Distan -- WordPress 插件

开发者	youheiokubo
更新时间	2026年10月10日 23:29
PHP版本:	8.0 及以上
WordPress版本:	7.1
版权:	GPLv2 or later
版权网址:	版权信息

Distan turns WordPress into a build environment for static HTML deliverables. You author in WordPress, then export clean, self-contained HTML that runs anywhere. No WordPress, no PHP, and no database are required on the production server. Distan is designed for projects where the agency maintains the source and hands over HTML files: recruitment sites, campaign pages, or a section of an existing corporate site where another WordPress installation is not an option. Edit in WordPress, regenerate, review the changes, and deliver the files through your existing publishing process. Built around the handoff

Keep updates reviewable. See added, changed and removed files, download a differential ZIP, and pass on a deletion list. Distan does not delete files on the production server.
Inspect what you actually deliver. Check generated links and development URLs, and review page-level JSON-LD diagnostics. Reports use the generation record, so changing settings later does not change the recorded findings for that build.
Respect existing structured data. Optional basic WebPage JSON-LD fills pages without existing JSON-LD. Theme, SEO-plugin and handwritten JSON-LD take precedence; diagnostics flag issues without repairing the data automatically.
Make the next person's job easier. Export one page with its referenced assets and editing instructions, or export selected content as Markdown for an AI-assisted workflow.
Keep publishing under your control. Download files or connect your deployment process through hooks. Distan does not store deployment credentials or require a Node.js build toolchain.

What it does

Generates every published page, the posts archive with pagination, category archives with pagination, and a 404 page.
Rewrites internal links as document-relative paths, so the output can be opened directly and placed in any directory.
Flattens WordPress paths: theme files move to assets/, uploads move to media/. The wp-content directory disappears from the deliverable.
Cleans the HTML: removes the generator tag, REST API links, oEmbed, emoji scripts, speculative loading rules, and the development noindex.
Keeps cache-busting query strings on assets, so file names stay stable and can be overwritten over FTP.
Optional sitemap.xml and a minimal robots.txt, built from the pages actually generated, using the production URL.
Optional Markdown export (content.md) that combines every page's main content into one file — handy for feeding a site into AI tools (Gemini Notebook / NotebookLM, and similar).
Exports a single page as a starting template for an outside coder: that page plus only the assets it references (following stylesheet url() and @import), laid out at their real paths, with a README of ground rules. In-page markers can drop the block-library CSS (distan:no-block-styles) or drop scripts and styles under a path prefix (distan:drop-assets).
Surfaces URLs that WordPress core's sitemap lists but enumeration missed — a plugin's thank-you page, a virtual route — and lets you include or ignore each. The choice is remembered across runs, so there is no filter to hand-write.
Reports a diff after each run — which files were added, changed, or need removing from production — and can hand you a ZIP of just the changed files, with a list of what to delete.
Audits internal links and reports any whose target was not generated, and compares against WordPress core's sitemap to surface pages that enumeration may have missed.
Optionally bundle extra files or directories that nothing links to (distan_extra_assets) — for assets referenced only from inside a script, such as a JSON file fetched by JavaScript.
Downloads the result as a ZIP.

Requirements Requires WordPress 6.0+, PHP 8.0+, writable uploads and working loopback HTTP (the site being able to request itself). ZIP downloads require ZipArchive. No direct SQL, no exec(), no external binaries, no external APIs. It runs on Local, MAMP, XAMPP, DDEV, Docker, or shared hosting, and works offline. Not for every case Distan is not suitable when the client updates the site themselves, or when forms, search, comments, or membership are required. If you need built-in deployment to many destinations, scheduled or change-based publishing, or managed static hosting, a larger tool such as Simply Static or Staatic will fit better. Structured data: add only what is known, inspect what is delivered Enable the schema.org option to add basic WebPage JSON-LD to exported pages without existing JSON-LD. Configure the production URL first. Directory and page exclusions use source paths relative to the WordPress home URL, one per line (for example /private/ and /contact/thanks/). Directory rules include descendants; page rules match only that path. Trailing slashes are ignored and query variants share the rule. Existing JSON-LD is retained even when excluded or disabled. Regenerate after changing settings. Every generated page is also checked for basic JSON-LD syntax and URL consistency, even when automatic additions are off. See the generation screen and delivery report for per-page outcomes, block numbers and issues. Existing data is never automatically repaired. Limited or failed inspection is explicitly recorded. Use Schema.org Validator and Google Rich Results Test for additional validation; this local audit does not guarantee search appearance.

