Hello,

Let us say I have a parsing library. Now, I want to define parsers in stand-alone modules -- for code structuration and reusability. Then, import them from apps that need them.
Is there another way than defining the parser (== list of patterns) at the module's top-level. I have nothing against this, since the module is dedicated to this anyway -- but dmd refuses.
More generally, I find myself unable to define structured objects at a modules's top level. for instancz:

=== mod.d ==
import std.stdio;

struct S {
    int i;
    void speak() {writeln("i: ",this.i);}
}
=== __trials__.d ===
import mod;

auto s = S();
s.speak();
s.i = 1;
writeln(s.i);

void main () {
}

=== compilation ===
__trials__.d(19): no identifier for declarator s.speak
__trials__.d(20): no identifier for declarator s.i
__trials__.d(21): found '.' when expecting ')'
__trials__.d(21): semicolon expected, not 'i'
__trials__.d(21): no identifier for declarator i
__trials__.d(21): semicolon expected, not ')'
__trials__.d(21): Declaration expected, not ')'

Why does dmd refuse? If I put the code in a function, I can compile, link, and run it. But this does not match my use case: I need to export symbols (patterns) defined there; so, i guess, they must be defined at the top of the module. Is there another solution?

I take the opportunity to ask whether it is possible to define which symbols are to be _ex_ported.


Denis
-- -- -- -- -- -- --
vit esse estrany ☣

spir.wikidot.com

spir:

> More generally, I find myself unable to define structured objects at a modules's top level. for instancz:

I suggest to declare global variables and then initialize them in a "static this() {...}". But generally I suggest you to minimize the number of global variables.

Bye,
bearophile

On 11/22/2010 08:18 PM, spir wrote:
> Hello,
>
> Let us say I have a parsing library. Now, I want to define parsers in stand-alone modules -- for code structuration and reusability. Then, import them from apps that need them.
> Is there another way than defining the parser (== list of patterns) at the module's top-level. I have nothing against this, since the module is dedicated to this anyway -- but dmd refuses.
> More generally, I find myself unable to define structured objects at a modules's top level. for instancz:
>
> === mod.d ==
> import std.stdio;
>
> struct S {
>      int i;
>      void speak() {writeln("i: ",this.i);}
> }
> === __trials__.d ===
> import mod;
>
> auto s = S();
> s.speak();
> s.i = 1;
> writeln(s.i);
>
> void main () {
> }
>
> === compilation ===
> __trials__.d(19): no identifier for declarator s.speak
> __trials__.d(20): no identifier for declarator s.i
> __trials__.d(21): found '.' when expecting ')'
> __trials__.d(21): semicolon expected, not 'i'
> __trials__.d(21): no identifier for declarator i
> __trials__.d(21): semicolon expected, not ')'
> __trials__.d(21): Declaration expected, not ')'
>
> Why does dmd refuse? If I put the code in a function, I can compile, link, and run it. But this does not match my use case: I need to export symbols (patterns) defined there; so, i guess, they must be defined at the top of the module. Is there another solution?
>
> I take the opportunity to ask whether it is possible to define which symbols are to be _ex_ported.
>
>
> Denis
> -- -- -- -- -- -- --
> vit esse estrany ☣
>
> spir.wikidot.com
>

To initialize at runtime:

    S s;
    static this() {
        s.i = 1;
    }

to initialize at compile time:

    S getS() { S s; s.i = 1; return s; }
    S s = getS();

Use public/private to control which symbols are availible. Unfortunately, this doesn't really work at the moment :-)

On Mon, 22 Nov 2010 15:11:29 -0500
bearophile <bearophileHUGS@lycos.com> wrote:

> spir:
> 
> > More generally, I find myself unable to define structured objects at a modules's top level. for instancz:
> 
> I suggest to declare global variables and then initialize them in a "static this() {...}". But generally I suggest you to minimize the number of global variables.
> 
> Bye,
> bearophile

Yes, I would generally agree (I normally have about nothing at module toplevel), but then how can they be exported? A parser module is precisely here to define symbols for export... How do you guys write that?


denis
-- -- -- -- -- -- --
vit esse estrany ☣

spir.wikidot.com

spir Wrote:


> === mod.d ==
> import std.stdio;
> 
> struct S {
>     int i;
>     void speak() {writeln("i: ",this.i);}
> }
> === __trials__.d ===
> import mod;
> 
> auto s = S();
> s.speak();
> s.i = 1;
> writeln(s.i);
> 
> void main () {
> }

