v0.9.0

The first release. Everything below is new, so this section describes what
Blogin does rather than what changed.

Commands

• blogin build renders a content tree to static HTML. A rebuild does only the
  work the change requires and lands byte-identical to a build from scratch.
• blogin serve previews the site on loopback and rebuilds as you edit, with
  the open page reloading itself over a WebSocket. It builds into
  .blogin-preview/, so serving never writes the directory you deploy.
• blogin init scaffolds a site that builds with nothing to fill in first,
  against none, bootstrap5, pico, or bulma.
• blogin new writes a post with its front matter filled in.
• blogin clean removes both trees a build can write and refuses any target
  outside the site. Naming one with --out makes it the only target.
• Every option has a long and a one-letter spelling, and short options combine
  as -fv. blogin --version reports the version it was built from.

Content

• CommonMark, with GitHub Flavored Markdown for tables, task lists,
  strikethrough, and fenced code.
• Footnotes, reference links, definition lists, and attribute lists on links and
  images.
• Math and Mermaid diagrams, parsed into markup a client-side renderer draws.
• Shortcodes, with youtube and figure built in and a shortcodes/ directory
  for your own.
• Front matter for the title, date, slug, summary, ordering, drafts, table of
  contents, and redirect aliases.
• Summaries from front matter, a <!--more--> marker, or the opening block.

Templates

• A HAML engine with an expression language of its own: no assignment, no
  user-defined functions, and no host language to call into. Anything outside it
  is an error naming the file, line, column, and construct.
• Layouts per section, by directory or by config.
• Themes, a directory of layouts and assets your own files override file by
  file.
• cache-fragment, which reuses a rendered fragment wherever the values it read
  are the same, so reuse cannot serve one page's markup to another.
• Data files, JSON and a YAML subset, site-wide and scoped to a directory.

Output

• Atom, RSS, and JSON feeds, site-wide and per section, plus a sitemap,
  robots.txt, and a 404 page.
• Client-side search against a prebuilt index, with the script and its styling
  generated by the build.
• Taxonomies, with a page per term and an index, paginated like any listing.
• Optional extensionless URLs.
• More than one language, each built into its own subtree with translations
  matched by filename.

Assets

• CSS and JavaScript minified, and assets named for a hash of their content so a
  far-future cache never serves a stale file.
• Responsive image variants with a srcset, when an image resizer is present.
• CSS framework profiles for Bootstrap 5, Pico, and Bulma, which class the same
  semantic HTML without a layout knowing which framework is selected.

Platforms

• macOS through Homebrew, and Linux x86_64 and arm64 as statically linked
  binaries that run on any distribution.
• Windows through WSL2.
• Builds from source with clang on macOS and Linux, and with GCC 14 or newer on
  Linux.
• No runtime dependencies. Markdown, HAML, JSON, YAML, the asset pipeline, and
  the preview server are all in the binary.