line continuation in a comment

[hjoliver]

Describe the bug

Commenting out a line that ends in a line-continuation back-slash causes the subsequent line to become part of the comment.

Release version(s) and/or repository branch(es) affected? Current 7.8.x and master

Steps to reproduce the bug

[scheduling]
   [[graph]]
      R1 = """
         # foo => \
            baz => \
            qux
              """

That is equivalent (wrongly) to this:

[scheduling]
   [[graph]]
      R1 = """
         # foo => baz => qux
              """

Expected behavior

We discourage use of line continuation characters in the graph because they are no longer needed. Using natural continuation (implied by trailing =>):

[scheduling]
   [[graph]]
      R1 = """
         # foo => 
            baz => 
            qux
              """

that is equivalent (correctly) to:

[scheduling]
   [[graph]]
      R1 = """
         # foo => 
            baz => qux
              """

Additional context

This comment-continuation effect will apply anywhere in the file, but it is most likely to be encountered in graph strings (because users quite commonly comment bits of a graph out, and may not be aware that trailing \ is not needed for continuation in graph strings).

Unfortunately line continuation markers are stripped first thing during file parsing, where the context is unknown. We could ignore continuation markers if preceded anywhere on the line by a comment character - but need to be careful about # inside a string, if not inside a graph string? (which might be fiendishly difficult? might require abandoning the simple handling of line continuation first up during file parsing?).

A user at NIWA hit this problem last week, but it is the first time I've ever seen it, so could be classified rare bug, no desperate need to fix it quickly if difficult to do.

Pull requests welcome! This is an Open Source project - please consider contributing a bug fix yourself (please read CONTRIBUTING.md before starting any work though).

[dpmatthews]

Potentially slightly dangerous to change the behaviour now? As a minimum we should document the behaviour.

[hjoliver]

Potentially slightly dangerous to change the behaviour now?

Tough call ... I don't think anyone would be relying on the current behaviour would they? At least not deliberately.

As a minimum we should document the behaviour.

Yes, I'll reference this page on cylc-doc for Cylc 8. We should also document this for 7.8.5 though.

[hjoliver]

Case 2

(From @dpmatthews on Riot chat)

       #c => \_
           d => e

I've added a space after the continuation marker and it gets ignored in this case. However, for an uncommented line this would result in an error so this behaviour is not consistent.

[hjoliver]

Stop joining lines within multi-line strings. Reimplement within graph parser.