add customFragmentsInScripts option

[gzzhanghao]

It leads to an issue with following html fragment:

<script>
function a() {
  b()
  <?php if (false): ?>
    c()
  <?php endif; ?>
}
</script>

After minified, it becomes:

<script>function a(){b(),}</script>

The trailing comma leads to a syntax error.

The only method I found is to prevent minifying javascripts when there's a customFragment inside the script tag.

[alexlamsl]

Can this not be implement with minifyJS by the use of function(text, inline){}?

[gzzhanghao]

@alexlamsl I believe not :(

Custom fragments were replaced by the placeholders at that moment, so you can't easily find out whether there's a customFragment inside the script.

Setting sequence: false for minifyJS may help in this issue, but I think it's better to have such an option to prevent other potential issues.

[alexlamsl]

Right you are. 👍

I think rather than introducing this flag, if we can literally make this custom fragment case works by default that would be better. So if this option you've introduced work without any regressions, we might as well have this behaviour by default.

Failing that, I still prefer offloading this to a custom minifyJS processor if possible, if not because we can employ the same tricks for minifyCSS. After all, uglify-js and clean-css do not have the same sort of ignoreCustomFragments functionality as html-minifier does.

Would this work if I were to restore all the substitutions before passing the script string to function(text, inline) {...}?

[alexlamsl]

So coming back to this. One way to think of it is that uglify-js and clean-css should be fully responsible for the content, so we should just restore any custom fragment substitutions and pass the origin string directly at them. This however would break a lot of existing cases where the current logic does work properly, and many of which produce good savings.

So I think the flag you've introduced is probably quite a good solution to this issue. A few things:

• the flag currently only takes care of JavaScript, but we do have similar issues with CSS as well
• one option is to make two flags, one for JS and one for CSS
• can we have the flag the other way round, i.e. set to true to pass custom fragments as-is?
• please exclude dist/htmlminifier.js from the commit