Improved parsing of Trans component

package.json

@@ -17,6 +17,7 @@
     "url": "https://github.com/i18next/i18next-parser"
   },
   "dependencies": {
+    "acorn-jsx": "^4.1.1",
     "colors": "~1.2.0-rc0",
     "commander": "~2.9.0",
     "concat-stream": "~1.6.0",

src/helpers.js

@@ -105,9 +105,16 @@ function populateHash(source, target = {}) {
   return target
 }
 
+class ParsingError extends Error {
+  constructor(message) {
+    super(message);
+    this.name = "ParsingError";
+  }
+}
 
 export {
   dotPathToHash,
   mergeHashes,
-  populateHash
+  populateHash,
+  ParsingError
 }

src/lexers/jsx-lexer.js

@@ -1,4 +1,8 @@
+import * as acorn from 'acorn-jsx';
+import assert from 'assert'
 import HTMLLexer from './html-lexer'
+import BaseLexer from './base-lexer';
+import { ParsingError } from '../helpers';
 
 export default class JsxLexer extends HTMLLexer {
   constructor(options = {}) {
@@ -44,7 +48,7 @@ export default class JsxLexer extends HTMLLexer {
       const key = attrs.keys
 
       if (matches[3] && !attrs.options.defaultValue) {
-        attrs.options.defaultValue = matches[3].trim()
+        attrs.options.defaultValue = this.eraseTags(matches[0]).replace(/\s+/g, ' ')
       }
 
       if (key) {
@@ -54,4 +58,59 @@ export default class JsxLexer extends HTMLLexer {
 
     return this.keys
   }
+
+  /**
+   * Recursively convert html tags and js injections to tags with the child index in it
+   *
+   * @param {string} string
+   *
+   * @returns string
+   */
+  eraseTags(string) {
+    const children = this.simplify(acorn.parse(string, {plugins: {jsx: true}}).body[0].expression.children, string);
+
+    const elemsToString = children => children.map((child, index) => {
+      switch(child.type) {
+        case 'text': return child.content;
+        case 'js': return `<${index}>${child.content}</${index}>`;
+        case 'tag': return `<${index}>${elemsToString(child.children)}</${index}>`;
+        default: throw new ParsingError('Unknown parsed content: ' + child.type);
+      }
+    }).join('');
+
+    return elemsToString(children);
+  }
+
+  /**
+   * Simplify the bulky AST given by Acorn
+   * @param {*} children An array of elements contained inside an html tag
+   * @param {string} originalString The original string being parsed
+   */
+  simplify(children, originalString) {
+    for (let i = 0; i < children.length; i ++) {
+      const child = children[i];
+      switch (child.type) {
+        case 'JSXText': children[i] = {
+          type: 'text',
+          content: child.value.replace(/^(?:\s*(\n|\r)\s*)?(.*)(?:\s*(\n|\r)\s*)?$/, '$2')
+        }; break
+        case 'JSXElement': children[i] = {
+          type: 'tag',
+          children: this.simplify(child.children, originalString)
+        }; break;
+        case 'JSXExpressionContainer': children[i] = {
+          type: 'js',
+          content: originalString.slice(child.start, child.end)
+        }; break;
+        default: throw new ParsingError("Unknown ast element when parsing jsx: " + child.type)
+      }
+    }
+
+    // Filter empty text elements. Using string.length instead of !string because
+    // '0' is a valid text element, and '' is not, and !string doesn't make a difference
+    // between the two.
+    children = children.filter(child => !(child.type === 'text' && child.content.length === 0));
+
+    return children;
+  }
 }