Add basic string interpolation / string templates

[dvdvgt]

Fixes #722 and adds basic support for string templates.

There are a few caveats to keep in mind with the current initial implementation:

• All arguments already have to be strings. It is basically just syntactic sugar so that you can write "${a.show} and ${b.show}" instead of a.show ++ " and " ++ b.show where a and b are some 'showable' values. Internally, however, this is just how it is desugared.
• String templates cannot be used to match string literals in pattern matches
• The generated JS code looks a bit weird:

  const v_r_897 = $effekt.println(((((((((("GET ") + (domain_0)))) + (((("/users/") + (user_0))))))) + (((("/resource/") + (('' + resourceId_0))))))));
  

• If one forgets to call show on arguments which are not already of type string, the error messages are hideous and possibly confusing:

error

-There are multiple overloads, which all fail to check:
-Possible overload: effekt::println of type Bool => Unit
-  Expected String but got Int.
-  Expected Bool but got String.
-
-Possible overload: list::println of type List[Bool] => Unit
-  Expected String but got Int.
-  Expected List[Bool] but got String.
-
-Possible overload: array::println of type Array[Int] => Unit
-  Expected String but got Int.
-  Expected Array[Int] but got String.
-
-Possible overload: effekt::println of type String => Unit
-  Expected String but got Int.
-
-Possible overload: list::println of type List[Int] => Unit
-  Expected String but got Int.
-  Expected List[Int] but got String.
-
-Possible overload: effekt::println of type Byte => Unit
-  Expected String but got Int.
-  Expected Byte but got String.
-
-Possible overload: list::println of type List[String] => Unit
-  Expected String but got Int.
-  Expected List[String] but got String.
-
-Possible overload: list::println of type List[Double] => Unit
-  Expected String but got Int.
-  Expected List[Double] but got String.
-
-Possible overload: array::println of type Array[String] => Unit
-  Expected String but got Int.
-  Expected Array[String] but got String.
-
-Possible overload: effekt::println of type Ordering => Unit
-  Expected String but got Int.
-  Expected Ordering but got String.
-
-Possible overload: array::println of type Array[Bool] => Unit
-  Expected String but got Int.
-  Expected Array[Bool] but got String.
-
-Possible overload: effekt::println of type Unit => Unit
-  Expected String but got Int.
-  Expected Unit but got String.
-
-Possible overload: option::println of type Option[Int] => Unit
-  Expected String but got Int.
-  Expected Option[Int] but got String.
-
-Possible overload: option::println of type Option[Double] => Unit
-  Expected String but got Int.
-  Expected Option[Double] but got String.
-
-Possible overload: array::println of type Array[Double] => Unit
-  Expected String but got Int.
-  Expected Array[Double] but got String.
-
-Possible overload: effekt::println of type Double => Unit
-  Expected String but got Int.
-  Expected Double but got String.
-
-Possible overload: effekt::println of type Int => Unit
-  Expected String but got Int.
-  Expected Int but got String.
-
-Possible overload: option::println of type Option[Bool] => Unit
-  Expected String but got Int.
-  Expected Option[Bool] but got String.
-
-  println("GET ${domain}/users/${user}/resource/${resourceId}")

Apart from that, the changes are really lightweight because of the early desugaring in the parser.

I am looking forward to your feedback and ideas.

[jiribenes]

This is a great first implementation, but I'm not sure that desugaring in parser is the best choice here :/
We talked about it at a meeting when #722 was created, but I don't remember what we agreed on.

Personally, I'd much rather see templates up until the transformation to core so that we can:

• have reasonable error messages in typer
• later on resolve typeclasses or something for the splices
• possibly have a more performant impl (?)

[dvdvgt]

Personally, I'd much rather see templates up until the transformation to core so that we can:

• have reasonable error messages in typer
• later on resolve typeclasses or something for the splices
• possibly have a more performant impl (?)

You raise some very good concerns and I completely agree. I will try to go the 'non-lazy route' and I actually completely thread through templates to core.

[jiribenes]

Don't get me wrong, having a version is still super helpful; I truly have no idea how difficult it really is to do the thing I outlined above. But I hope it's worth it, especially wrt technical debt. :)

[b-studios]

I am not completely against desugaring in parser, but maybe we could actually desugar

"foo ${exp1} bar ${exp2} baz"

to something like:

splice(["foo ", " bar ", "baz"], [exp1, exp2])

this way, as a user I can still overload splice for types other than exp1 exp2: String :) Not that don't have enough overloaded functions already...

[dvdvgt]

I just pushed a prototype that threads through string templates all the way until code generation.

This has the added benefit that we remain somewhat flexible for later changes, as Jiří mentioned. Furthermore, we can re-use the string interpolation capabilities of each respective backend, e.g. currently the JS string templates are generated as template literals. As a trade-off, we need to add a TemplateStr AST node to Core, CPS and to each backend. As I have already heard, not everybody is a proponent of further expanding Core ;)

Note: I currently only fully implemented it for JS. If we come to an agreement that this is the right way to proceed, I will implement for the remaining backends.

(I already started working on this before Jonathan's remark and I'd rather discuss this implementation briefly before scrapping it right away :) )

[jiribenes]

I am not completely against desugaring in parser, but maybe we could actually desugar

"foo ${exp1} bar ${exp2} baz"

to something like:

splice(["foo ", " bar ", "baz"], [exp1, exp2])

this way, as a user I can still overload splice for types other than exp1 exp2: String :) Not that don't have enough overloaded functions already...

