Add URL rewriting for imported SQL data

[adamziel]

Summary

• Adds a db-apply command that reads db.sql, rewrites URLs using wp-php-toolkit structured processors (BlockMarkupUrlProcessor, URLInTextProcessor), and executes statements against a target MySQL database
• Domain discovery runs inline during db-sync, collecting all HTTP/HTTPS domains from decoded base64 values into .import-domains.json
• Uses WP_MySQL_Naive_Query_Stream (vendored from WordPress/sqlite-database-integration#264) for streaming SQL parsing with cursor-based resumability

URL rewriting pipeline

For each base64-decoded string value in INSERT statements:

1. Serialized PHP → detected via ContentClassifier (port of is_serialized()) → skipped (adjusting s:N: length prefixes is out of scope)
2. JSON → decoded, all string values walked recursively, URLs rewritten via URLInTextProcessor, re-encoded
3. Everything else (HTML, block markup, plain text, markdown) → wp_rewrite_urls() which handles HTML attributes, block comment JSON, text nodes, and CSS url() in style attributes

No preg_match or DOMDocument — only the structured data processors from wp-php-toolkit/data-liberation.

New components

File	Purpose
importer/lib/Base64ValueScanner.php	Scans SQL for FROM_BASE64('...') expressions, decodes values
importer/lib/ContentClassifier.php	Classifies values as serialized PHP, JSON, or text/HTML
importer/lib/DomainCollector.php	Collects unique domains from string values via URLInTextProcessor
importer/lib/SqlValueUrlRewriter.php	Rewrites URLs in a single value using the appropriate processor
importer/lib/SqlStatementRewriter.php	Orchestrates per-statement rewriting
importer/lib/mysql-query-stream/	Vendored WP_MySQL_Naive_Query_Stream + WP_MySQL_Lexer

Usage

# 1. Download SQL (domains discovered automatically)
php import.php db-sync https://source.example.com /path/to/import --secret=TOKEN

# 2. Apply with URL rewriting
php import.php db-apply - /path/to/import \
  --target-user=root --target-db=wp_new \
  --url-mapping=https://source.example.com::https://target.example.com

Test plan

• 72 unit tests across 5 test files (Base64ValueScanner, ContentClassifier, DomainCollector, SqlValueUrlRewriter, SqlStatementRewriter)
• E2E round-trip test (import-36-url-rewriting.test.js — scaffolded, needs implementation)
• Manual test with a real WordPress SQL dump

🤖 Generated with Claude Code

[adamziel]

It seems like the domain detection could break if the download stream pauses mid-query, e.g.

INSERT INTO ... VALUES
(FROM_BASE64("STRING1")),
<paused right here>

So we'll need some way of resuming from there – even if that way is oversimplified for now and requires going back to the initial INSERT. Maybe that could happen in a follow-up PR, it's a relatively low stakes error. Worst-case scenario we'll report an incomplete list of detected domains. We'll still be able to replace any listed domains when inserting the data.

It would be nice to auto-detect the source site domain and mark it for rewriting, maybe assets domain as well if it's separate.

Other than that, I'll need to take this for a spin.

[adamziel]

Also – it would be cool to support downloading the media files loaded from external domains, but we don't strictly need it to land this feature.