Add WP_XML_Tag_Processor and WP_XML_Processor #6713
Conversation
|
The following accounts have interacted with this PR and/or linked issues. I will continue to update these lists as activity occurs. You can also manually ask me to refresh this list by adding the Core Committers: Use this line as a base for the props when committing in SVN: To understand the WordPress project's expectations around crediting contributors, please review the Contributor Attribution page in the Core Handbook. |
Test using WordPress PlaygroundThe changes in this pull request can previewed and tested using a WordPress Playground instance. WordPress Playground is an experimental project that creates a full WordPress instance entirely within the browser. Some things to be aware of
For more details about these limitations and more, check out the Limitations page in the WordPress Playground documentation. |
| * | ----------------|---------------------------------------------------------------------| | ||
| * | *Prolog* | The parser is parsing the prolog. | | ||
| * | *Element* | The parser is parsing the root element. | | ||
| * | *Misc* | The parser is parsing miscellaneous content. | |
There was a problem hiding this comment.
It might be helpful to start without an XML Tag Processor and go straight to the XML Processor, following the design of the HTML Processor with its step() and step_in_body() following the different contexts of the HTML API spec.
Prolog, for example, sounds an awful lot like the IN PROLOG context, where the rules can be contained and indicate a state transition rather than embedding the rules for staging in the lower-level tag processing.
There was a problem hiding this comment.
Shipping only an XML Processor was quite tedious
out of curiosity, what was tedious about it? I have a hard time understanding the need for an XML tag processor, since the nesting rules are straightforward when compared with HTML, where the complicated semantic rules forced a split
There was a problem hiding this comment.
I found it difficult to:
- Think about two different levels of parsing (tokenization, context) in a single class
- Avoid intertwining them
- Come up with names for all the methods
That being said, there's nothing inherently wrong with the approach and I can see how it could work.
| * We limit this scan to 30 characters, which allows twenty zeros at the front. | ||
| */ | ||
| 30 | ||
| ); |
There was a problem hiding this comment.
in this case I don't think we want a limit on these. for one, XML requires the trailing semicolon, and for two, we don't want to add rules to the spec and accidentally let something slip in, for example, by prepending a long reference in the front.
I can clean this up here, but I think it's important we not mush together all of the characters. something like this…
if ( '#' === $text[ $next_character_reference_at + 1 ] ) {
$is_hex = 'x' === $text[ $next_character_reference_at + 2 ] || 'X' === $text[ … ];
$zeros_start_at = $next_character_reference_at + 3 + ( $is_hex ? 1 : 0 );
$zeros_length = strspn( $text, '0', $zeros_start_at );
$digits_start_at = $zeros_start_at + $zeros_length;
$digit_chars = $is_hex ? '0123456789abcdefABCDEF' : '0123456789';
$digits_length = strspn( $text, $digit_chars, $digits_start_at );
$semicolon_at = $digits_start_at + $digits_length;
// Must be followed by a semicolon.
if ( ';' !== $text[ $semicolon_at ] ) {
return false;
}
// Null bytes cannot be encoded in XML.
if ( 0 === $digits_length ) {
return false;
}
/*
* Must encode a valid Unicode code point.
* (Avoid parsing more than is necessary).
*/
$max_digits = $is_hex ? 6 : 7;
if ( $digits_length > $max_digits ) {
return false;
}
$base = $is_hex ? 16 : 10;
$code_point = intval( substr( $text, $digits_start_at, $digits_length ), $base );
if ( if_allowable_code_point( $code_point ) ) {
$decoded .= WP_HTML_Decoder::code_point_to_utf8_bytes( $code_point );
continue;
}
return false;
}
// Must be a named character reference.
$name_starts_at = $next_character_reference_at + 1;
$standard_entities = array(
'amp;' => '&',
'apos;' => "'",
'gt;' => '>',
'lt;' => '<',
'quot;' => '"',
);
foreach ( $standard_entities as $name => $replacement ) {
if ( substr_compare( $text, $name, $name_starts_at, strlen( $name ) ) ) {
$decoded .= $replacement;
continue;
}
}granted, we want to perform all length checks to avoid crashing, but I think we have to scan the entire thing to avoid mis-parsing and security issues
There was a problem hiding this comment.
