I think this is an artifact of the type checker. It will run on a single file before any of the lines in the file are run. So, if you're trying to use a class that's only being imported with a Machine.add declaration, that declaration is not going to run before the type checker gets to the line where you use it. But, if a file has two Machine.add declarations, then the type isn't used in that file, so the type checker doesn't complain, then at runtime the first .add is run, followed by the second. I guess a Machine.import would need to compile and run the file during compile time, which might be non-trivial because I'm not sure that the compiler can be gracefully interrupted. Maybe the "import" keyword is the way to go. This might be straightforward to do by adding a few rules to the grammar and making import be a reserved word, and allow a number of import statements (only?) at the top of of a program. I have definitely faced the same issue when I was working on utility classes. ~Jack On Sat, Aug 8, 2020 at 11:46 AM Curtis Ullerich wrote:

Thanks for confirming. I subscribed to the issue in case it gains traction.

I found it curious that Machine.add used in the header of control.ck doesn't work, but it works if the libs and control.ck are Machine.added in the same file. Why is that?

On Sat, Aug 8, 2020, 11:35 Michael Heuer wrote:

...

Hello Curtis,

In LiCK there is one big import.ck file (your second method)

https://github.com/heuermh/lick/blob/master/import.ck

I typically use it with two terminal windows, in one

$ chuck --loop

and in the other

$ chuck + import.ck $ chuck + other-stuff.ck

See also

Add namespaces and import statements https://github.com/ccrma/chuck/issues/109

Cheers,

michael

On Aug 8, 2020, at 12:56 PM, Curtis Ullerich wrote:

What's the state of the art for imports/includes?

If I have files lib0.ck and lib1.ck that declare public classes both used in control.ck, I understand these to be the two options for running them:

chuck lib0.ck lib1.ck control.ck

or, make another file control-main.ck: Machine.add("lib0.ck"); Machine.add("lib1.ck"); Machine.add("control.ck");

and run it as: chuck control-main.ck

I thought it would work to use Machine.add("lib0.ck"); Machine.add(" lib1.ck"); as the first line of control.ck and then just run chuck control.ck, but the included classes are not found.

Are these the two options, or is there another way that can support transitive inclusion (not having to list each util file for every program that uses them)?

Thanks, Curtis _______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

The most straightforward version of an import statement would just add the public class in that file to the VM if it hasn't been added already. Do you think an import statement should also add the ability to splice in functions and other classes too? I get an email when new pull requests are made, but I don't get the chance to look through them very often. ~Jack On Sat, Aug 15, 2020 at 9:31 PM Curtis Ullerich wrote:

Makes sense; thank you for the details. A more restricted sense of importing than normal sporking might be good, like only allowing importing of class and method definitions. Pre-constructor class code makes that a little messier.

By the way, does anyone get notified when new pull requests are made on the ChucK repo, or should I be tagging someone?

On Fri, Aug 14, 2020 at 9:01 AM Jack Atherton wrote:

...

I think this is an artifact of the type checker. It will run on a single file before any of the lines in the file are run. So, if you're trying to use a class that's only being imported with a Machine.add declaration, that declaration is not going to run before the type checker gets to the line where you use it. But, if a file has two Machine.add declarations, then the type isn't used in that file, so the type checker doesn't complain, then at runtime the first .add is run, followed by the second.

I guess a Machine.import would need to compile and run the file during compile time, which might be non-trivial because I'm not sure that the compiler can be gracefully interrupted. Maybe the "import" keyword is the way to go. This might be straightforward to do by adding a few rules to the grammar and making import be a reserved word, and allow a number of import statements (only?) at the top of of a program.

I have definitely faced the same issue when I was working on utility classes.

~Jack

On Sat, Aug 8, 2020 at 11:46 AM Curtis Ullerich wrote:

...

Thanks for confirming. I subscribed to the issue in case it gains traction.

I found it curious that Machine.add used in the header of control.ck doesn't work, but it works if the libs and control.ck are Machine.added in the same file. Why is that?

On Sat, Aug 8, 2020, 11:35 Michael Heuer wrote:

...

Hello Curtis,

In LiCK there is one big import.ck file (your second method)

https://github.com/heuermh/lick/blob/master/import.ck

I typically use it with two terminal windows, in one

$ chuck --loop

and in the other

$ chuck + import.ck $ chuck + other-stuff.ck

See also

Add namespaces and import statements https://github.com/ccrma/chuck/issues/109

Cheers,

michael

On Aug 8, 2020, at 12:56 PM, Curtis Ullerich wrote:

What's the state of the art for imports/includes?