1.9.2

Fixed failed pages being listed as production deletion candidates. Failed or interrupted runs retain the last successful diff baseline, so a successful retry includes all changed pages and genuine removals.
Blocked full, differential and template ZIP exports and manual dispatch until generation succeeds. Incomplete status and recorded errors remain visible after reloading the admin screen.
The distan_after_generate hook now fires only when generation finishes without recorded errors. Review integrations that previously used this hook to observe failed runs.
Asset read/copy failures now mark the run incomplete. Assets discovered inside CSS are recorded and retained during output cleanup.

1.9.1

Refactored: output-file and JSON-LD development URL checks now share one read-only detector with explicit environment inputs. Generated HTML and existing JSON-LD are not rewritten by diagnostics.
Fixed: escaped slashes and Unicode escapes in URL text are now inspected; similar hostnames and URL user-info no longer count as the development host. Source path boundaries, ports, Japanese paths and protocol-relative URLs use the shared rules. A production base matching the source base is not flagged.
Improved: development URL warnings explain that production URL settings alone may not replace escaped URLs. The environment check continues to read the saved generation record.
Updated: regression coverage and upgrade-testing instructions; upgrade tests preserve the installed plugin settings instead of deleting the old installation.

1.9.0

Added: optional basic WebPage JSON-LD for pages without existing JSON-LD. Uses the production URL and actual output path, plus the title, description and language available in the HTML. Off by default; requires a production URL and skips 404 pages.
Added: separate directory and exact-page exclusions for automatic additions. Existing theme, SEO-plugin and handwritten JSON-LD takes precedence, including empty or malformed blocks; exclusion does not remove existing data.
Added: read-only diagnostics on generated HTML, even when automatic additions are off. Detects empty or malformed JSON, development URLs including escaped URLs, page-type URL/canonical mismatches, and repeated or conflicting same-page @id definitions. Reference-only @id nodes are not treated as duplicate definitions.
Added: per-page addition and inspection results in the admin and delivery reports, with relevant results in differential and template ZIPs. Generation records preserve the findings; failed or incomplete inspections are explicitly marked.
Improved: warns when automatic additions are enabled without a production URL. Template handoff instructions now explain which page-specific metadata must be updated when reusing a page.
Scope: basic local syntax and URL checks only. No automatic repair, external context fetching, complete Schema.org validation, or guarantee of rich results or AI-search visibility.

1.8.0

Added: the Markdown export (content.md) can now be narrowed by post type and by publish date, and specific fixed pages can be force-included, so a site can be delivered whole while only, say, its articles within a date window are transcribed for an AI. With no filter set the export still covers every page as before (posts, custom types, fixed pages, the front page and archives). When you do filter, post types are chosen with checkboxes and the publish-date range is inclusive on both ends (either bound may be left open); while filtering, pages and archives drop out unless a page is added individually from the live title search, which force-includes it. The static HTML output (dist/) is unchanged by any of this.
Added: an optional "remove HTML comments" setting. When enabled, plain HTML comments left by plugins and themes are stripped from the delivered pages at write time. Comment-like text inside script, style, textarea and pre elements is shielded, and IE conditional comments are kept because they can wrap real markup. Off by default; the static output is unchanged unless you turn it on.
Changed: the admin screen's own styles and scripts are now cache-busted by file modification time, so an updated build always loads its current assets instead of a stale cached copy.