Oh, the decoder won't just skip over the 30 bytes when it stumbles upon ģ. If there's no valid reference within those 30 bytes, that's a parse failure and we halt. Still, I think you're right and it wouldn't be that bad. If someone loaded 1GB of data into memory, they must be expecting the parser will scan through it eventually.
|
Have you seen the Extensible Markup Language (XML) Conformance Test Suites? It may be a helpful resource to find a lot of test cases, although after a quick scan many cases seem to violate "No DTD, DOCTYPE, ATTLIST, ENTITY, or conditional sections" |
… the MISC context where whitespace text nodes are always complete and non-whitespace ones are syntax errors
I haven't seen that one, thank you for sharing! With everything that's going on in Playground I may not be able to iron out all the remaining nuances here – would you like to take over at one point once XML parsing becomes a priority? |
| * | ||
| * @var bool | ||
| */ | ||
| protected $is_incomplete_text_node = false; |
There was a problem hiding this comment.
this name confused me. I think something like $inside_document_trailing_whitespace could be clearer, but I could still be confused
|
Remember too that text appears in XML outside of In effect, Probably need to examine the comment parser because XML cannot contain |
|
@dmsnell Text is supported, or at least should be – is there a specific scenario that this PR doesn't handle today? |
|
Oh that replacement is just overly selective, it can target text nodes, too. By the way, there's no easy way of setting text in a tag without any text nodes. I'm noodling on emitting an empty text node, but haven't yet found a good way of doing it. Another idea would be to couple the tag tag token with the text that follows it, but that sounds unintuitive for the consumer of this API. |
You're way too fast for me to keep up on this, but it shouldn't be that hard, because we will know where the start and end of a tag is, or if it's an empty tag, we know the full token. We'll just need to create a function like |
|
@adamziel I pushed some changes to the decoder to avoid mixing the HTML and XML parsing rules. Sadly, while I thought that PHP's In mucking around I became aware of how much more the role of errors is going to have to play in an XML API. I don't have any idea what's best. Character encoding failures I would assume are going to be fairly benign as long as we treat those failures as plaintext instead of actually decoding them, but that's a point to ponder. This is going to get interesting with documents mixing HTML and XML, such as WXR. We're going to need to ensure that the tag and text parsing rules are properly separated. I'm still not sure what that means for us when we find something like a WXR without proper escaping of the content inside. |
Streaming makes it a bit more difficult, e.g. we may not have the closer yet, or we may not have the opener anymore. Perhaps pausing on incomplete input before yielding the tag opened would be useful here. |
Oh dang it! Too bad.
Yes, that struck me too. At minimum, we'll need to communicate on which byte offset the error has occurred. Ideally, we'd show the context of the document, highlight the relevant part, and give a highly informative error message.
I'm not sure I follow. By escaping of the content do you mean, say, missing |
My plan with the HTML API is to allow a "soft limit" on memory use. If we need to we can add an additional "hard limit" where it will fail. Should content come in and we're still inside a token, we just keep ballooning past the soft limit until we run out of memory, hit the hard limit, or parse the token. So in this way I don't see streaming as a problem. The goal is to enable low-latency and low-overhead processing, but if we have to balloon in order to handle more extreme documents we can break the rules as long as it's not too wild. I prototyped this with the 🔔 Notifications on WordPress.com though in a slightly different way. The Notifications Processor has a runtime duration limit and a length limit, counting code points in the text nodes of the processed document. If it hits the runtime duration limit, it stops processing formatting and instead rapidly joins the remaining text nodes as unformatted plaintext. If it hits the length limit it stops processing altogether. I believe that this Processor framework opens up new avenues for constraints and graceful degradation beyond those limits.
This could be an interesting operating mode: when bailing, produce an Elm/Rust-quality error message. The character reference errors make me think we also have some notion of recoverable and non-recoverable errors. The character reference error does not cause syntactic issues, so we can zoom past them if we want, collecting them along the way into a list of errors. Errors for things like invalid tag names, attribute names, etc… are similar. Missing or unexpected tags though I could see as being more severe since they have ambiguous impact on the rest of the document.