If I have files lib0.ck and lib1.ck that declare public classes both used in control.ck, I understand these to be the two options for running them:

chuck lib0.ck lib1.ck control.ck

or, make another file control-main.ck: Machine.add("lib0.ck"); Machine.add("lib1.ck"); Machine.add("control.ck");

and run it as: chuck control-main.ck

I thought it would work to use Machine.add("lib0.ck"); Machine.add(" lib1.ck"); as the first line of control.ck and then just run chuck control.ck, but the included classes are not found.

Are these the two options, or is there another way that can support transitive inclusion (not having to list each util file for every program that uses them)?

Thanks, Curtis _______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

This is what I was thinking as well. I don't think the dependencies of the import statements should be exposed transitively, but they need to be added to the vm. I think it's reasonable to only have public classes be exposed by an import, and leave bare functions and non-public classes private. Jack, no worries about review speed, I just wanted to make sure I was following the preferred procedure. On Wed, Aug 19, 2020 at 3:41 PM Michael Heuer wrote:

On Aug 19, 2020, at 12:25 PM, Jack Atherton wrote:

The most straightforward version of an import statement would just add the public class in that file to the VM if it hasn't been added already. Do you think an import statement should also add the ability to splice in functions and other classes too?

In LiCK many public classes depend on non-public classes local to the file. A quick count shows 437 public classes of approximately 939 total classes:

$ chuck import.ck ... [chuck](VM): sporking incoming shred: 437 (RubberBand.ck)... LiCK imported at path ~/working/lick/

$ find lick -name "*.ck" | xargs grep class | wc -l 939

Imports should also be recursive, e.g. the public class in the imported file may import other classes.

Finally, as described in the linked issue below, having namespaces via a namespace or package statement would also be quite useful.

Cheers,

michael

I get an email when new pull requests are made, but I don't get the chance to look through them very often.

~Jack

On Sat, Aug 15, 2020 at 9:31 PM Curtis Ullerich wrote:

...

Makes sense; thank you for the details. A more restricted sense of importing than normal sporking might be good, like only allowing importing of class and method definitions. Pre-constructor class code makes that a little messier.

By the way, does anyone get notified when new pull requests are made on the ChucK repo, or should I be tagging someone?

On Fri, Aug 14, 2020 at 9:01 AM Jack Atherton wrote:

...

I think this is an artifact of the type checker. It will run on a single file before any of the lines in the file are run. So, if you're trying to use a class that's only being imported with a Machine.add declaration, that declaration is not going to run before the type checker gets to the line where you use it. But, if a file has two Machine.add declarations, then the type isn't used in that file, so the type checker doesn't complain, then at runtime the first .add is run, followed by the second.

I guess a Machine.import would need to compile and run the file during compile time, which might be non-trivial because I'm not sure that the compiler can be gracefully interrupted. Maybe the "import" keyword is the way to go. This might be straightforward to do by adding a few rules to the grammar and making import be a reserved word, and allow a number of import statements (only?) at the top of of a program.

I have definitely faced the same issue when I was working on utility classes.

~Jack

On Sat, Aug 8, 2020 at 11:46 AM Curtis Ullerich wrote:

...

Thanks for confirming. I subscribed to the issue in case it gains traction.

I found it curious that Machine.add used in the header of control.ck doesn't work, but it works if the libs and control.ck are Machine.added in the same file. Why is that?

On Sat, Aug 8, 2020, 11:35 Michael Heuer wrote:

...

Hello Curtis,

In LiCK there is one big import.ck file (your second method)

https://github.com/heuermh/lick/blob/master/import.ck

I typically use it with two terminal windows, in one

$ chuck --loop

and in the other

$ chuck + import.ck $ chuck + other-stuff.ck

See also

Add namespaces and import statements https://github.com/ccrma/chuck/issues/109

Cheers,

michael

On Aug 8, 2020, at 12:56 PM, Curtis Ullerich wrote:

What's the state of the art for imports/includes?

If I have files lib0.ck and lib1.ck that declare public classes both used in control.ck, I understand these to be the two options for running them:

chuck lib0.ck lib1.ck control.ck

or, make another file control-main.ck: Machine.add("lib0.ck"); Machine.add("lib1.ck"); Machine.add("control.ck");

and run it as: chuck control-main.ck

I thought it would work to use Machine.add("lib0.ck"); Machine.add(" lib1.ck"); as the first line of control.ck and then just run chuck control.ck, but the included classes are not found.

Are these the two options, or is there another way that can support transitive inclusion (not having to list each util file for every program that uses them)?

Thanks, Curtis _______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users

_______________________________________________ chuck-users mailing list chuck-users@lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/chuck-users