Hi,

Since macro is reserved perhaps D can introduce AST manipulating macros.

Suminda

On 2013-04-21 09:57, Suminda Dharmasena wrote:
> Hi,
>
> Since macro is reserved perhaps D can introduce AST manipulating macros.
>
> Suminda

The original idea with the "macro" keyword was to introduce AST macros. Although nothing has happened yet. People have been talking about it from time to time but no one has had a real proposal.

I've got this: https://dl.dropboxusercontent.com/u/18386187/ast_macros.html

But I haven't turned it into a DIP (D Improvement Proposal) yet.

-- 
/Jacob Carlborg

On Sunday, 21 April 2013 at 07:57:51 UTC, Suminda Dharmasena wrote:
> Hi,
>
> Since macro is reserved perhaps D can introduce AST manipulating macros.
>
> Suminda

Hi!

As far as I can remember, macros have been deferred to D3 because it has mixins and templates, so one can easily manipulate code in D. However, it's been long time since there was serious discussion about the implementation details. I'm happy to see that the topic emerged again just before DConf - maybe the discussion will continue there.
Several non-lispy languages introduced macros long ago so the developers may learn from others mistakes and they will avoid the pitfalls. These languages include Groovy,
Scala, Nimrod and Rust.

The Rust implementation is much like text substituation when used, so one can easily learn it in an hour. Its syntax is a little distinct from the base language, and there has been a problem with lexing it according to a past issue: https://github.com/mozilla/rust/issues/2755
It's still unfinished, there are some limitations. ( although it's been always hard to debug them ) Have a look at: http://static.rust-lang.org/doc/0.6/tutorial-macros.html
and also see the language manual ( at http://rust-lang.org )

Beyond AST macros, Nimrod ( http://nimrod-code.org ) has text substituation mechanisms called templates - much like C macros or templates in C++ and D. It also has term rewriting macros. In my opinion, Nimrod is a language where a lot of features can produce dirty code ( although some of them are very well done and they are extremely powerful tools ), and it's the developer's choice to use the features for the right purposes ( especially true with term rewriting ).
So obfuscation is just as easy in that language as in C++ or Perls before version 6 despite the fact that its sysntax is derived from Python.

I can't say much about Scala and Groovy, but their syntax for defining macros doesn't use as cryptic symbols as Nimrod. For example I couldn't find out for first time what is nnk the nnk prefix in the nnkInfix symbol.

Zsombor

On 2013-04-21 12:36, Belzurix wrote:

> Hi!
>
> As far as I can remember, macros have been deferred to D3 because it has
> mixins and templates, so one can easily manipulate code in D. However,
> it's been long time since there was serious discussion about the
> implementation details. I'm happy to see that the topic emerged again
> just before DConf - maybe the discussion will continue there.
> Several non-lispy languages introduced macros long ago so the developers
> may learn from others mistakes and they will avoid the pitfalls. These
> languages include Groovy,
> Scala, Nimrod and Rust.

There's also Nemerle which lets you create new syntax for your AST macros.

http://nemerle.org/wiki/index.php?title=Macros

-- 
/Jacob Carlborg

> Beyond AST macros, Nimrod ( http://nimrod-code.org ) has text substituation mechanisms called templates - much like C macros or templates in C++ and D. It also has term rewriting macros.

Nimrod's templates do operate on the AST too, they don't perform text substitution.

> In my opinion, Nimrod is a language where a lot of features can produce dirty code ( although some of them are very well done and they are extremely powerful tools ), and it's the developer's choice to use the features for the right purposes ( especially true with term rewriting ).

Term rewriting macros can be turned off on both local and global levels to ensure they don't affect semantics. You can also exclude a TR macro symbol explicitly in an "import" statement (import module except broken_tr_macro)  should the transformation turn out to be buggy.

> So obfuscation is just as easy in that language as in C++ or Perls before version 6 despite the fact that its sysntax is derived from Python.
>

The idea that a statically typed language with checked exceptions and an effect system (both features which didn't make it into 0.9.0 though) is somehow less maintainable than Python with its dynamism everywhere is absurd.

> I can't say much about Scala and Groovy, but their syntax for defining macros doesn't use as cryptic symbols as Nimrod. For example I couldn't find out for first time what is nnk the nnk prefix in the nnkInfix symbol.

Admittedly the macros API has been rushed out and the nnk prefixes are ugly. However there are addons that improve on the API. The core API will be improved too but it needs some transition path as unfortunately quite some code now depends on it. ;-)

Also Redex Racket / Racket can give some inspiration.

Also perhaps Nimrod can get a D back end?

On Sunday, 21 April 2013 at 10:13:43 UTC, Jacob Carlborg wrote:
> On 2013-04-21 09:57, Suminda Dharmasena wrote:
>> Hi,
>>
>> Since macro is reserved perhaps D can introduce AST manipulating macros.
>>
>> Suminda
>
> The original idea with the "macro" keyword was to introduce AST macros. Although nothing has happened yet. People have been talking about it from time to time but no one has had a real proposal.
>
> I've got this: https://dl.dropboxusercontent.com/u/18386187/ast_macros.html
>
> But I haven't turned it into a DIP (D Improvement Proposal) yet.

This is amazing, its not 100% perfect but it hits all the main things I want out of a macro system. How long ago did you come up with this?

On Sunday, 21 April 2013 at 10:13:43 UTC, Jacob Carlborg wrote:
> On 2013-04-21 09:57, Suminda Dharmasena wrote:
>> Hi,
>>
>> Since macro is reserved perhaps D can introduce AST manipulating macros.
>>
>> Suminda
>
> The original idea with the "macro" keyword was to introduce AST macros. Although nothing has happened yet. People have been talking about it from time to time but no one has had a real proposal.
>
> I've got this: https://dl.dropboxusercontent.com/u/18386187/ast_macros.html
>
> But I haven't turned it into a DIP (D Improvement Proposal) yet.

My favorite part was the statement macros and the attribute macros. I want this so much.

On 2014-01-24 11:20, Tofu Ninja wrote:

> This is amazing, its not 100% perfect but it hits all the main things I
> want out of a macro system. How long ago did you come up with this?

I don't know. Dropbox says the file was created 2013 in February. I've created a DIP out of this:

http://wiki.dlang.org/DIP50

-- 
/Jacob Carlborg

On Friday, 24 January 2014 at 13:22:53 UTC, Jacob Carlborg wrote:
> On 2014-01-24 11:20, Tofu Ninja wrote:
>
>> This is amazing, its not 100% perfect but it hits all the main things I
>> want out of a macro system. How long ago did you come up with this?
>
> I don't know. Dropbox says the file was created 2013 in February. I've created a DIP out of this:
>
> http://wiki.dlang.org/DIP50

Is this avaliable as some sort of preprocessor?
Looks really amazing!