Some WXRs I've seen will have something like I believe that I've seen WXRs that are broken by not encoding the HTML either way, and these are the ones that scare me: Because the embedded document isn't encoded there's no boundary to detect it. I think this implies that for WordPress, our XML parser could have reason to be itself a blend of XML and HTML. For example:
This exploration is quite helpful because I think it's starting to highlight how the shape of WordPress' XML parsing needs differ from those of HTML. |
|
💡This will generally not be in the most critical performance hot path. We can probably relax some of the excessive optimizations, like relying on more functions to separate concepts like Update: You already had this idea 😆 |
Brings together a few explorations to stream-rewrite site URLs in a WXR file coming from a remote server. All of that with no curl, DOMDocument, or other PHP dependencies. It's just a few small libraries built with WordPress core in mind: * [AsyncHttp\Client](WordPress/blueprints#52) * [WP_XML_Processor](WordPress/wordpress-develop#6713) * [WP_Block_Markup_Url_Processor](https://github.com/adamziel/site-transfer-protocol) * [WP_HTML_Tag_Processor](https://developer.wordpress.org/reference/classes/wp_html_tag_processor/) Here's what the rewriter looks like: ```php $wxr_url = "https://raw.githubusercontent.com/WordPress/blueprints/normalize-wxr-assets/blueprints/stylish-press-clone/woo-products.wxr"; $xml_processor = new WP_XML_Processor('', [], WP_XML_Processor::IN_PROLOG_CONTEXT); foreach( stream_remote_file( $wxr_url ) as $chunk ) { $xml_processor->stream_append_xml($chunk); foreach ( xml_next_content_node_for_rewriting( $xml_processor ) as $text ) { $string_new_site_url = 'https://mynew.site/'; $parsed_new_site_url = WP_URL::parse( $string_new_site_url ); $current_site_url = 'https://raw.githubusercontent.com/wordpress/blueprints/normalize-wxr-assets/blueprints/stylish-press-clone/wxr-assets/'; $parsed_current_site_url = WP_URL::parse( $current_site_url ); $base_url = 'https://playground.internal'; $url_processor = new WP_Block_Markup_Url_Processor( $text, $base_url ); foreach ( html_next_url( $url_processor, $current_site_url ) as $parsed_matched_url ) { $updated_raw_url = rewrite_url( $url_processor->get_raw_url(), $parsed_matched_url, $parsed_current_site_url, $parsed_new_site_url ); $url_processor->set_raw_url( $updated_raw_url ); } $updated_text = $url_processor->get_updated_html(); if ($updated_text !== $text) { $xml_processor->set_modifiable_text($updated_text); } } echo $xml_processor->get_processed_xml(); } echo $xml_processor->get_unprocessed_xml(); ```
…ools (#1888) Let's officially kickoff [the Data Liberation](https://wordpress.org/data-liberation/) efforts under the Playground umbrella and unlock powerful new use cases for WordPress. ## Rationale ### Why work on Data Liberation? WordPress core _really_ needs reliable data migration tools. There's just no reliable, free, open source solution for: - Content import and export - Site import and export - Site transfer and bulk transfers, e.g. mass WordPress -> WordPress, or Tumblr -> WordPress - Site-to-site synchronization Yes, there's the WXR content export. However, it won't help you backup a photography blog full of media files, plugins, API integrations, and custom tables. There are paid products out there, but nothing in core. At the same time, so many Playground use-cases are **all about moving your data**. Exporting your site as a zip archive, migrating between hosts with the [Data Liberation browser extension](https://github.com/WordPress/try-wordpress/), creating interactive tutorials and showcasing beautiful sites using [the Playground block](https://wordpress.org/plugins/interactive-code-block/), previewing Pull Requests, building new themes, and [editing documentation](#1524) are just the tip of the iceberg. ### Why the existing data migration tools fall short? Moving data around seems easy, but it's a complex problem – consider migrating links. Imagine you're moving a site from [https://my-old-site.com](https://playground-site-1.com) to [https://my-new-site.com/blog/](https://my-site-2.com). If you just moved the posts, all the links would still point to the old domain so you'll need an importer that can adjust all the URLs in your entire database. However, the typical tools like `preg_replace` or `wp search_replace` can only replace some URLs correctly. They won't reliably adjust deeply encoded data, such as this URL inside JSON inside an HTML comment inside a WXR export: The only way to perform a reliable replacement here is to carefully parse each and every data format and replace the relevant parts of the URL at the bottom of it. That requires four parsers: an XML parser, an HTML parser, a JSON parser, a WHATWG URL parser. Most of those tools don't exist in PHP. PHP provides `json_encode()`, which isn't free of issues, and that's it. You can't even rely on DOMDocument to parse XML because of its limited availability and non-streaming nature. ### Why build this in Playground? Playground gives us a lot for free: - **Customer-centric environment.