Juliette: World Age

This project contains:

Formalization of world age, a Julia language mechanism for efficient implementation of eval.
Implemented in Redex (Racket-based DSL).
Analysis of eval usage in registered Julia packages (primarily static analysis).
Implemented in Julia.

World Age

World age is a language mechanism that prevents new methods defined in eval to be called from an already running function. For example, consider the following program:

f(x) = 1

# g(x) calls f(x) once, then
# redefines f(x) in eval, and calls f(x) again
function g(x)
  v1 = f(x)
  v2 = (eval(:(f(x) = 0)); f(x))
  v1 * v2
end

# at this point, there are two methods:
# f(x) = 1 and g(x) = ...
g(5)     # 1

# at this point, methods are:
# g(x) = ... and f(x) = 0
g(666)   # 0

Without the world age feature, call g(5) would return 0. This is because the first call to f returns 1, the second would return 0 (because f was redefined in eval), and 1*0 is 0.

However, in Julia, g(5) will actually return 1. Why? Because the redefinition f(x) = 0 from eval is not visible while g(5) is running. We can think of this mechanism in the following way: Julia's run-time takes a snapshot of method definitions before the call g(5), and then uses the snapshot to resolve nested method calls.
But once g(5) is done, the new definition of f becomes visible, so the next call g(666) will return 0.

Repository Organization

README.md this file
Project.toml and Manifest.toml files with Julia dependencies required for the project.
install-julia-deps.jl script installing all necessary Julia dependencies.
src/ folder with the implementation of the calculus, analysis, and related utilities:
- redex/ Redex prototype of the calculus;
- jl-transpiler/ transpiler from a subset of Julia to the surface language of the Redex model;
- test-gen Julia scripts for generating Juliette programs in Redex based on Julia test programs;
analysis/ folder with static and dynamic analysis of eval/invokelatest usage in Julia:
- static-analysis/ text- and parse-based static analysis of source code of registered Julia packages;
- dynamic-analysis/ dynamic analysis of eval/invokelatest calls in tests of some Julia packages.
tests/ folder with runnable tests related to the world age semantics:
litmus/ folder with summaries of litmus tests (files in this folder are not supposed to be executed):
- world-age.jl short Julia programs demonstrating the interaction of eval and world age.
- optimizations.jl Julia programs used to test the optimization algorithm implemented in Redex.

Dependencies

Racket with Redex
Julia with the following packages:
- ArgParse
- JSON
We used Julia 1.4.2, but the version should not make a big difference.

Installing Julia dependencies and running Julia

Since we are using pretty basic Julia packages, we don't expect any problems with their versions. Thus, most likely, Option 1 will work just fine. However, if it does not, try Option 2.

Option 1 (easiest)

Run the following from the main directory (jl-wa):

jl-wa$ julia install-julia-deps.jl

After that, run any Julia code with julia command.

Option 2

To use the same versions of the packages that we did, run the following from the main directory (jl-wa):

jl-wa$ julia -e 'using Pkg; Pkg.activate("."); Pkg.instantiate()'

After that, to run Julia code anywhere in jl-wa, use julia --project=@. command instead of julia.

Notes on running Julia

Small scripts (for IO) are much faster if run with:

$ julia -O 0 --compile=min <script.jl>

Static analysis script supports parallel cloning of Julia packages, so it's better to run it with -p N argument where N specifies the number of processors, for example

jl-wa$ julia -p 4 src/analysis/static-analysis/run-all.jl 40

to download and analyze 40 most starred Julia packages.

prl-julia / juliette-wa