1.7.0

Improved: the distan:no-block-styles / distan:drop-assets markers now also take effect in a normal "static HTML" run, not only in the template export. On a marked page, the declared link and script tags and inline block styles are removed from that page's output; shared asset files are left in place so other pages that use them keep working.
Improved: the Markdown export (content.md) now preserves much more structure, so it reads better when fed to an AI. Tables become Markdown tables, code becomes fenced or inline code, images keep their alt text, and block quotes, bold/italic, ordered-list numbering, and horizontal rules are carried across. Each page section also records its source URL and meta description, and its title falls back to the first heading when a theme emits no title tag.

1.6.4

Updated: the plugin screenshots — the full admin screen and the in-app help.

1.6.3

Added: a "設定" link on the plugin's row on the Plugins screen, so you can reach the Distan screen from there.
Improved: the template export page selector can now be filtered live by title. Type in the box above the list and the candidates narrow as you go — useful when a site has many pages.

1.6.2

Fixed: the copy buttons for the template markers did nothing on a plain-http development host (for example http://site.local). The clipboard API is only available in a secure context, so a fallback is now used there.

1.6.1

Improved: the admin screen now follows the WordPress admin colour scheme instead of a custom palette. The accent tracks the colour scheme each user picks, and status badges are shown in neutral greys with red reserved for real problems such as broken links.
Improved: the section sub-navigation no longer sticks to the top of the screen, so it never covers a heading while scrolling. Dates are all shown in the site's timezone. The uncovered-URL list scrolls once it grows long, and the settings save bar keeps a little breathing room.
Clarified: the descriptions for "WordPress の痕跡を除く" and the template export now spell out that the former only removes invisible metadata (block CSS and plugin JS are left to the per-page markers), so the two are not confused.
Fixed: block styles the template export could leave behind. The distan:no-block-styles marker now also removes the inline block styles WordPress emits (wp-block-*-inline-css, the style/global-styles placeholders, and the wp-img-auto-sizes helper), not just the stylesheet link and global-styles block.
Updated: bundled Alpine.js to 3.17.1.

1.6.0

Added: template markers for the template export. A page can declare, in an HTML comment, what to drop from its one-page template -- the distan:no-block-styles marker removes the block library CSS and global styles, and distan:drop-assets wp-includes/ wp-content/plugins/foo/ removes scripts and stylesheets whose output path starts with the given prefixes (directory or single file). The referenced tags are removed with the files so nothing dangles, and the marker comments are stripped from the delivered template. Distan does not guess what is unused; it acts on what the template declares.
Added: take-up of uncovered URLs. Enumeration reads the database, so it cannot know about URLs a plugin registers dynamically — a form's thank-you page, a virtual route. Distan already surfaces these (WordPress core's sitemap declares them, or a generated page links to them), but acting on the gap used to mean hand-writing a distan_sources filter. The generate screen now shows those uncovered URLs and lets you decide per URL: include it on the next run, stop offering it, or leave it pending. The default is opt-in — nothing is taken up unless you choose it, so a page the site should not ship never sneaks into the deliverable — and your decisions are remembered, so you do not pick the same URLs again every time. A free-text box takes any other same-site URL you want generated (for example one you spotted in the broken-links report). Taken-up URLs join the queue on the same footing as distan_sources: counted, deduplicated, and shown in the diff like any other page.

1.5.0

Added: a template export. Pick one generated page and Distan hands you a ZIP of just that page plus only the assets it actually references — its CSS, JS, fonts and images, followed recursively through stylesheet url() and @import — laid out at their real relative paths so the page opens and renders as-is. It is meant for handing an outside coder a shell to build a new special page against the site's shared header and footer, without shipping the whole site's imagery. Links to other pages are expected to dangle in a one-page handoff, and a README.md at the ZIP root spells out the ground rules for the coder (leave the head/header/footer untouched, keep relative paths, add new assets under their own folder). The page selector appears on the generate screen after a run and can be turned off with the new テンプレート書き出し setting. Distan does not attempt to carve out the content region — that was the failure mode of the earlier structure map — so the full page is delivered and the coder is told which part to replace.
Changed: the manifest now records the environment it was generated in (site URL, site name, link style, production URL). The report, differential ZIP and template export read these from the manifest instead of asking WordPress, so a deliverable reflects the site as it was generated. In particular, changing the Production URL or link style after a run no longer disturbs an export taken from an earlier run. Manifests written before this release fall back to the current values.

