Full-site Export / Import Format

[adamziel]

Let's standardize the site export format used by Playground to:

• Provide users with a reliable, useful export that will continue to work in the future
• Explore a format we'll eventually propose for WordPress core
• Unlock integrating Playground with webhosts

With #1398, no platform-level files live in the WordPress site directory – this makes Playground so much more portable.

Technical implementation

I propose the following format:

• The export is a ZIP file
• In the archive root, there’s wp-admin, wp-content, wp-includes, other WordPress files
  • wp-config.php is shipped and contains all the data it did contain in Playground
• No "storage optimization" is done
  • All the WordPress static assets are included in the export, even if Playground was booted from a minified build.
  • All the exported .php files are unminified, even if Playground was booted from a minified build.
  • The default WordPress theme is present in wp-content/themes.
• In wp-content, two non-standard artifacts are shipped:
  • A series of .sql files containing the full database dump (to stream-write, e.g. up to 500KB in each)
  • The SQLite database dump to enable fast loading in Playground (importing .sql can take noticeable time)
• No other non-standard artifacts are present in the ZIP. In particular:
  • The SQLite integration plugin is not shipped
  • Drop-in plugins like db.php are not shipped
  • Playground mu-plugins are not shipped

This would enable:

• Booting Playground directly from the zip file (both in CLI and in the browser)
• Importing both into Playground and with regular PHP
• Importing both into MySQL and SQLite
• Streamed download

Related work

Let's reuse as much of the work already done as possible:

• PHP lib: SQL -> ZIP exporter (needs refactor to stream instead of buffering). Related: Add commands to export and import SQLite databases wp-cli/ideas#188
• PHP lib: files -> ZIP streaming exporter
• Blueprint libraries including native PHP ZIP support
• URL rewriter: https://github.com/adamziel/site-transfer-protocol
• https://github.com/adamziel/playground-docs-workflow/
• Site Transfer Protocol – Trac discussion with many relevant resources

Outside of scope

• Rewriting links and assets URLs. This will need to land soon after the format, but is not strictly required for the format itself. The URLs would be stream-rewritten on export from the stored site URL to https://wordpress.internal URL using the rewriting flow we'll need for markdown importing. The importer would stream-rewrite them before inserting each post into the database.
• Support for non-standard wp-content path
• Support for exporting env variables or server-provided constants
• Partial export ("I want just these three plugins")

[adamziel]

Related discussions:

• WordPress Playground Integration Planning wporg-theme-directory#14
• Improve and explain migration of media from standard WP install to playground #1525