flowr-analysis / flowr

A program slicer and dataflow analyzer for the R programming language.

flowR is a static dataflow analyzer and program slicer for the R programming language (currently tested for versions 4.x and 3.6.x). It is available as a Visual Studio Code extension and as a docker image.

⭐ Getting Started

To get started with flowR, please check out the Overview. The Setup wiki page explains how you can download and setup flowR on your system. For Visual Studio Code, please see the marketplace entry. With docker 🐳️, the following line should be enough (and drop you directly into the REPL):

docker run -it --rm eagleoutice/flowr

📜 More Information

For more details, see the wiki pages and the deployed code documentation.

🚀 Contributing

We welcome every contribution! Please check out the contributing guidelines for more information.

Contributors

Florian Sihler 💻 🤔 🚧 📆 🔬 ⚠️ 📢	Ell 💻 🚧 ⚠️ 🔌	Lukas Pietzschmann 💻 ⚠️	Benedikt Jutz 💻 ⚠️	Core5563 💻 ⚠️	Ehcsan 💻 ⚠️
Add your contributions

---

flowr is actively developed by Florian Sihler (contact at: [email protected]) under the GPLv3 License.

---

flowr's Issues

Produce rdf from ast

Produce quads with file information attached.

Unify statistics folders

Make it statistics-out across the board and register the rest as subfolders.

Transform the returned XML structure of R into a stable AST

`retrieveXmlFromRCode` and family should allow to reuse a given session

otherwise the performance is horrible and if we need to install a package once we have to do so every time!

Parse Calls

Parse function-calls

Automatically extract TODO statements as issues

Maybe just use a simple action for that? otherwise i may loose track regarding whats to do.

parse `break` and `next`

Switch formatter from eslint

See https://typescript-eslint.io/linting/troubleshooting/formatting/

Refactor parser to remove class nesting which is not necessary (configuration can be made static as we do not change how the input object looks)

Refactor Id and Parent fold so they do not have so many local functions.

Mocha schould work with WebSorm Again

it stopped working after updating the multifile reporter setting.

Revisit the Dataflow algorithm

Refactor based on systematic expression-level extraction.

Parse Vectors, Lists, and Data-Frames

Parse constructs like c(1, 2, 3) or list(1,1,2)

Test constant if-then etc. with slightly nested expressions

The xpath queries seem to pick up on the wrong tokens sometimes.

Extract AST of single R script as XML

Use existing RShellSession and xmlparsedata on the R side.

RShellSession should be independent from implicit buffers

onData is dependent on the buffer separation which causes tests etc. to fail sometimes although the data is there! similarly, AST extraction could be fooled by too many output.

Important in the context of #1

Tests for Ranges

Test functions on SourceRange

Parser: allow to install hooks on parsing X -> so that we can count etc.

Parse Sequences

Constructs like 1:3

Parse switch statements

Extract usage information from R code

Parse named arguments

Rudimentary dataflow folder

Tests that install packages (and need a lot of time) should not be run by default and only with the "full suite")

Improve Diff Output of Dataflow tests

While it helps to get links to the corresponding dataflow graph. Currently we get 2 links
(expected and got) and have to compare them manually.

Support ';' and test dataflow extraction of expression-lists

Resolve Namespaces with Symbols

2 + x::b produces:

<exprlist>
  <expr line1="1" col1="1" line2="1" col2="8" start="10" end="17">
    <expr line1="1" col1="1" line2="1" col2="1" start="10" end="10">
      <NUM_CONST line1="1" col1="1" line2="1" col2="1" start="10" end="10">2</NUM_CONST>
    </expr>
    <OP-PLUS line1="1" col1="3" line2="1" col2="3" start="12" end="12">+</OP-PLUS>
    <expr line1="1" col1="5" line2="1" col2="8" start="14" end="17">
      <SYMBOL_PACKAGE line1="1" col1="5" line2="1" col2="5" start="14" end="14">x</SYMBOL_PACKAGE>
      <NS_GET line1="1" col1="6" line2="1" col2="7" start="15" end="16">::</NS_GET>
      <SYMBOL line1="1" col1="8" line2="1" col2="8" start="17" end="17">b</SYMBOL>
    </expr>
  </expr>
</exprlist>

Allow to disable ci with commit msg and run CI in Pull Requests

Use source location as default id

Simple Slicing Based on Dataflow

Refine Readme

Add development hints, write about the commit-message rules, explain the stats and testing script, ...

Parsing Lambdas

Parse for-loops (and others)

Parse for-loop constructs

Currently missing are:

• #27
• #28
• #86
• #38
• tapply, apply, lapply, etc.

Parsing Unary Operators

like - or +

Integrate and display test coverage

ESLint should respect function parameter indentation rule

Currently the configuration in the .eslintrc.json seems to be ignored:

"rules": {
    "indent": [
        "error", 2, {"FunctionDeclaration": {"parameters": "first"},
            "SwitchCase": 1
        }],
}

Parse Semicolon to split multiple Statements

Parsing while loops

Refactor: Revisit the normalized Ast and deal with comments etc.

Parsing repeat loops

Differentiate between packages and namespaces

Support 'T' and 'F' for booleans

Resolve Functions Calls

Disable logging output during tests

We do not want to have logging output for (successful) tests, we need some way to reconfigure logging globally to achieve this.

Setup typedoc

Constant Naming Scheme in Tests

Clean up main folder and move configurations together

Like the mocha-config.json.

Undo #14 and no longer re-run test suite on PR

we already run it on every push... this is redundant.

See #14

Using `const enum` to optimize some enums away

`xmlparsedata` fails for other encodings

the used gsub call can't deal with foreign encodings...