change property name transform method for html id

[YusukeSuzuki]

In the current version, html_id is not generated correctly when non-ascii characters such as Japanese are used in property names. I have changed from escaping to encoding in order to generate unique IDs while targeting non-ascii characters.

{
  "名前": {
    "description": "顧客の名前",
    "type": "string"
  },
  "年齢": {
    "description": "顧客の年齢",
    "type": "integer"
  }
}

[dblanchette]

Hi @YusukeSuzuki! Thanks for your contribution.

The reason the ID is escaped is to preserve some sort of of readability when sending direct links to users. For example, sending a link to the root property named "person" would be http://example.com/schema.html#person. I think changing this to something like http://example.com/schema.html#706572736f6e is a downgrade.

However, there has been a probelm since the beginning where several strings that escapes to the same value will cause duplicate ids (example: person and person_ will both become person_)

What I did is to combine the "human readable" id with a hex (limited to 8 chars). This should avoid most collisions since the hex uses the "real" value. Collisions are still possible because I limit the number of chars for the value not to explose in length.

The link would then be http://example.com/schema.html#person-72736f6e

This change will break all existing direct links to elements unfortunately.

Any thoughts, objections?

[dblanchette]

I completely changed my mind. Turns out it is totally acceptable to use kanji and other special characters in HTML IDs with HTML 5 which all modern browser supports.

I changed the PR to use all characters in IDs except spaces and escape characters. I also added a flag to restore the old behavior.

[dblanchette]

Released in version 0.44: https://github.com/coveooss/json-schema-for-humans/releases/tag/v0.44