Escape indented code block before render

[seaoak]

What does it do?

Supplement of the PR #4171, for the issue #4087.

Since the PR #4171 changes rendering order,
Nunjucks tags in indented code blocks become targets of rendering by Nunjucks.
This is undesirable.

This situation can be observed in the issue #4087.

This patch escapes Nunjucks tags in indented code blocks before rendering by Nunjucks.

How to test

git clone -b feature/escape_indented_code_block_before_render https://github.com/seaoak/hexo.git
cd hexo
npm install
npm test

Pull request tasks

• Add test cases for the changes.
• Passed the CI test.

[seaoak]

Code highlighting might be unnecessary for indented code blocks.

[SukkaW]

Code highlighting might be unnecessary for indented code blocks.

Yeah, there's no way for indented code blocks to add language mark.
Simply wrapped it inside <pre><code> </code></pre> will be ok then.

[SukkaW]

test test test
  hello world
test test
  hello
  world
test test

- a
  - b
    - c
      - d

What about adding test cases above to see if it can handle the situation.

---

Update

The current approach can not handle the second case above correctly.

[seaoak]

@SukkaW I updated.
Your test cases are imported into test codes. Thank you!

[SukkaW]

Your test cases are imported into test codes. Thank you!

@seaoak I get a better idea.
Since it is hard for a regexp to do marked.js's job (to identify which line is nested list, which line is indented code block). What about not to wrapped code block inside <pre><code></code></pre> (this should be done by markdown renderer), only escape { & } for indented line then.

[stevenjoezhang]

Users may use tags in nested list

- a
  - b
    - c
      - {% post_link 'hello world' %}

[seaoak]

As @stevenjoezhang 's commnet, @SukkaW 's idea can not be used.
But I tried to make many test cases, the regexp will be OK.

[SukkaW]

But I tried to make many test cases, the regexp will be OK.

They might be many other cases that should be covered:

- a
  - b
    * c
      1. a
    * b

> 1. a
>   - b
>     1. c

And if the regexp is over complicated, the performance might be affected.

[seaoak]

The condition to recognize "indented code block" is not so complicated.
It is whether the preceding line is a blank line or not.

Only in blockquote, the condition is a bit more complicated.

I think it is acceptable.

[SukkaW]

Suggested change

	data.content = data.content.replace(rIndentedCodeBlock, ($0, before, content, start, $4, $5, end) => {
	// Return early if there is no 4 consecutive spaces for better performance
	if (!data.content.includes(' ')) return;
	data.content = data.content.replace(rIndentedCodeBlock, ($0, before, content, start, $4, $5, end) => {

[seaoak]

To support single-tab-indented code block, this optimization can not be used.

[seaoak]

I Updated.

[seaoak]

Unfortunately, I awake to the danger that this filter spoils indented codes in code block tag.

For example,

{% codeblock lang:javascript %}
if (a) {
  if (b) {
    if (c) {
      doSomething1();

      doSomething2();
    }
  }
}
{% endcodeblock %}

this filter escapes the line 7-8 with <pre><code>...</code></pre>.
This is undesirable.

I have no idea to resolve this. 😢

[seaoak]

I'll discard this patch.