haxe.macro.Compiler.setFieldType

[filt3rek]

Hej !

Happy Christmas everyone !

I'm sorry, I always come with strange things...
Last changes in RecordMacros.hx related to :

var get = {
	args : [],
	params : [],
	ret : t,
	expr: macro return @:privateAccess $i{tname}.manager.__get(this, $v{f.name}, $v{relKey}, $v{lock}),
	//expr : Context.parse("return untyped "+tname+".manager.__get(this,'"+f.name+"','"+relKey+"',"+lock+")",pos),
};

record-macros/src/sys/db/RecordMacros.hx

Line 1380 in aaa5f70

var get = {

Makes Haxe complain when trying to build my code.
I've done a small example, let's say we have that :
Main.hx

class Main{
	static var foo : Foo;
	public static function main(){}
}

Foo.hx

class Foo extends sys.db.Object{
	var id		: sys.db.Types.SId;
	@:relation( barID )
	public var bar	: Bar;
}

Bar.hx

class Bar extends sys.db.Object{
	var id : sys.db.Types.SId;
}

my/Bar.hx

package my;
class Bar extends sys.db.Object{
	var id	: sys.db.Types.SId;
	var b	: Bool;
}

build.hxml

-lib record-macros

#--macro haxe.macro.Compiler.setFieldType( "Foo", "bar", "my.Bar" )

-main Main
-php www

If this line is commented, all runs well, but when it's set, I get this error : record-macros/git/src/sys/db/RecordMacros.hx:1391: characters 46-53 : Unknown identifier : my.Bar

Is there something we could do to make it work again like before please ? (Before, untyped was used but now, with the strict synthax, the compiler complains...)

Thanks for reading,
Cheers,

Michal

[filt3rek]

Hej,

I'm sorry, there is nothing to do with type patching.
In fact, you use $i{ tname } but tname can be dotted so ident does'nt work here.
Doing that it works fine $p{ p } :

var ttype = t, p;
while( true )
	switch(ttype) {
		case TPath(t):
			if( t.params.length == 1 && (t.name == "Null" || t.name == "SNull") ) {
				ttype = switch( t.params[0] ) {
					case TPType(t): t;
					default: throw "assert";
				};
				continue;
			}
			p = t.pack.copy();
			p.push(t.name);
			if( t.sub != null ) p.push(t.sub);
			break;
		default:
			Context.error("Relation type should be a type path", f.pos);
	}
	function e(expr) return { expr : expr, pos : pos };
	var get = {
		args : [],
		params : [],
		ret : t,
		expr: macro return @:privateAccess $p{p}.manager.__get(this, $v{f.name}, $v{relKey}, $v{lock}),
	};
	var set = {
		args : [{ name : "_v", opt : false, type : t, value : null }],
		params : [],
		ret : t,
		expr: macro return @:privateAccess $p{p}.manager.__set(this, $v{f.name}, $v{relKey}, _v),
	};

[grepsuzette]

Got the same exact problem as you.

[filt3rek]

Hej Emugel !
Just replace $i{ tname } with $p{ p } and then it works fine.

[grepsuzette]

Thank you sir!

I had one error with String instead of Array<String> (inside the macro),
I really am not sure why, but your listing has removed a certain line tname = p.join(".");,
for me it will work if I change this line to tname = p;, like below:

                             var ttype = t, tname;
                              while( true )
                                  switch(ttype) {
                                  case TPath(t):
                                      if( t.params.length == 1 && (t.name == "Null" || t.name == "SNull") ) {
                                          ttype = switch( t.params[0] ) {
                                          case TPType(t): t;
                                          default: throw "assert";
                                          };
                                          continue;
                                      }
                                      var p = t.pack.copy();
                                      p.push(t.name);
                                      if( t.sub != null ) p.push(t.sub);
~                                     // tname = p.join(".");
+                                     tname = p; 
                                      break;
                                  default:
                                      Context.error("Relation type should be a type path", f.pos);
                                  }
                              function e(expr) return { expr : expr, pos : pos };
                              var get = {
                                  args : [],
                                  params : [],
                                  ret : t, 
                                 expr: macro return @:privateAccess $p{tname}.manager.__get(this, $v{f.name}, $v{relKey}, $v{lock}),
                              };
                              var set = {
                                  args : [{ name : "_v", opt : false, type : t, value : null }],
                                  params : [],
                                  ret : t, 
~                                 expr: macro return @:privateAccess $p{tname}.manager.__set(this, $v{f.name}, $v{relKey}, _v), 
                              };

Do you see why?

Edit: even though it compiles and seems to work I am not sure at all it is correct, it would be better we inform of a likely regression here.

[filt3rek]

Hmmm I don't know why it throws you this error. I'm on Haxe 4, are you too ?
And It works just fine with tname = p.join("."); but maybe It's a specific case I didn't noticed ?

[grepsuzette]

Right now Haxe 4.2.1.
In your snippet you renamed tname to p, since there is a local var p below is it possible it's because of this?

I mean, according to the doc, $p{}: Array<String> -> Expr Generates a field expression from the given string array.. But you use $p{p} and not $p{tname}, so even if you reintroduce tname = p.join("."); it will not produce the String should be Array<String> error.

[grepsuzette]

I had actually misread your patch,
the original var p = t.pack.copy(); became p = t.pack.copy(); in your code.
So it seems a potential fix, thanks!
Hope the lib can be updated though :)

[jonasmalacofilho]

@filt3rek

Can you please open a PR with your fix and a small test case?

(EDIT: please @ me directly on the PR so I can review/approve it, I'm not longer watching/officially maintaining this repository).

[filt3rek]

Hej Jonas !

Thanks for your reply even if you don't maintain this anymore !
I'm not really comfortable with PR and have not much time to look at that this time.
I'll try to take a look at this as soon I've time.

That's pitty that noone want to maintain this lib anymore, I still use it in my projects and find it really light and useful.
IMHO it needs just a big refactoring to put it up to date with new Haxe features...

Thanks anyway for your interest.

[jonasmalacofilho]

I'll try to take a look at this as soon I've time.

Let me know if I can help you get started.

And maybe in the future you can help maintain it too : )

[filt3rek]

Jonas,

I'll be on holiday soon, I'll contact you by PM and yes if you could help me get started and maybe talk a little about how I can help in maintaining this lib, I will be thanksful.

See you soon ;)