** The need to move data around is so natural in Playground. So many people asked for reliable WXR imports, site exports, synchronization with git, and the ability to share their Playground. Playground allows us to get active users and customer feedback every step of the way. - **Free QA**. Anyone can share a testing link and easily report any problems they found. Playground is the perfect environment to get ample, fast moving feedback. - **Space to mature the API**. Playground doesn’t provide the same backward compatibility guarantees as WordPress core. It's easy to prototype a parser, find a use case where the design breaks down, and start over. - **Control over the runtime.** Playground can lean on PHP extensions to validate our ideas, test them on a simulated slow hardware, and ship them to a tablet to see how they do when the app goes into background and the internet is flaky. Playground enables methodically building spec-compliant software to create the solid foundation WordPress needs. ## The way there ### What needs to be built? There's been a lot of [gathering information, ideas, and tools](https://core.trac.wordpress.org/ticket/60375). This writeup is based on 10 years worth of site transfer problems, WordPress synchronization plugins, chats with developers, analyzing existing codebases, past attempts at data importing, non-WordPress tools, discussions, and more. WordPress needs parsers. Not just any parsers, they must be streaming, re-entrant, fast, standard compliant, and tested using a large body of possible inputs. The data synchronization tools must account for data conflicts, WordPress plugins, invalid inputs, and unexpected power outages. The errors must be non-fatal, retryable, and allow manual resolution by the user. No data loss, ever. The transfer target site should be usable as early as possible and show no broken links or images during the transfer. That's the gist of it. A number of parsers have already been prototyped. There's even [a draft of reliable URL rewriting library](https://github.com/adamziel/site-transfer-protocol). Here's a bunch of early drafts of specific streaming use-cases: - [A URL parser](https://github.com/adamziel/site-transfer-protocol/blob/trunk/src/WP_URL.php) - [A block markup parser](https://github.com/adamziel/site-transfer-protocol/blob/trunk/src/WP_Block_Markup_Processor.php) - [An XML parser](WordPress/wordpress-develop#6713), also explored by @dmsnell and @jonsurrell - [A Zip archive parser](https://github.com/WordPress/blueprints-library/blob/87afea1f9a244062a14aeff3949aae054bf74b70/src/WordPress/Zip/ZipStreamReader.php) - [A multihandle HTTP client](https://github.com/WordPress/blueprints-library/blob/trunk/src/WordPress/AsyncHttp/Client.php) without curl dependency - [A MySQL query parser](WordPress/sqlite-database-integration#157) started by @zieladam and now explored by @JanJakes - [A stream chaining API](adamziel/wxr-normalize#1) to connect all these pieces On top of that, WordPress core now has an HTML parser, and @dmsnell have been exploring a [UTF-8](WordPress/wordpress-develop#6883) decoder that would to enable fast and regex-less URL detection in long data streams. There are still technical challenges to figure out, such as how to pause and resume the data streaming. As this work progresses, you'll start seeing incremental improvements in Playground. One possible roadmap is shipping a reliable content importer, then reliable site zip importer and exporter, then cloning a site, and then extends towards full-featured site transfers and synchronization. ### How soon can it be shipped? Three points: * No dates. * Let's keep building on top of prior work and ship meaningful user flows often. * Let's not ship any stable public APIs until the design is mature. For example, the [Try WordPress extension](https://github.com/WordPress/try-wordpress/) can already give you a Playground site, even if you cannot migrate it to another WordPress site just yet. **Shipping matters. At the same time, taking the time required to build rigorous, reliable software is also important**. An occasional early version of this or that parser may be shipped once its architecture seems alright, but the architecture and the stable API won't be rushed. That would jeopardize the entire project. This project aims for a solid design that will serve WordPress for