Swift REPL

[j-woz]

REPL model for Swift/T.

[bsubei]

I'm starting to tinker with the Turbine API and calling std input from within a Swift leaf function here.

@jmjwozniak I was wondering, is there a way to get variable ids or probe for data members using the Turbine API? Currently, my REPL loop is sitting in a Swift leaf function, and I don't have references to the variables. Should I write functionality that probes and gets all ADLB data member ids?

[bsubei]

since all the generated ids of ADLB data members seem to be consecutive (1, 2, 3, 4... etc), can't I just call a adlb::retrieve on numbers starting from zero until we run out?

it seems like a terribly hacky way of doing it, but I can't think of anything else (perhaps due to my lack of knowledge of adlb internals)

[timarmstrong]

I'll jump in if that's not a problem.

If there are multiple ADLB servers they are not assigned consecutively like
that, but even with a single server you're going to run into more problems
soon.

Looking at the code, the intent is that you should be able to refer to the
global variable a from within the Tcl function without explicitly passing a
into repl_tcl(), right?

You will have trouble achieving this without modification to the compiler.
The compiler has no idea that you intended to implicitly pass a into
repl_tcl(), so it will feel free to optimise a out. Looking at the
compiled code you checked in, I don't see any references to a, so that's
what happened.

The REPL can presumably keep track of any variables it created itself so
it's mainly a problem of getting predeclared variables into the REPL.

I think the options are:
a) explicit pass in any variables you want the REPL to have access to.
This seems easiest
b) modify the compiler so that all global variables are implicitly passed
into certain functions like repl_tcl(). I can help spec out the changes if
that would be useful, but it involves getting into some of the guts of the
compiler.
c) disable the dead code elimination (and possibly other) optimisations in
REPL mode and add functionality to ADLB to enumerate variables. I'm not
sure if there's enough debug information attached to ADLB variables to
actually determine whether a random data item you pulled out of ADLB is a
Swift global, though...

On 12 July 2015 at 17:06, Basheer Subei notifications@github.com wrote:

since all the generated ids of ADLB data members seem to be consecutive
(1, 2, 3, 4... etc), can't I just call a adlb::retrieve on numbers
starting from zero until we run out?

it seems like a terribly hacky way of doing it, but I can't think of
anything else (perhaps due to my lack of knowledge of adlb internals)

—
Reply to this email directly or view it on GitHub
#62 (comment).

[bsubei]

relevant ADLB functions (these can all be called from within Tcl inside the turbine namespace):

• retrieve (gets a data member given an id)
• exists (checks whether data given an id exists)
• closed (checks whether data given an id is closed [is assigned a value?])
• locate (returns which server data given an id is located)
• servers (returns number of servers)

[bsubei]

oh hey @timarmstrong, thanks for your advice!

I was thinking to completely ignore optimizations for now, perhaps just in the initial stages. But now that you mention it, option a) sounds like a very good idea! If the REPL were incorporated as part of the Turbine engine (or maybe even externally), it could keep track of the all the variables (since they all happen within the REPL, whether if the user interactively types them in or tells it to run a script). I guess I need to sit down and flesh this out more conceptually.

Much appreciated! :)

[bsubei]

I have no idea why I was thinking that the REPL needs to sort of "attach" itself to an existing Swift session (or predeclared variables) and map out all the variables... There is no need for this; the user will run the REPL environment and then they can run Swift code interactively. (still not sure how exactly but I'll think about it more)

[j-woz]

I would try the "STC object file" subproject first- I think that will be a better learning curve.

[bsubei]

@jmjwozniak I'm not sure what you mean by STC object file. Is there a webpage where it's discussed in detail?