Add a new HTMLEntry

[stloyd]

Resolves: #1914

Change Log

---

Added

• Add a new `HTMLEntry`

Fixed

Changed

Removed

Deprecated

Security

[codecov]

Codecov Report

❌ Patch coverage is 92.00000% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.72%. Comparing base (9e7c5f7) to head (d18f6af).
⚠️ Report is 2 commits behind head on 1.x.

Additional details and impacted files

@@            Coverage Diff             @@
##              1.x    #1940      +/-   ##
==========================================
+ Coverage   79.67%   79.72%   +0.05%     
==========================================
  Files         832      833       +1     
  Lines       24748    24771      +23     
==========================================
+ Hits        19718    19749      +31     
+ Misses       5030     5022       -8     

Components	Coverage Δ
etl	88.84% <90.24%> (+0.02%)	⬆️
cli	85.96% <ø> (ø)
lib-array-dot	95.00% <ø> (ø)
lib-azure-sdk	60.39% <ø> (ø)
lib-doctrine-dbal-bulk	95.14% <ø> (ø)
lib-filesystem	80.68% <100.00%> (ø)
lib-types	88.88% <100.00%> (+0.54%)	⬆️
lib-parquet	68.35% <ø> (ø)
lib-parquet-viewer	83.04% <ø> (ø)
lib-snappy	90.18% <ø> (ø)
bridge-filesystem-async-aws	91.00% <ø> (ø)
bridge-filesystem-azure	89.47% <ø> (ø)
bridge-monolog-http	96.91% <ø> (ø)
bridge-openapi-specification	94.50% <ø> (ø)
symfony-http-foundation	73.17% <ø> (ø)
adapter-chartjs	86.36% <ø> (ø)
adapter-csv	89.08% <ø> (ø)
adapter-doctrine	90.97% <ø> (ø)
adapter-elasticsearch	97.17% <ø> (ø)
adapter-google-sheet	91.40% <ø> (ø)
adapter-http	60.56% <ø> (ø)
adapter-json	89.80% <ø> (ø)
adapter-logger	83.33% <ø> (ø)
adapter-meilisearch	97.87% <ø> (ø)
adapter-parquet	78.91% <ø> (ø)
adapter-text	88.09% <ø> (ø)
adapter-xml	83.07% <ø> (ø)

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[norberttech]

there is a bug in https://github.com/flow-php/flow/blob/1.x/phpunit.xml.dist#L75C1-L80C17 (those should have lib- prefix) which prevents types tests from being executed on the CI/CD 🤦‍♂️ I noticed because on my local machine I have overwritten phpunjt.xml.dist and noticed that html type related tests are failing

[norberttech]

(it was automatically closed since #1944 mentioned this as something that should be resolved)

[norberttech]

I checked out locally to that branch (to figure out why infection is complaining) and when I execute:

 nix-shell --arg php-version 8.2 --run "composer test:lib:types"

I'm getting:

1) Flow\Types\Tests\Unit\Value\HTMLDocumentTest::test_create_with_invalid_html_on_old
Failed asserting that two strings are identical.
--- Expected
+++ Actual
@@ @@
-'<p>invalid</p>'
+''

/Users/norbert/Workspace/flow-php/flow/src/lib/types/tests/Flow/Types/Tests/Unit/Value/HTMLDocumentTest.php:48

Which makes sense, that test is expecting that DOMDocument is going to wrap value with <p></p> but you added \LIBXML_HTML_NOIMPLIED that should prevent that and I have no idea why it's passing on GitHub Actions :|

Different version of libxml perhaps? 🤔

[norberttech]

As for CodeCove / Infection - this PR just does not have enough code coverage (I thought it's something else related to tools config).

https://app.codecov.io/gh/flow-php/flow/pull/1940?dropdown=coverage&src=pr&el=h1&utm_medium=referral&utm_source=github&utm_content=checks&utm_campaign=pr+comments&utm_term=flow-php

It adds some lines of code, but around 27% of HTMLEntry is not covered by any test

[norberttech]

Different version of libxml perhaps? 🤔

Yes, here is the result of the investigation around this issue:

Breaking change: libxml2 2.14.0 (March 27, 2025)
The LIBXML_HTML_NOIMPLIED behavior changed fundamentally in libxml2 2.14.0 when the HTML parser was rewritten to use an HTML5-compliant tokenizer.
Key changes:
Before 2.14.0:

HTML parser based on HTML4 spec (underspecified)
LIBXML_HTML_NOIMPLIED worked leniently with plain text and fragments
Plain text like 'body' was tolerated even without proper HTML tags

From 2.14.0 onwards:

HTML tokenizer fully conforms to HTML5 specification
Tree construction still uses custom algorithm (not fully HTML5)
Stricter handling of invalid markup
LIBXML_HTML_NOIMPLIED now requires actual HTML elements - plain text without tags results in empty output

[norberttech]

HTMLEntry needs better coverage

[norberttech]

After taking another look at \Flow\Types\Value\HTMLDocument I think there is a bigger problem with the implementation.

\Flow\Types\Value\Uuid - when you pass string that is not Uuid or uuid string it will simply throw an exception.

    public function test_construct_with_invalid_string_uuid_throws_exception() : void
    {
        $this->expectException(InvalidArgumentException::class);
        new Uuid('invalid-uuid-string');
    }

\Flow\Types\Value\HTMLDocument is making some weird flips

    #[RequiresPhp('>= 8.4')]
    public function test_create_with_proper_html_on_newer() : void
    {
        $html = '<html><body><div><span>bar</span></div></body></html>';
        $document = new HTMLDocument($html);

        self::assertSame($html, (string) $document);
    }

So what needs to happen is following:

HTMLDocument needs to validate the input and check if passed value is either:

\DOM\HTMLDocument - and extract string from it
\DOMDocument - and extract string from it
string - nothing to extract

Once values are extracted it should validate if given string is a valid HTML5 by checking if:

• starts with <!DOCTYPE html ...>
• root element is named html
• root element has exactly one body
• root element has exactly one head

So now the test that is failing because of different versions of libxml should simply throw an exception.