This relates to some strategy we'll need to come up with in order to represent custom string interpolators (which is very outside of the scope of this PR, I believe...)

Then I'd expect some desugaring like, for example:

val name = "Joe"
val age = 42

////

val s: SQL::Stmt = 
  sql"SELECT * FROM Person p WHERE p.last_name = ${name} AND p.age = ${age}"

// ~>

val s: SQL::Stmt = sql { // sql: { prog: () => Unit / { Splice[String], Splice[Int] } }: SQL::Stmt
  do literal("SELECT * FROM Person p WHERE p.last_name = ")
  do splice(name)
  do literal(" AND p.age = ")
  do splice(age)
}

[dvdvgt]

There are three things keeping this from merging:

• The test fails on LLVM with error code 139 (segmentation fault)
• For some reason the test now also fails on JS even though the output is apparently the same as the expected. Maybe some weird non-displayable characters are inserted by accident?
• I would like to include the splice module in the prelude however, I am prompted with Effekt Compiler Crash: null

[jiribenes]

I would like to include the splice module in the prelude however, I am prompted with Effekt Compiler Crash: null

I'll merge #780 ASAP, then it should at least tell you what's wrong :)
(You can also review it and merge it ;D)

[jiribenes]

I think this could be made somewhat more efficient -- resize copies into a newly allocated bytearray, but toString doesn't (since on LLVM strings are bytearrays)
So it should be possible to just clear the buffer without allocating it again?

Suggested change

	// resize buffer to strip trailing zeros that otherwise would be converted into 0x00 characters
	buffer = bytearray::resize(buffer, pos)
	val str = buffer.toString()
	// after flushing, the stringbuffer should be empty again
	buffer = bytearray::allocate(initialCapacity)
	// resize buffer to strip trailing zeros that otherwise would be converted into 0x00 characters
	val str = bytearray::resize(buffer, pos).toString()
	// after flushing, the stringbuffer should be empty again
	buffer = buffer.fill(0)

where buffer.fill(0) means memset(buffer, 0) (which we don't have yet) :)

[dvdvgt]

Yes, you could just clear the buffer, however, then the buffer's size will not be reset to a (potentially) smaller size and instead just buffer's previous size is kept. This probably would not be a big deal, though.

[b-studios]

Regarding the prelude, we have an undocumented invariant: all imports of prelude modules have to be in the prelude as well...

That is, the prelude in the JS runner needs to look like this:

override def prelude: List[String] = List("effekt", "option", "list", "result", "exception", "array", "string", "ref", "bytearray", "stringbuffer", "splice")

and bytearray, and stringbuffer also need to import effekt and dependencies that they need from the prelude manually.

That is, the following changes would be necessary (+ LLVM Runner)

diff --git a/effekt/jvm/src/main/scala/effekt/Runner.scala b/effekt/jvm/src/main/scala/effekt/Runner.scala
index ab26700e..6dbaac0b 100644
--- a/effekt/jvm/src/main/scala/effekt/Runner.scala
+++ b/effekt/jvm/src/main/scala/effekt/Runner.scala
@@ -151,7 +151,7 @@ object JSNodeRunner extends Runner[String] {
 
   override def includes(path: File): List[File] = List(path / ".." / "js")
 
-  override def prelude: List[String] = List("effekt", "option", "list", "result", "exception", "array", "string", "ref")
+  override def prelude: List[String] = List("effekt", "option", "list", "result", "exception", "array", "string", "ref", "bytearray", "stringbuffer", "splice")
 
   def checkSetup(): Either[String, Unit] =
     if canRunExecutable("node", "--version") then Right(())
diff --git a/examples/pos/string_interpolation.effekt b/examples/pos/string_interpolation.effekt
index 60aeb0b0..54bf9a20 100644
--- a/examples/pos/string_interpolation.effekt
+++ b/examples/pos/string_interpolation.effekt
@@ -1,6 +1,3 @@
-import splice
-import stringbuffer
-
 type Expr {
   Var(id: String)
   Abs(param: String, body: Expr)
@@ -38,4 +35,3 @@ def main() = {
   val fixpoint = Abs("f", App(Abs("x", App(Var("f"), App(Var("x"), Var("x")))), Abs("x", App(Var("f"), App(Var("x"), Var("x"))))))
   println(pretty"Fix point combinator: ${fixpoint}")
 }
-
diff --git a/libraries/common/bytearray.effekt b/libraries/common/bytearray.effekt
index d7552919..d7232e44 100644
--- a/libraries/common/bytearray.effekt
+++ b/libraries/common/bytearray.effekt
@@ -1,5 +1,7 @@
 module bytearray
 
+import effekt
+
 /**
  * A memory managed, mutable, fixed-length array of bytes.
  */
diff --git a/libraries/common/splice.effekt b/libraries/common/splice.effekt
index 9ba74d15..7de68c0d 100644
--- a/libraries/common/splice.effekt
+++ b/libraries/common/splice.effekt
@@ -1,5 +1,7 @@
 module splice
 
+import effekt
+
 import stringbuffer
 
 effect literal(s: String): Unit
diff --git a/libraries/common/stringbuffer.effekt b/libraries/common/stringbuffer.effekt
index 3b6d3acb..bfa57e62 100644
--- a/libraries/common/stringbuffer.effekt
+++ b/libraries/common/stringbuffer.effekt
@@ -1,5 +1,6 @@
 module stringbuffer
 
+import effekt
 import bytearray
 
 interface StringBuffer {