@gracexwho/model-card-generator/test_lale/getHyperparameter.js

Tool for generating model cards for Jupyter Notebook.

var args = process.argv.slice(2);

var fs = require('fs');
var py = require("modified-python-program-analysis/dist/es5");

var filename = "/../lib/lale/sklearn/logistic_regression.py"

//var file = fs.readFileSync(schemaname.toString());

// @TODO: Also parse the "properties" of relevant hyperparameters!


const testFolder = '/../lib/lale/sklearn/';
var schemas = [];
filenames = fs.readdirSync(__dirname + testFolder);
filenames.forEach(file => {
        file = file.replace("_", "");
        file = file.replace(".py", "");
        schemas.push(file);
    });
//console.log(schemas);

var input = "import sklearn.logisticregression";
input = input.toLowerCase();

for (let s of schemas) {
    if (input.includes(s)) {
        console.log("MATCH!");
    }
}


// So now for each import statement "input" want to check if each KEY of the array is in that input


//var schema = JSON.parse();
var contents = fs.readFileSync(__dirname + filename, "utf8");
var flag = false;
var hyperparams = "";
var properties = "";
var pflag = false;
var hyperflag = false;
var brackets = 0;

for (let line of contents.split("\n")) {
    //console.log(line);
    if (line.includes("_hyperparams_schema =")) {
        hyperflag = true;

    }
    if (hyperflag) {

        brackets += (line.match(/{/g)||[]).length
        brackets -= (line.match(/}/g)||[]).length
        //console.log(line);
        //console.log(brackets);
        //console.log();
    }

    if (line.includes("relevantToOptimizer")) {
        flag = true;
    }
    if (flag) {
        hyperparams += line;
    }
    if (line.includes("],")) {
        flag = false;
    }
    if (line.includes("properties")) {
        pflag = true;
    }
    if (pflag && hyperflag) {
        properties = properties + line + "\n";
    }
    if (hyperflag && brackets == 0) {
        break;
    }
}

//console.log("PROPERTIES", properties);

//@TODO //CHANGE IT SO THAT PROPERTIES RELIES ON OPEN BRACKET COUNT TO END PFLAG

hyperparams = hyperparams.substr(hyperparams.indexOf('[')+1);
hyperparams = hyperparams.split("]")[0];
hyperparams = hyperparams.split(",");
var parameters = [];
for (let s of hyperparams) {
    s = s.replace(/['"]+/g, "");
    s = s.trim();
    if (s) {
        parameters.push(s);
    }
}
console.log(parameters);

var openbrackets=0;
var desc = "";
var propertyflag = false;
var hyperparam_descriptions = "";

function containsAny(str, substrings) {
    for (var i = 0; i != substrings.length; i++) {
        var substring = "'" + substrings[i] + "'";
        if (str.indexOf(substring) != - 1) {
            return substring;
        }
    }
    return null;
}
var substring = null;
var param = "";

for (let line of properties.split("\n")) {
    if (!propertyflag) {
        substring = containsAny(line, parameters);
        if (substring != null) {
            //parameters.some(function(v) {hyperparam_descriptions["'" + v + "'"] = ""; return line.indexOf("'" + v + "'") >= 0; })
            propertyflag = true;
            param = substring;
        }
    }
    if (propertyflag) {
        if (line.includes("{")) {
            openbrackets += 1;
        }
        if (line.includes("}")) {
            openbrackets -=1;
        }
        desc = desc + line + "\n";
        if (openbrackets <= 0) {
            propertyflag = false;
            //console.log("PARAM: ", param)
            //console.log(desc);
            hyperparam_descriptions += desc;
            var desc = "";
        }

    }
}

console.log(hyperparam_descriptions);

//hyperparams = hyperparams.split("]")[0];
//console.log(hyperparams);
//console.log(hyperparams.split(","));

//console.log(contents);

//console.log(schema);
//console.log("Hyperparameters: ", Object.keys(schema['allOf'][0]['properties']));

/**

let tree = py.parse(contents);
let cfg = new py.ControlFlowGraph(tree);
const analyzer = new py.DataflowAnalyzer();
const flows = analyzer.analyze(cfg).dataflows;
var lineToCode = {};
// https://github.com/nodeca/argparse
var count = 1

for (let line of contents.split("\n")) {
    lineToCode[count] = line;
    count += 1;
}

var hyp_first = -1;
var hyp_last = -1;
var hyperparams_schema = "";

for (let flow of flows.items) {
    let fromNode = py.printNode(flow.fromNode).split("\n");
    let toNode = py.printNode(flow.toNode).split("\n");
    console.log(flow.fromNode.type);
    console.log(fromNode[0]);
    if (flow.fromNode.type == "assign") {
        if (fromNode[0].includes("_hyperparams_schema")) {
            hyp_first = flow.fromNode.location.first_line;
            hyp_last = flow.fromNode.location.last_line;
        }

    }
}
 **/

/**
var h = hyp_first;
var hyperparams = "";
var flag = false;

while (h <= hyp_last) {
    hyperparams_schema += lineToCode[h];
    h += 1;
    if (lineToCode[h].includes("relevantToOptimizer")) {
        flag = true;
    }
    if (flag) {
        hyperparams += lineToCode[h];
    }
    if (lineToCode[h].includes("],")) {
        flag = false;
    }
}

//console.log(hyperparams);
hyperparams = hyperparams.split("[")[1];
hyperparams = hyperparams.split("]")[0];
hyperparams = hyperparams.split(",");

//hyperparams = hyperparams.replace(/\s/g, "");
var parameters = [];
for (let s of hyperparams) {
    s = s.replace(/['"]+/g, "");
    s = s.trim();
    if (s) {
        parameters.push(s);
    }
}

console.log(parameters);

**/
/**
var chars = {'_':'','F':'f','T':'t', "'":'"'};
hyperparams_schema = hyperparams_schema.replace(/[_FT']/g, m => chars[m]);

hyperparams_schema= hyperparams_schema.substr(hyperparams_schema.indexOf('{'));
console.log(hyperparams_schema);

hyperparams_schema = JSON.parse(hyperparams_schema);
**/