search

postlight / parser-api

🚀 A drop-in replacement for the Postlight Parser API.
https://reader.postlight.com/
Apache License 2.0
282 stars 113 forks source link

<img ....> gets corrupted when using parse-html #20

Open fappelman opened 5 years ago

fappelman commented 5 years ago

Expected Behavior

The parser should not corrupt the \ content.

Current Behavior

The \ tag originally is 

  <img src=\"https://cdn.example-domain.com/example1.jpg"/>

and after parsing

 <img src="https://www.example-domain.com/%22https://cdn.example-domain.com/example1.jpg/%22/">

Steps to Reproduce

  1. Take the following HTML

<html>
<head>
<body>     
Main content
<br/>
<img src="https://cdn.example-domain.com/example1.jpg"/>
More content
<br/>
More Content to Simulate main content.
<img src="https://cdn.example-domain.com/example2.jpg"/>
</body>
</html>
  1. Call the api with the path /parse-html. The API takes a POST with a JSON object containing a URL and HTML. The HMTL is the HTML as provided in step 1 but is first converted to the following format:
<html>\\n<head>\\n<body>\\nMain content\\n<br/>\\n<img src=\"https://cdn.example-domain.com/example1.jpg\"/>\\nMore content\\n<br/>\\nMore Content to Simulate main content.\\n<img src=\"https://cdn.example-domain.com/example2.jpg\"/>\\n</body>\\n</html>\\n

and the URL value that is passed is https://www.example-domain.com

  1. The JSON result content being returned contains the main content including the images. The image values are however corrupted:
<img src="https://www.example-domain.com/%22https://cdn.example-domain.com/example1.jpg/%22/">
<img src="https://www.example-domain.com/%22https://cdn.example-domain.com/example2.jpg/%22/">

Question/Comment

Am I using the API in a correct way? I could not find any documentation so this is a bit of reverse engineering.

The reason for not doing this directly, i.e. using the /parser?url=..... is that I am trying to work around a problem where a TypeError is returned. See. The page gives back a 202 which the parser cannot handle. I am now downloading the content and try to pass the HTML into the API as a workaround instead. Unfortunately it doesn't react as I expected it would.