years. The progress will be communicated in the open, while maintaining feedback loops and using the work to ship new Playground features. ## Plans, goals, details ### Next steps Let's start with building a tool to export and import _a single WordPress post_. Yes! Just one post. The tricky part is that all the URLs will have to be preserved. From there, let's explore the breadth and depth of the problem, e.g.: * Rewriting links * Frontloading media files * Preserving dependent data (post meta, custom tables, etc.) * Exporting/importing a WXR file using the above * Pausing and resuming a WXR export/import * Exporting/importing a full WordPress site as a zip file Ideally, each milestone will result in a small, readily reusable tool. For example "paste WordPress post, paste a new site URL, get your post migrated". There's an ample body of existing work. Let's keep the existing codebases (e.g. WXR, site migration plugins) and discussions open in a browser window during this work. Let's involve the authors of these tools, ask them questions, ask them for reviews. Let's publish the progress and the challenges encountered on the way. ### Design goals - **Fault tolerance** – all the data tools should be able to start, stop, resume, tolerate errors, accept alternative data from the user, e.g. media files, posts etc. - **WordPress-first** – let's build everything in PHP using WordPress naming conventions. - **Compatibility** – Every WordPress version, PHP version (7.2+, CLI), and Playground runtime (web, CLI, browser extension, desktop app, CI etc.) should be supported. - **Dependency-free** – No PHP extensions required. If this means we can't rely on cUrl, then let's build an HTTP client from scratch. Only minimal Composer dependencies allowed, and only when absolutely necessary. - **Simplicity** – no advanced OOP patterns. Our role model is [WP_HTML_Processor](https://developer.wordpress.org/reference/classes/wp_html_processor/) – a **single class** that can parse nearly all HTML. There's no "Node", "Element", "Attribute" classes etc. Let's aim for the same here. - **Extensibility** – Playground should be able to benefit from, say, WASM markdown parser even if core WordPress cannot. - **Reusability** – Each library should be framework-agnostic and usable outside of WordPress. We should be able to use them in WordPress core, WP-CLI, Blueprint steps, Drupal, Symfony bundles, non-WordPress tools like https://github.com/adamziel/playground-content-converters, and even in Next.js via PHP.wasm. ### Prior art Here's a few codebases that needs to be reviewed at minimum, and brought into this project at maximum: - URL rewriter: https://github.com/adamziel/site-transfer-protocol - URL detector : WordPress/wordpress-develop#7450 - WXR rewriter: https://github.com/adamziel/wxr-normalize/ - Stream Chain: adamziel/wxr-normalize#1 - WordPress/wordpress-develop#5466 - WordPress/wordpress-develop#6666 - XML parser: WordPress/wordpress-develop#6713 - Streaming PHP parsers: https://github.com/WordPress/blueprints-library/tree/trunk/src/WordPress - Zip64 support (in JS ZIP parser): #1799 - Local Zip file reader in PHP (seeks to central directory, seeks back as needed): https://github.com/adamziel/wxr-normalize/blob/rewrite-remote-xml/zip-stream-reader-local.php - WordPress/wordpress-develop#6883 - Blocky formats – Markdown <-> Block markup WordPress plugin: https://github.com/dmsnell/blocky-formats - Sandbox Site plugin that exports and imports WordPress to/from a zip file: https://github.com/WordPress/playground-tools/tree/trunk/packages/playground - WordPress + Playground CLI setup to import, convert, and exporting data: https://github.com/adamziel/playground-content-converters - Markdown -> Playground workflow _and WordPress plugins_: https://github.com/adamziel/playground-docs-workflow - _Edit Visually_ browser extension for bringing data in and out of Playground: WordPress/playground-tools#298 - _Try WordPress_ browser extension that imports existing WordPress and non-WordPress sites to Playground: https://github.com/WordPress/try-wordpress/ - Humanmade WXR importer designed by @rmccue: https://github.com/humanmade/WordPress-Importer ### Related resources - [Site transfer protocol](https://core.trac.wordpress.org/ticket/60375) - [Existing data migration plugins](https://core.trac.wordpress.org/ticket/60375#comment:32) - WordPress/data-liberation#74 - #1524 - WordPress/gutenberg#65012 ### The project structure The structure of the `data-liberation` package is an open exploration and will change multiple times. Here's what it aims to achieve. **Structural goals:** - Publish each library as a separate Composer package - Publish each WordPress plugin separately (perhaps a single plugin would be the most