Others have answered how to initialize variables, but haven't explained the compilation error. A simple example of what you are doing wrong is:

import std.stdio;

writeln("foo");

void main () {
}

How are you to execute writeln()? You probably wouldn't try this under normal conditions, but is what you are trying to do. You can place it in a static this constructor:

import std.stdio;

static this() { writeln("foo"); }

void main () {
}

Note you can move the calls into main and it compiles fine:

void main () {
s.speak();
s.i = 1;
writeln(s.i);
}

On Mon, 22 Nov 2010 22:17:46 +0100
Pelle Månsson <pelle.mansson@gmail.com> wrote:

> On 11/22/2010 08:18 PM, spir wrote:
> > Hello,
> >
> > Let us say I have a parsing library. Now, I want to define parsers in stand-alone modules -- for code structuration and reusability. Then, import them from apps that need them.
> > Is there another way than defining the parser (== list of patterns) at the module's top-level. I have nothing against this, since the module is dedicated to this anyway -- but dmd refuses.
> > More generally, I find myself unable to define structured objects at a modules's top level. for instancz:
> >
> > === mod.d ==
> > import std.stdio;
> >
> > struct S {
> >      int i;
> >      void speak() {writeln("i: ",this.i);}
> > }
> > === __trials__.d ===
> > import mod;
> >
> > auto s = S();
> > s.speak();
> > s.i = 1;
> > writeln(s.i);
> >
> > void main () {
> > }
> >
> > === compilation ===
> > __trials__.d(19): no identifier for declarator s.speak
> > __trials__.d(20): no identifier for declarator s.i
> > __trials__.d(21): found '.' when expecting ')'
> > __trials__.d(21): semicolon expected, not 'i'
> > __trials__.d(21): no identifier for declarator i
> > __trials__.d(21): semicolon expected, not ')'
> > __trials__.d(21): Declaration expected, not ')'
> >
> > Why does dmd refuse? If I put the code in a function, I can compile, link, and run it. But this does not match my use case: I need to export symbols (patterns) defined there; so, i guess, they must be defined at the top of the module. Is there another solution?
> >
> > I take the opportunity to ask whether it is possible to define which symbols are to be _ex_ported.
> >
> >
> > Denis
> > -- -- -- -- -- -- --
> > vit esse estrany ☣
> >
> > spir.wikidot.com
> >
> 
> To initialize at runtime:
> 
>      S s;
>      static this() {
>          s.i = 1;
>      }
> 
> to initialize at compile time:
> 
>      S getS() { S s; s.i = 1; return s; }
>      S s = getS();
> 
> Use public/private to control which symbols are availible. Unfortunately, this doesn't really work at the moment :-)

Thank you very much!
Unfortunately, a parser is typically a bunch a definitions. I cannot expect the user (first, myself) to write a func for every little setting needed.
I could write setters on classes that require the user to set attributes, but as shown in the example, method calls are also refused.
Where is the issue, actually?


Denis
-- -- -- -- -- -- --
vit esse estrany ☣

spir.wikidot.com

On Mon, 22 Nov 2010 16:20:07 -0500
Jesse Phillips <jessekphillips+D@gmail.com> wrote:

> spir Wrote:
> 
> 
> > === mod.d ==
> > import std.stdio;
> > 
> > struct S {
> >     int i;
> >     void speak() {writeln("i: ",this.i);}
> > }
> > === __trials__.d ===
> > import mod;
> > 
> > auto s = S();
> > s.speak();
> > s.i = 1;
> > writeln(s.i);
> > 
> > void main () {
> > }
> 
> Others have answered how to initialize variables, but haven't explained the compilation error. A simple example of what you are doing wrong is:
> 
> import std.stdio;
> 
> writeln("foo");
> 
> void main () {
> }
> 
> How are you to execute writeln()? You probably wouldn't try this under normal conditions, but is what you are trying to do. You can place it in a static this constructor:
> 
> import std.stdio;
> 
> static this() { writeln("foo"); }
> 
> void main () {
> }
> 
> Note you can move the calls into main and it compiles fine:
> 
> void main () {
> s.speak();
> s.i = 1;
> writeln(s.i);
> }
> 

Right. Seems I may put the whole parser definition into static this(). I'll try it. Will work if symbols defined in there are available for export.


Denis
-- -- -- -- -- -- --
vit esse estrany ☣

spir.wikidot.com