如何将类 JSON（不是 JSON）字符串转换为对象？

Question

We all know we can use JSON.parse() to convert the string '{"a":0,"b":"haha"}' to the object {a: 0, b: 'haha'} .

But can we convert the string '{a: 0, b: "haha"}' to the object {a: 0, b: 'haha'} ?

I'm writing a web crawler and I need to get the data in the page. But the complete data is not in DOM but in one <script> element. So I got the useful content in the <script> and converted that string (like 'window.Gbanners = [{...}, {...}, {...}, ...];' ) to a JSON-like string (like '{banners: [...]}' ). However, I couldn't parse the "JSON-like" string. Does anyone have a solution?

Answer 1

A string like {a: 0, b: "haha"} is not JSON, but just a bunch of JavaScript code.

Best way to get a JSON representation of data inside is to run it through a JS parser (such as Esprima ), traverse the syntax tree and build a json object out of it. This needs some work, but at least you'll have the parsing done correctly, with proper handling of escape sequences.

Here's a starting point:

const esprima = require("esprima");
const code = '({a: 0, b: "haha"})';
const ast = esprima.parse(code);

const properties = ast.body[0].expression.properties;
const output = properties.reduce((result, property) => { 
  result[property.key.name] = property.value.value;
  return result;
}, {});
console.log(output);

This code assumes a lot about what the input code looks like - might be OK for a prototype, but still needs error checking and handling nested objects.

(A more generic approach could involve a recursive function that takes an ObjectExpression and returns an equivalent JSON.)

I also had to wrap your input in parentheses so that it's an expression (not a block statement) according to JS grammar.

Answer 2

Something like this might work:

function evalJsString(str) {
    let a = null;
    try {
        eval('a = ' + str);
    } catch (err) {
        console.error(err);
    }
    if(typeof a === "object")
      return a;
    else
      return null;
}

evalJsString('({a: 0, b: "haha"})');

Answer 3

As eval() has security flaws, it's better not to use it. A possible way would be creating a own parser to convert it to JSON string and then apply JSON.parse() . Something like below

function toJSONString(input) {
   const keyMatcher = '([^",{}\\s]+?)';
   const valMatcher = '(.,*)';
   const matcher = new RegExp(`${keyMatcher}\\s*:\\s*${valMatcher}`, 'g');
   const parser = (match, key, value) => `"${key}":${value}`
   return input.replace(matcher, parser);
}

JSON.parse(toJSONString('{a: 0, b: "haha"}'))