useful?) - No duplication of libraries between WordPress plugins - Easy installation in Playground via Blueprints, e.g. no `composer install` required - Compatibility with different Playground runtimes (web, CLI) and versions of WordPress and PHP **Logical parts** - First-party libraries, e.g. streaming parsers - WordPress plugins where those libraries are used, e.g. content importers - Third party libraries installed via Composer, e.g. a URL parser **Ideas:** - Use Composer dependency graph to automatically resolve dependencies between libraries and WordPress plugins - or use WordPress "required plugins" feature to manage dependencies - or use Blueprints to manage dependencies cc @brandonpayton @bgrgicak @mho22 @griffbrad @akirk @psrpinto @ashfame @ryanwelcher @justintadlock @azaozz @annezazu @mtias @schlessera @swissspidy @eliot-akira @sirreal @obenland @rralian @ockham @youknowriad @ellatrix @mcsf @hellofromtonya @jsnajdr @dawidurbanski @palmiak @JanJakes @luisherranz @naruniec @peterwilsoncc @priethor @zzap @michalczaplinski @danluu
A part of #1894. Follows up on #1893. This PR brings in a few more PHP APIs that were initially explored outside of Playground so that they can be incubated in Playground. See the linked descriptions for more details about each API: * XML Processor from WordPress/wordpress-develop#6713 * Stream chain from adamziel/wxr-normalize#1 * A draft of a WXR URL Rewriter class capable of rewriting URLs in WXR files ## Testing instructions * Confirm the PHPUnit tests pass in CI * Confirm the test suite looks reasonabel * That's it for now! It's all new code that's not actually used anywhere in Playground yet. I just want to merge it to keep iterating and improving.
…essor (#1960) Merge `WP_XML_Tag_Processor` and `WP_XML_Processor` into a single `WP_XML_Processor` class. This reduces abstractions, enables keeping more properties as private, and simplifies the code. Related to #1894 and WordPress/wordpress-develop#6713 ## Testing instructions Confirm the CI tests pass.
|
I ported this work to https://github.com/WordPress/wordpress-playground/ and experimentally merged the two XML processors into a single |
Trac Ticket: Core-61365
Important
This PR remains open but is not being updated. The XML parser is actively evolving in the WordPress-Playground repository. If you'd like to contribute to the XML API, please do so in there. Once the API matures, this PR will be updated and reviewed for inclusion in WordPress core.
What
Proposes an XML Tag Processor and XML Processor to complement the HTML Tag processor and the HTML Processor.
The XML API implements a subset of the XML 1.0 specification and supports documents with the following characteristics:
The API and ideas closely follow the HTML API implementation. The parser is streaming in nature, has a minimal memory footprint, and leaves unmodified markup as it was originally found.
It also supports streaming large XML documents, see adamziel#43 for proof of concept.
Design decisions
Ampersand handling in text and attribute values
XML attributes cannot contain the characters
<or&.Enforcing
<is fast and easy. Enforcing&is slow and complex because ampersands are actually allowed when used as the start. This means we'd have to decode all entities as we scan through the document – it doesn't seem to be worth it.Right now, the WP_XML_Tag_Processor will only bale out when attempting to explicitly access an attribute value or text containing an invalid entity reference.
Accepting all byte sequences up to
<as text dataXML spec defines text data as follows:
Currently WP_XML_Tag_Processor does not attempt to reject bytes outside of these ranges. It will treat everything between valid elements as text data. On the upside, we avoid using an expensive
preg_match()call for processing text. On the downside, we won't bale out early when processing a malformed, truncated, or a non-UTF-8 document.I think it's a good trade-off to take, but I'm also happy to discuss.
Entity decoding
This API only decodes the following entities:
&,<,>,",'{orOther entities trigger a parse error. This API could use the full HTML decoder, but it would be wrong and slow here. If lack of support for other entities ever becomes a problem,
html_entity_decode()would be a good replacement for the custom implementation. It's worth noting that inXML1mode it doesn't decode many HTML entities:Follow-up work
WP_XML_Processorclass as well as theWP_HTML_Processorclass is documented today.Out of scope for the foreseeable future
<!DOCTYPE, see https://www.w3.org/TR/xml/#sec-prolog-dtd<!ATTLIST, see https://www.w3.org/TR/xml/#attdecls<!ENTITY, see https://www.w3.org/TR/xml/#sec-entity-decl<!NOTATION, see https://www.w3.org/TR/xml/#sec-entity-declcc @dmsnell @sirreal