Feature #35035

Refactor text formatting to HTML::Pipeline

Added by Martin Cizek 2 months ago. Updated 2 months ago.

Status:NewStart date:
Priority:NormalDue date:
Assignee:-% Done:


Category:Text formatting
Target version:-


After merging the CommonMark patch #32424, Redmine will contain an HTML::Pipeline implementation limited to this new format. It makes then perfect sense to extend the new concept to text formatting generally.

The current implementation looks like:

While the new Markdown (CommonMark) implementation is designed like:

Migrating text rendering to HTML::Pipeline should provide several benefits:
  • Better separation of concerns.
  • Cleaner, safer and DRY code.
  • Individual formatters can be freed from excessive restrictions that were introduced as XSS prevention. Especially the absolute prohibition of HTML tags can be pain in specific cases.
  • More consistent behavior among formatters (e.g. HTML is currently escaped in Textile, but dropped in Markdown),
  • Controlled level of security without the need to rely on individual formatter implementations that often use poorly maintained libraries.
  • More flexibility for Redmine developers and Redmine plugin developers. Adding some hooks would be nice and would allow e.g. implementing mermaid charts as a custom syntax highlighter - in the same way as GitLab.

Some inspiration can be found in GitLab sources, but our pipelines would probably won't need to go that complex.

If there were no other timing considerations, it would make sense to implement this together with Redcarpet removal.

Text_Formatting-Redmine-4.2.png (69.9 KB) Martin Cizek, 2021-04-05 15:18

CommonMark-HTML_Pipeline.png (29.8 KB) Martin Cizek, 2021-04-05 15:19


#1 Updated by Jens Krämer 2 months ago

Totally agree, the potential benefits of switching to html pipeline are quite big. This kind of modularization will allow for many interesting features to be implemented in core or by plugins in the future.

As stated in #32424-37, I'd even like to move the redcarpet renderer to the pipeline, so it could still be used in the future as part of core, or extracted into a plugin, by those unwilling or unable to switch to common mark.

Also available in: Atom PDF