Started off as a test PR: The PR humbly proposes HTML Minifier Next (HMN) as a compatible, maintained, faster, and more secure alternative to HTML Minifier Terser (it’s the official successor to HTML Minifier).

I took care of some small fixes and improvements—quick summary:

• npm install failures (pre-existing conflicts, unrelated to HMN):
  • typescript@4.9.4 was one patch below what cosmiconfig required—bumped to 4.9.5
  • pug-loader@2.4.0 declares a peer dep on pug@^2, but the project uses pug 3—added legacy-peer-deps=true to .npmrc (pug-loader works fine at runtime with pug 3)
  • webpack-recompilation-simulator@3.2.0 only declares support for webpack 4—same .npmrc fix covers this
• Tests:
  • Example fixture files (examples/*/dist/webpack-5/) were stale—regenerated
  • One test expected bare defer (development mode, no minifier)—updated to defer="defer" after changing how the plugin sets the attribute
    • Note: I preferred this to stay consistent with the codebase—normally I’d recommend to minimize attributes and enable collapseBooleanAttributes: true (maybe worth considering!)
  • tsc was failing on ...new Set(...) spread because no target was set in tsconfig (defaulting to ES3)—added "target": "es2017" to match the existing lib setting
• Lint:
  • “Minifier” (as in “HTML Minifier Next”) wasn’t in cspell’s word list—added to .cspell.json
• Behavioral finding:
  • HMN converts bare defer → defer=defer (valid but unquoted) when the minifier receives it, which differed from the old fixtures. To get the consistent defer="defer" output, I changed attributes.defer = true to attributes.defer = "defer" in generatedScriptTags. This is again to keep things as consistent with possible as with the previous setup.
• Version and changelog:
  • I didn’t propose anything as it depends entirely on your preferences and decision. If accepted, although fundamentally compatible, I recommend to add a note to double-check output and update settings if needed (also see the following general note).

General note: As mentioned, HMN is fundamentally compatible, but a couple of major changes were introduced, and then it also has some convenience features (like presets) and some new flags that may be worth checking out. I only scanned the list of issues in this repo, but the fixes made in HMN might even help address some issues here.

Happy to help a bit more, but as I maintain HMN I also don’t want to impose—I rather propose this for you to make a call if this is a useful update for HTML Webpack Plugin or not. Cheers!