Reuse results from previous scala snippets

[riedelcastro]

For the more complex examples in Wolfe it would be nice to break up scala snippets and interleave them with text. It would be great if I could decide for a given cell whether it starts a new scope, or reuses the one from above. By default we can start new scopes to get the behaviour we currently have.

Several obvious ways come to mind:

• Simply prepend the previous snippets to the current one when compiling. This is pretty clean, but maybe not the most efficient way.
• Collect all snippets in the same scope, and then create one scala block from it. Inside this block we inject calls that append the sub-snippet results to a list, and moro just renders the results from this list. This means only one compilation per scope.
• Maybe each snippet in the same scope becomes a function with parameters corresponding to the variables its using from the previous snippet in the same scope. Then we could re-run on a per-snippet basis, and compile incrementally.

[riedelcastro]

Most awesome! Can I control whether there is reuse or not?

[sameersingh]

Yes, see aggregate option in reference.conf.

For a running example, see wolfe/docs/examples/03_grids, the server is up on earth.

[riedelcastro]

This is nice. Another idea to cache compilation is to wrap each snippet cell in a "object" and create an own compilation unit for it. If a cell uses a previous cell, it simply imports that object into scope.

[riedelcastro]

Can I change aggregation on a per-cell basis? That is, I want some cells to start a fresh scope?

[sameersingh]

The problem with combined compiling are the following:

• I don't know how to get the return value of an object (as Moro prints out the last value). Currently, I wrap it inside an apply, but then I don't know how to import the inside of the apply in another object. There must be a way, I'll have to think.
• I don't know how to cache the objects. On a per <user, doc, cell> level, but somehow keeping track of when it has been modified since last run? I'm thinking md5sum of the actual code to object, and send all of the previous code anyway, to be compiled if the md5sum of the code is not found in map.

No aggregation is currently at Application level. It's very easy to otherwise set it up on a cell-level, I just don't know what the interface should be for such configurations. A row of config checkboxes over the editor? A collapsible toolbar? Actually, maybe a form that appears if you click a gears icon.

[riedelcastro]

Can't you do

object Cell1 {
   val x = 5
   val y = 10
   def apply() = x + y
}
object Cell2 {
  import Cell1._
  val z = y + x + 5
  def apply() = z
}

[riedelcastro]

As for the interface, if it was only for defining scopes I would just add a single checkbox/toggle "start a new scope". But I assume there may be more of such properties coming, so yeah, a form makes sense.

[sameersingh]

The problem is that I can't assume just the last line is the one returning something, i.e.

Map("key1" -> 10,
    "key2" -> 100)

So given a String snippet of code, I have no direct way of figuring out where to insert def apply =.

[riedelcastro]

Maybe you can use a light-weight scala parser. Not sure if something like the parser from scalariform would work.