Change Log for rib

0.9.0.0 (UNRELEASED, DEV)

API
- Dropped path and path-io in favour of good ol' FilePath
  - This also lifts the restriction with absolute paths
Misc changes
- #145: CLI arguments have been revamped
  - Added --input-dir/--output-dir to override these paths
  - Accept host string in addition to port number
  - Exposed Rib.Shake.getCliConfig to get full CLI configuration
  - Allow customizing fsnotify ignore list
- #141: Allow quiet logging (useful when rib is used as a library)

Dependency upgrades
- mmark: 0.0.7.2
- megaparsec: 0.8
- clay: 0.14
- shake: 0.8.15
New features:
- Added Dhall parser, Rib.Parser.Dhall
- Add Rib.Extra containing useful but non-essential features
MMark, extensions removed:
- ghcSyntaxHighlighter: we already have skylighting (which supports more parsers than Haskell)
- obfuscateEmail: requires JS, which is not documented.
API changes:
- Introduced Route functionality for simpler management of static routes.
  - Removed buildHtmlMulti, buildHtml, readSource functions and Source type.
- Introduced Rib.Shake.forEvery to run a Shake action over a pattern of files when they change.
- Exposed Rib.Shake.writeFileCached
- MMark.parse and Pandoc.parse now automatically append path to ribInputDir and do not return Either.
- Added MMark.parseWith (and parsePureWith), to specify a custom list of mmark extensions
Bug fixes
- #95: Fix Shake error resource busy (file is locked)
- #97: Fix Shake error AsyncCancelled when server thread crashes
- #96 & #108: Drop problematic use of Shake cacheActionWith

Advance nixpkgs; require Shake >=0.18.4
Major API simplication: no more type class!
- Allow user to specify their own source parser as a Haskell function
- Removed types Document and Markup in favour of Source
- Expose ribInputDir and ribOutputDir for use in custom Shake actions
Bug fixes:
- #63: create intermediate directories when generating post HTML
- #70: Don't crash on Shake errors
- Fix unnecessary rebuild of all files when only one file changed
  - #66: Use caching (via Shake's cacheActionWith), to avoid writing HTML to disk until it has changed.

This release comes with a major API refactor. Key changes:

Added MMark support, as an alternative to Pandoc
Allows using arbitrary records to load metadata
- This replaces the previous complex metadata API
Added Document type that uses the custom metadata record
Add top-level Rib import namespace for ease of use
Remove the following:
- JSON cache
- Rib.Simple
Support for Table of Contents via MMark

Other changes:

Rib.Pandoc:
- Export render' and renderInlines' (the non-Lucid versions)
- Re-export Text.Pandoc.Readers so the library user does not have to directly depend on pandoc only to render its documents.
Rib.App: The run funtion now takes two more arguments, specifying the input and output directory, which are no longer hardcoded.
Rib.Simple: add LaTeX to default list of readers
Rib.Server: Remove ".html" detection magic from URLs