Newsgroups: php.internals,php.internals
Path: news.php.net
Xref: news.php.net php.internals:120979 php.internals:120980
Reply-To: Dennis Snell <dennis.snell@automattic.com>
Message-ID: <3B406FC1-DEC9-4CB3-80F7-CB90B2F5AA71@automattic.com>
Content-Type: multipart/alternative;
	boundary="Apple-Mail=_16DD6D99-C3C6-4AF2-8455-90EA3211D453"
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3731.700.6\))
Date: Mon, 4 Sep 2023 12:54:47 -0700
In-Reply-To: <c7a7f77b-8187-23f7-7c02-c92f1b538039@gmail.com>
Cc: PHP Internals <internals@lists.php.net>
To: Niels Dossche <dossche.niels@gmail.com>
References: <c7a7f77b-8187-23f7-7c02-c92f1b538039@gmail.com>
Subject: Re: [RFC] [Discussion] DOM HTML5 parsing and serialization support
From: internals@lists.php.net ("Dennis Snell via internals")

--Apple-Mail=_16DD6D99-C3C6-4AF2-8455-90EA3211D453
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain;
	charset=utf-8

Thanks for the proposal Niels,

I=E2=80=99ve dealt with my own grief working through issues in =
DOMDocument and wanting it to work but finding it inadequate.

> HTML5

This would be a great starting point; I would love it if we took the =
opportunity to fix named character reference decoding, as PHP has (to my =
knowledge) never respected (at least in HTML5) that they decode =
differently inside attributes as they do inside markup, considering =
rules such as the ambiguous ampersand and decode errors.

It=E2=80=99s also been frustrating that DOMDocument parses tags in =
RCDATA sections where they don=E2=80=99t exist, such as in TITLE or =
TEXTAREA elements, escapes certain types of invalid comments so that =
they appear rendered in the saved document, and misses basic semantic =
rules (e.g. creating a BUTTON element as a child of a BUTTON element =
instead of closing out the already-open BUTTON).

I=E2=80=99d like to share some what a few of us have been working on =
inside WordPress, which is to build a conformant streaming HTML5 parser:
 - =
https://developer.wordpress.org/reference/classes/wp_html_tag_processor/
 - https://make.wordpress.org/core/2023/08/19/progress-report-html-api/

It=E2=80=99s just food for thought right now because adding HTML5 =
support to DOMDocument would benefit everyone, but we decided we had =
common need in PHP to work with HTML not in a DOM, but in a streaming =
fashion, one with very little runtime overhead. My long-term plan has =
been to get a good grasp for the interface needs and thoroughly test it =
within the WordPress community and then propose its inclusion into PHP. =
It=E2=80=99s been incredibly handy so far, and on my laptop runs at =
around 20 MB/s, which is not great, but good enough for many needs. My =
naive C port runs on the same laptop at around 80 MB/s and I believe =
that we can likely triple or quadruple that speed again if any of us =
working on it knew how to take advantage of SIMD instrinsics.

It tries to accomplish a few goals:
 - be fast enough
 - interpret HTML as an HTML5-compliant browser will
 - find specific locations within an HTML document and then read or =
modify them
 - pass through any invalid HTML it encounters for the browser to =
resolve/fix unless modifying the part of the document containing those =
invalid constructions

I only bring up this different interface because once we started digging =
deep into DOMDocument we found that the problems with it were far from =
superficial; that there is a host of problems and a mismatched interface =
to our common needs. It has surprised me that PHP, the language of the =
web, has had such trouble handling HTML, the language of the web, and we =
wanted to completely resolve this issue once and for all within =
WordPress so we can clean up decades=E2=80=99 old problems with =
encoding, decoding, security, and sanitization.

Warmly,
Dennis Snell

> On Sep 2, 2023, at 12:41 PM, Niels Dossche <dossche.niels@gmail.com =
<mailto:dossche.niels@gmail.com>> wrote:
>=20
> I'm opening the discussion for my RFC "DOM HTML5 parsing and =
serialization support".
> https://wiki.php.net/rfc/domdocument_html5_parser
>=20
> Kind regards
> Niels

--Apple-Mail=_16DD6D99-C3C6-4AF2-8455-90EA3211D453--