daanx/oopsla22-namedh repository overview

⁠OOPSLA'22 Paper Artifact: First Class Names for Effect Handlers

Docker image: daanx/oopsla22-namedh:1.0

Digest: sha256:d15cf6b8a105160e550eaa64c639b97289504cf7fe5bf6c54000ce84b730891a

⁠Getting Started

We provide a docker image to run the examples of the paper.

> docker pull daanx/oopsla22-namedh:1.0
> docker run -it daanx/oopsla22-namedh:1.0

We now see the koka interpreter prompt as:

 _         _
| |       | |
| | _ ___ | | _ __ _
| |/ / _ \| |/ / _' |  welcome to the koka interactive compiler
|   ( (_) |   ( (_| |  version 2.4.1, Jul 11 2022, libc x64 (gcc)
|_|\_\___/|_|\_\__,_|  type :? for help, and :q to quit

loading: std/core
loading: std/core/types
loading: std/core/hnd

>

We can then see the provided examples by typing :l samples/named-handlers/ <tab>:

> :l samples/named-handlers/
 1 net/             4 file.kk          7 file-scoped.kk
 2 ask.kk           5 heap.kk
 3 net.kk           6 unify.kk

When loading ask.kk for example we see:

> :l samples/named-handlers/ask.kk
compile: samples/named-handlers/ask.kk
loading: std/core
loading: std/core/types
loading: std/core/hnd
check  : samples/named-handlers/ask
modules:
  samples/named-handlers/ask

>

and we can run each sample by running main():

> main()
check  : interactive
check  : interactive
add default effect for std/core/exn
linking: interactive
created: .koka/v2.4.1/gcc-debug/interactive

hello world

We can see the highlighted source of each sample using the :s command.

We can see the inferred type of an expression by using the :t <expr> command.

Exit the interpreter (and container) by using the :q command.

⁠Step-by-Step Guide

The included examples are:

1. ask.kk: A simple reader effect discussed in Section 2.2 of the paper
2. file.kk: A non-scoped file effect discussed in Section 2.3.1 (and Figure 1)
3. file-scoped.kk: A scoped file effect. This is not directly discussed in the paper but very similar to the scoped reader of Figure 4.
4. heap.kk: A first-class heap discussed in Section 3.2.4 (and Figure 5)
5. unify.kk: A unification effect discussed in Section 7.2
6. net.kk: A neural network discussed in Section 7.1

Each of these show aspects of the named handler design as discussed in the paper, like named handlers, scoped handlers, rank-2 types, etc.

⁠Ask

This is the reader effect discussed in Section 2.2. The output should look like:

> :l samples/named-handlers/ask.kk
...
> main()
...

hello world

⁠File

A non-scoped file effect discussed in Section 2.3.1 and Figure 1. Reads the first lines of artifact1.txt and artifact2.txt. The output should look like:

> :l samples/named-handlers/file.kk
...
> main()
...

Hello 
world!

⁠File-Scoped

A scoped file effect. This is not directly discussed in the paper but very similar to the scoped reader of Figure 4. The output should look like:

> :l samples/named-handlers/file-scoped.kk
...
> main()
...

artifact1.txt: line 1
artifact2.txt: line 1

It is interesting to uncomment the "wrong" functions at the end of the source to see the static compile time errors that result when a file escapes its scope (see the later section on how to modify sources and run them).

⁠Heap

A first-class heap with scoped references discussed in Section 3.2.4. This demonstrates we can model first-class polymorphic heaps with just (named) handlers.

> :l samples/named-handlers/heap.kk
...
> main()
...


42

⁠Unify

The unification example discussed in Section 7.2.

> :l samples/named-handlers/unify.kk
...
> main()
...

unified type: list int -> list int

⁠Net

The neural network example discussed in Section 7.1.

> :l samples/named-handlers/net.kk
...
> main()
...

epochs 1: (1, 1): Var(data: [0.65791339619326994], grad: [1] )
epochs 100: (1, 1): Var(data: [0.51198566799805911], grad: [1] )
...
epochs 4900: (1, 1): Var(data: [0.067684855309453493], grad: [1] )
epochs 5000: (1, 1): Var(data: [0.066819144610864939], grad: [1] )

The sampling is randomized so the actual numbers will be different from run to run. The sample also generates the files ./netplot/plot.py and ./netplot/data.json which can be used to see the "learned" sine function. To see the plot graphically, you should exit the interpreter (with :q), get the docker container name, and use that to copy the generated files out of the container onto the local file system:

> docker ps -a 
CONTAINER ID   IMAGE                        COMMAND   CREATED        STATUS                     PORTS     NAMES
8fa5d65107b7   daanx/oopsla22-namedh:1.0    "koka"    24 hours ago   Exited (0) 24 hours ago              gifted_saha
...

> docker cp gifted_saha:/build/koka/netplot/plot.py   plot.py
> docker cp gifted_saha:/build/koka/netplot/data.json data.json

(here the name gifted_saha will be something else that was generated on your system).

Now one can run Python locally to render a plot:

> python3 plot.py

⁠Changing The Examples

You can start the docker container with a bash prompt as:

> docker run -it daanx/oopsla22-namedh:1.0  bash

and use vim to edit the samples in the samples/named-handlers directory. You can use the koka command to start the interpreter, or directly run the examples as:

> koka -e samples/named-handlers/ask.kk

for example.

⁠Docker

The provide container is just a build of Koka of the artifact-namedh branch. One can just checkout that branch as:

> git clone --recursive https://github.com/koka-lang/koka -b artifact-namedh
> cd koka
> stack build --fast
> stack exec koka

and run all the provided examples as described here.