1.4.1

Changed: marked as tested up to WordPress 7.1.

1.4.0

Changed: refreshed the admin screen. The help ("使い方") now opens as a centred modal from a button in the page header, instead of a floating panel in the corner, and shows a short guide to Distan and each setting. Type sizes, line heights, and spacing were tightened to a consistent scale, and the settings sections (基本設定 / 追加オプション / 公開・デプロイ) read as clearer breaks so the page is easier to scan. No change to what Distan generates.

1.3.0

Added: a differential download. Alongside the existing full ZIP, Distan can now hand you a ZIP of only the files added or changed since the last generation, laid out at their real paths so it unzips straight onto production — no re-uploading the whole site. Files that were in the previous delivery but were not generated this time are listed in a bundled DELETE.txt (with a short distan-diff.md delivery note), so removals are explicit rather than forgotten. The full ZIP remains for first delivery and larger milestones; the differential ZIP is for routine updates.
Added: content-hash change detection. Each generated page's HTML is hashed, so a page whose path is unchanged but whose content actually changed — an edited notice, a refreshed "latest posts" list — is now reported as modified. Previously the diff could only see paths that were added or removed, so an in-place edit slipped through. Assets remain tracked by add/remove. On the first run after upgrading, hashes are recorded but nothing is flagged as modified; detection begins on the next run, so upgrading never mass-flags every page.
Added: the distan_manifest_source filter (db by default, or output). In output mode the diff baseline is carried as a portable manifest inside the output tree (.distan/manifest.json) so it travels with the deliverable — the right choice when you generate in one place and deploy from another (local or CI to a separate server). A deny rule is written beside it (.distan/.htaccess) to keep the meta directory out of public serving on Apache; other servers need an equivalent location rule. The default db mode keeps the baseline in the WordPress option exactly as before, so nothing changes unless you opt in.
Added: a read-only diff status line on the generate screen — the current baseline source, the last generation time, and a notice when there is no previous baseline, so the "everything counts as new next time" case is visible before it happens. The differential download can be turned off with the new 差分 ZIP setting; change detection keeps running regardless, so enabling it later still produces a correct diff from that point on.

1.2.0

Added: optional support for pages whose content changes with a URL query parameter (for example /guide/?tab=faq versus /guide/?tab=news). List the query keys that matter with the new distan_variant_keys filter, and each value is written as its own static file under a path like /guide/tab-faq/, with internal links to those URLs rewritten to match. Register the specific variant URLs through the existing distan_sources filter so they are discovered and generated. Off by default: with no keys listed, query strings on page URLs are dropped exactly as before, so nothing changes unless you opt in. Query keys you do not list (utm_source, fbclid and the like) are ignored, so this cannot multiply into unwanted files. The folded path format can be changed with the distan_query_variant_segment filter. Because a plain static host can only address files by path, the query has to be folded into the path — the published URL becomes /guide/tab-faq/ rather than /guide/?tab=faq.

1.1.9

Added: a help tool pinned to the bottom-right of the Distan admin screen. A small "使い方" button opens a short panel that walks a first-time user through the flow — check environment, generate, then receive or deploy — with each step jumping to the matching section, plus pointers to where the detailed, contextual help already lives. It does not repeat that detail. Behaviour is open/close only (Alpine, already bundled); no new JavaScript file is added. This is admin-only UI — nothing in the generated output changes.
Changed: simplified the admin masthead to a single calm line under the title, replacing the uppercase English tagline and the two-sentence lede. Admin-only cosmetic change; nothing in the generated output changes.
Changed: updated the admin colour scheme to match the Distan book cover — a warm cream background with a brown accent on buttons and highlights, in place of the previous grey background and green accent. Admin-only; nothing in the generated output changes.

