the same techniques in .Net
(still the majority of my job), which also has pragmas (called attributes).
I also prefer the ad-hoc nature of pragmas, I can just make them up as I
need them without the need to try and fit them into a big description
hierarchy.

Ramon Leon
http://onsmalltalk.com

Trygve

Re: [squeak-dev] Re: Re: pragmas usage

In reply to this post by Sophie424

Pragmas has given me a powerful hook into the Compiler/Parser complex. I may be misusing them, but my experiment worked and gave me the insights I needed into the relationships between static and dynamic variables.

A small and simple extension to the Parser made it generate in-line code for special pragmas. (Thanks to Klaus for helping me with this). An example method:

animateStars
    " Add and remove stars randomly, trying to keep total number of stars = 'numberOfStars' "
    <Role: #Collab> <Role: #StarsCount>
    [maestro currentState == #STARS] whileTrue:
        [Collab refresh.
        StarsCount >= maxNumberOfStars ifTrue: [self deleteStar].
        StarsCount <= maxNumberOfStars ifTrue: [self placeStar]]

The Parser extends role variable names such as StarsCount into

self playerForRole: #StarsCount

which looks up the binding for #StarsCount in a dynamic name space. (The name space object lives on the stack)

This was just a step in a larger experiment; more later.
--Trygve

On 19.03.2008 14:57, itsme213 wrote:

"Ramon Leon" [hidden email] wrote in message

Pragmas are static and aren't nearly as flexible as
Magritte descriptors, which are real objects and don't face the 
limitations
of pragmas.


Perhaps it need not be either-or? e.g. The pragma could be evaluated to 
record the "static" descriptions on the classes as any kind of real objects, 
including Magritte-like Descriptions:

Foo>>x
  <setUpMeta: #bar>

PragmaInterpreter>>setUpMeta: aSymbol for: aClass
  aClass recordMeta: aSymbol


And there could still be instance-level access from objects, which by 
default just uses the "static" descriptions that PragmaInterpreter recorded 
on the classes, but has a hook for dynamic (instance-specific) behavior:

Object>>getMeta: aSymbol
 (self class getMeta: aSymbol)
    overridenBy: (self dynamicGetMeta: aSymbol)

Just a thought.

- Sophie

--
--

Trygve Reenskaug mailto: [hidden email]

Morgedalsvn. 5A http://heim.ifi.uio.no/~trygver

N-0378 Oslo Tel: (+47) 22 49 57 27

Norway