@codingame/vscode-languagedetection
v1.0.23
Published
An npm package that uses guesslang's ML model to detect source code languages
Downloads
4,301
Keywords
Readme
vscode-languagedetection
An npm package that uses machine learning to detect source code languages. Powered by @yoeo's guesslang model!
Usage
First install it in your project:
npm install --save @vscode/vscode-languagedetection
# or using yarn
yarn add @vscode/vscode-languagedetection
Then instantiate a ModuleOperations and run the run the model on a string of code:
import { ModelOperations } from "@vscode/vscode-languagedetection";
const modulOperations = new ModelOperations();
const result = await modulOperations.runModel(`
function makeThing(): Thing {
let size = 0;
return {
get size(): number {
return size;
},
set size(value: string | number | boolean) {
let num = Number(value);
// Don't allow NaN and stuff.
if (!Number.isFinite(num)) {
size = 0;
return;
}
size = num;
},
};
}
`);
which will give you the following in order of confidence:
[
{ languageId: 'ts', confidence: 0.48307517170906067 },
{ languageId: 'rs', confidence: 0.10045434534549713 },
{ languageId: 'js', confidence: 0.07833506911993027 },
{ languageId: 'c', confidence: 0.045049071311950684 },
{ languageId: 'lua', confidence: 0.044198162853717804 },
{ languageId: 'cpp', confidence: 0.03847603127360344 },
{ languageId: 'cs', confidence: 0.03298814222216606 },
{ languageId: 'mm', confidence: 0.02999635599553585 },
{ languageId: 'html', confidence: 0.01874217577278614 },
{ languageId: 'sql', confidence: 0.01811739057302475 },
{ languageId: 'swift', confidence: 0.01418407540768385 },
{ languageId: 'pl', confidence: 0.014126052148640156 },
{ languageId: 'md', confidence: 0.01112559624016285 },
{ languageId: 'java', confidence: 0.009976979345083237 },
{ languageId: 'ps1', confidence: 0.009242385625839233 },
{ languageId: 'php', confidence: 0.008150739595293999 },
{ languageId: 'go', confidence: 0.0069260732270777225 },
{ languageId: 'tex', confidence: 0.006594990845769644 },
{ languageId: 'scala', confidence: 0.00619362760335207 },
{ languageId: 'py', confidence: 0.004240741487592459 },
{ languageId: 'r', confidence: 0.0033439004328101873 },
{ languageId: 'matlab', confidence: 0.0030552551615983248 },
{ languageId: 'css', confidence: 0.0026798006147146225 },
{ languageId: 'sh', confidence: 0.0023688252549618483 },
{ languageId: 'ipynb', confidence: 0.002114647999405861 },
{ languageId: 'bat', confidence: 0.0018151027616113424 },
{ languageId: 'hs', confidence: 0.001677449094131589 },
{ languageId: 'erl', confidence: 0.0014191442169249058 },
{ languageId: 'coffee', confidence: 0.000696933304425329 },
{ languageId: 'rb', confidence: 0.0006357143283821642 }
]
Advanced options
By default, this library will work in Node.js. It uses the fs
and path
modules provided by Node.js to read in the model.json
file and the weights file, group1-shard1of1.bin
, that are contained in this repo.
You can overwrite that behavior using the first two optional parameters of ModelOperations
:
modelJSONFunc?: () => Promise<any> // This must return a JSON.parse() object
weightsFunc?: () => Promise<ArrayBuffer>
These allow you to overwrite the model loading behavior of this package if you happen to be in a non-traditional environment. For an example of this, check out how VS Code is doing it.
The third parameter is the options bag that has:
minContentSize?: number
- The minimum number of characters in a file to be considered for language detection. Defaults to20
.maxContentSize?: number
- The maximum number of characters that will be used in a file to be considered for language detection. Defaults to100000
.
Local development
To build from source, follow these steps:
- Clone the repository
- Run
npm install
- Run
npm run watch
To run the tests, simply run npm test
.
To build a production package:
- Run
npm run build
- Run
npm pack
To publish this package, run npm publish
.
Contributing
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.
When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.
Trademarks
This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.