1.1.8

Added: a sticky in-page navigation bar on the Distan admin screen. It follows the page as you scroll and jumps to the environment / generate / settings sections, with the current section highlighted as you move through them. This is admin-only UI — nothing in the generated output changes.

1.1.7

Added: provenance on every generated entry. Each page now records what produced it (the post and its ID, the taxonomy term, the archive page), so the generation report names a change — for example "記事タイトル [投稿 #123]" — instead of only its file path. Removed pages keep their label, carried forward from the previous run. This is descriptive only: nothing is skipped or cached, and every page is still rendered on every run.
Added: the distan_sources filter, a first-class way to register URLs that enumeration cannot discover on its own (plugin-generated routes, virtual pages). Entries are built with Distan_Collector::make_item(), merged before de-duplication, and carry provenance, so an added URL is counted and appears in the diff like any built-in page. This differs from distan_collect, which remains available for raw last-resort edits.
Fixed: a URL added through the distan_collect filter could collide with a built-in page's output path and silently overwrite it, because the filter ran after de-duplication. The queue is now de-duplicated once more after the filter, so a collision drops the later entry instead of overwriting an already-generated file.
Added: core-sitemap reconciliation. After a run, Distan reads WordPress core's own sitemap (wp-sitemap) in-process — no HTTP, no crawling — and lists in the report any URL the sitemap declares but the run did not generate, so a coverage gap is visible rather than silent. Optionally (filter distan_use_core_sitemap) those URLs can also seed the queue as a supplementary source; it stays a supplement, since core sitemaps honour noindex and can be disabled.

1.1.6

Added: the distan_extra_assets filter, to bundle files or directories that no page links to. Distan collects assets by inspecting the generated HTML and CSS, so a file referenced only from inside a script (for example fetch('../assets/json/data.json')) is not found and is left out. Listing a file or directory here includes it anyway, through the same pipeline as linked assets: the theme (and uploads) path is flattened the same way, so a script that fetches a relative path keeps working; the file is recorded in the report and is never removed by the cleanup; executable extensions are still refused; and anything resolving outside the WordPress root is skipped. URLs inside these files are not rewritten, so they suit data files such as JSON. The default remains reference-based — nothing extra is copied unless you list it.

1.1.5

Added: a guard against two people (or two browser tabs) starting a generation at the same time. Progress and output are tracked in a single shared job and written to one output directory, so overlapping runs previously corrupted each other. Starting a run while one is already in progress is now refused with a message naming who started it and when. A run that was abandoned mid-way (browser closed) is detected as stale after a couple of minutes (filter distan_job_stale_after) so it never blocks future runs.

1.1.4

Fixed: on top-level pages, the import map address for block-theme modules (e.g. @wordpress/interactivity, used by the navigation block) was written as a bare relative path (wp-includes/…) with no ./ prefix. An import map address that is a relative URL must start with /, ./ or ../, so browsers treated the bare form as a null value and every module import on the page failed ("Failed to resolve module specifier … blocked by a null value"). Deeper pages already got a ../ prefix and worked; top-level pages now get ./ so interactive block-theme features work in relative-link mode too. Sites delivered with an absolute public URL were unaffected.

1.1.3

Fixed: pages with a multibyte (e.g. Japanese) slug are now written to disk under their real decoded name (テスト/index.html) instead of a literal percent-encoded name (%E3%83%86…/index.html), while links keep the percent-encoded form. A browser or web server decodes an href before it looks for the file, so the previous literal name only matched when a URL was double-encoded: such pages opened by double-click but broke when reached through their own link (and were silently served by WordPress when the export happened to sit inside a live site). Assets already followed this rule; pages now match. The link audit decodes links before checking the filesystem so it does not report false broken links.

