In regex_replace action, consider doing line-by-line matching instead of entire file

[drevell]

TL;DR

As a template developer, I want ^ and $ to match beginning and end of line, because that's what I'm used to from grep. As a template developer I'm probably not aware of how to do RE2 multiline regexes to manually enable this behavior using (?m:my_regex).

cc @sethvargo @pdewilde @verbanicm who have been involved in related conversations.

Upside:

• Users don't have to understand RE2 multiline syntax and semantics to match beginning/end of line
  Downside:
• If users want to match a string that spans multiple lines within a file, we have to give them a special mode.

Implementation ideas:

• split each input file on \n before doing the regex replace
• add a new mode for users that want to do a search-and-replace that spans multiple lines. Something like:

  action: `regex_replace`
  params:
    paths: ['my.txt']
    replacements:
      mode: 'file_at_a_time'  # <---- here we need a new mode
      regex: |
        my
        regex
        with
        newlines
      replace_with: 'whatever'
  

Detailed design

No response

Alternatives considered

No response

Additional information

No response

[pdewilde]

split each input file on \n before doing the regex replace

This would make it more grep-like, though it would have different semantics than for example my text editor, where $ and ^ are done per-line, but I can still replace \n if I need to.

I was thinking just having users use \n instead of $ or ^, but unfortunately that breaks on the first/last line of a file, where there is no preceding/trailing \n.

[drevell]

This would make it more grep-like, though it would have different semantics than for example my text editor, where $ and ^ are done per-line, but I can still replace \n if I need to.

@pdewilde, could it be that your editor is in multiline mode by default? It doesn't look like we have the ability to make that happen in Go 😞 . Unless we resort to using https://pkg.go.dev/regexp/syntax#Prog which is a lower-level API to regexes that looks like dragons. But at first glance it does seem desirable to default to multiline mode.

We could wrap the user's regex in (?m:their_regex), but that seems gross. Maybe acceptably gross though.