Samsung/mimic
Synthesizer for JavaScript API models
Readme.md
# Mimic: Computing Models for Opaque Code
Mimic is a research prototype to automatically compute models for JavaScript functions whose source code is not available or not easily processed by automatic tools. Our research paper at the ACM SIGSOFT Symposium on the Foundations of Software Engineering 2015 conference gives some more details, and is freely available:
- [Mimic: Computing Models for Opaque Code](http://stefanheule.com/publications/fse15-mimic/).
## Installation
This project requires `node`, its package manager `npm`, as well as `grunt` to run. It also requires `python` and `make`. Make sure they are installed first (Ubuntu and alike):
sudo apt-get install nodejs-legacy npm make
sudo npm install -g grunt-cli
Then, install all dependencies via `npm`:
npm install
Finally build the project:
make
You may want to make sure that all tests pass to verify the build by running `make test`.
Note: We have mostly tried running this prototype on Ubuntu. A quick test on OS X seemed to have trouble installing some of the dependencies.
## Usage
To run the model synthesis, the script `mimic` can be used. It's usage is as follows:
> ./mimic --help
usage: run.py [-h] [-t <n>] --function <body> --arguments <arglist>
[<arglist> ...] [--argnames <names>] [--args <args>]
[--metric <m>] [--out OUT] [--nocolor] [--debug]
[--parallel_t0 <t0>] [--parallel_f <f>]
Run Mimic to compute models for opaque code
optional arguments:
-h, --help show this help message and exit
-t <n>, --threads <n>
Number of threads (-1 = 1/2 of cores available)
(default: -1)
--function <body> The function body of the opaque code (as JavaScript
source code) (default: None)
--arguments <arglist> [<arglist> ...]
One (or more) initial inputs to the opaque code as a
comma-separated list (default: None)
--argnames <names> The name of the arguments (default: arg0, arg1, arg2,
arg3, arg4, arg5, arg6)
--args <args> Arguments to be passed to mimic-core (default: )
--metric <m> The metric to use (0 for default, 1 for naive metric)
(default: 0)
--out OUT Location where the resulting function should be
written to (default: result.js)
--nocolor Don't use any color in the output (default: False)
--debug Output debug information, and only do a single run of
mimic-core (default: False)
--parallel_t0 <t0> The timeout to be used in the first phase (default: 3)
--parallel_f <f> The factor with which to increase the timeout (based
on a single thread, and scaled appropriately for more
threads) (default: 1.025)
As an example, the following command synthesizes code for the identify function:
./mimic --function "return x" --argnames "x" --arguments "1" "2"
Mimic works by running multiple instances of `mimic-core` internally, each with a different random seed, and stops as soon as one of them succeeds. To understand the tool a little better, it is possible to run `mimic` with the option `--debug`, which will only launch a single copy of `mimic-core`, and output various information along the way. For example:
./mimic --function "return x" --argnames "x" --arguments "1" "2" --debug
It is also possible to run `mimic-core` directly. Furthermore, we have collected a few examples from the array standard library, and the script `scripts/example.py` can be used to conveniently access them. For instance, the following command synthesizes a JavaScript model for the Array.prototype.pop function:
scripts/example.py pop
Finally, there are two more useful scripts included: `scripts/experiment.py` repeats runs of mimic for all examples and gathers statistics. The script `scripts/process.py` can then be used to analyze that information. Pass `--help` to them to get more information on how to use them.
## Tests
Tests are written using mocha, and can be run by the following command:
make test
License
-------
The model synthesis code is distributed under the [Apache License](http://www.apache.org/licenses/LICENSE-2.0.html).