1.1.2

Changed: the Markdown export (content.md) is now streamed one page at a time instead of collecting every page's text and joining it at the end. On sites with a very large number of pages this keeps memory flat and stops the job data from growing with page count, the same way binary assets were made stream-copied in 1.1.1. The output file is byte-for-byte identical to before.

1.1.1

Changed: binary assets (images, PDFs, fonts, video) are now stream-copied instead of read fully into memory, so large files no longer inflate memory during generation. Stylesheets are still read in memory because their url() references are rewritten.
Added: files above a size threshold (10 MB by default, filter distan_large_file_threshold) are listed in the report as "large files". They are still copied automatically; the list is a heads-up for delivery size and CDN decisions, not a manual step.

1.1.0

Added: optional sitemap.xml generation. Built from the pages that were actually generated, using the production URL, in the standard format Google Search Console accepts. Author and date archives are never collected, so IDs like ?author=1 cannot leak into it.
Added: sitemap exclusions by slug prefix (/private/ excludes everything under it) or substring (draft excludes any URL containing it), via a setting or the distan_sitemap_exclude filter.
Added: optional minimal robots.txt (Allow: /, plus a Sitemap: line when the sitemap is enabled).
The development-URL audit now also scans sitemap.xml and robots.txt, so a leftover development host in them is reported like anywhere else.
The diff-based cleanup now keeps sitemap.xml, robots.txt and the Markdown files, so they are not removed on the next run.

1.0.0

First stable release. Classic and block (FSE) themes are both supported and have been verified across the hard cases: internal links and multibyte slugs, JSON-LD and canonical/OGP, srcset and path flattening, CSS url() and webfonts, custom post types and fields, diff-based cleanup, Markdown export, hooks, and a safe uninstall that never touches the delivered files.

0.9.18

Fixed: multibyte (e.g. Japanese) page slugs in internal links were corrupted (0x80-0x9F bytes turned into "_"), breaking the link. WordPress emits same-origin links as raw UTF-8; Distan now normalises every link path to a consistent percent-encoded form, so multibyte slugs survive intact in both the output filename and the rewritten link.
Added: the generation report and the environment check now surface how many development-domain URLs remain in the output (e.g. in JSON-LD or canonical/OGP tags when the Production URL is unset), so a forgotten Production URL is caught before and after delivery.
Fixed: the link audit reported existing pages with multibyte slugs as broken links, because it looked for a decoded name while the file is written with a percent-encoded name. The audit now matches the encoded form.
Fixed: url() references inside external CSS were copied verbatim, so theme fonts/images went uncopied and uploads-absolute paths (url("/wp-content/uploads/…")) leaked into delivered files. CSS url() is now resolved, flattened (theme -> assets/, uploads -> media/), its targets copied, and the reference rewritten relative to the stylesheet.

0.9.17

Added the distan_dispatch action hook and an optional manual "デプロイ" (dispatch) button on the generate screen. After reviewing the generated output, pressing the button fires distan_dispatch so a project can promote the reviewed build (git push, rsync, a build webhook). Distan stores no approval state — only the time of the last dispatch, shown on screen. The button is off by default; enable it in settings.

0.9.16

Added the distan_after_generate action hook, fired when a generation run completes, for wiring up automatic deployment (git push, rsync, build webhooks, etc.).

0.9.15

Redesigned the admin screen ("Distan Dispatch"): a packing-slip worksheet look with monospace figures for counts and file paths, a single stamp-green accent, and improved contrast.

0.9.14

Added the distan_url_replacements filter for custom production URL replacements (applied across the whole output, including JSON-LD).
Added an optional Markdown export (content.md) that combines every page's main content into one file for AI tools such as Gemini Notebook (formerly NotebookLM). URLs are rewritten to the production site URL; an optional content.local.md keeps development URLs.

0.9.13

Initial public release.

Distan

标签

下载

详情介绍:

安装:

屏幕截图:

常见问题:

更新日志: