Floki mistakes &adxnnl=1; in attributes for character code and crashes

[umuro]

Description

attribute contents crash Floki when they contain & and later ;. For example "&adxnnl=1;" .The tag example below is from https//:nytimes.com

To Reproduce

Steps to reproduce the behavior:

• Using Floki v0.23.0
• Using Elixir v1.9.1
• Using Erlang OTP v22.1.1
• With this code:

  "<meta data-rh=\"true\" name=\"msapplication-task\" content=\"name=Homepage;action-uri=https://www.nytimes.com?src=iepin&adxnnl=1;icon-uri=https://static01.nyt.com/images/icons/homepage.ico\"/>" |> Floki.parse

Expected behavior

Floki should not mistake it for a special character whenever there is an & around... In some sites, query strings are used as attribute contents. Query strings mean a lot of &'s around.

Workaround

Before Floki.parse get rid of the annoying pattern

 ~r/&(?=[[:alnum:]]+=.+;)/ |> Regex.replace(string, "\+\*\+\*")

It's easy to revert this also. But a lot of precious CPU time is lost

[philss]

@umuro Thank you for opening the issue! It was fixed in version 0.23.1. Can you try again with that version?