Inconsistencies in Replace_Trans_Action and Add_Trans_Action

[podcherklife]

Replace_* and Add_* families of operations share a broad semantics of "mutate value if condition matches", but have some idiosyncrasies, here I am taking specifically about transition actions.

About transition actions:

• These two declarations produce different values for action at compilation time:

IF ~~ EXIT // None()
IF ~~ DO ~~ EXIT // Some("")

• when transition is serialized to disk, Some("") is essentially converted to None():

  weidu/src/dlg.ml

  Lines 311 to 317 in fc5a92f

  	(match t.action with
  	| None -> ()
  	| Some(s) when
  	(try let _ = Str.search_forward letter_regexp s 0 in
  	false
  	with Not_found -> true) -> ()
  	| Some(real_action) -> flags := !flags lor 4) ;

  
  (Actually, that regex does more, but it is reasonable to think about is as Some("") -> None() normalizer)
• All Replace_* operations completely skip None() values.
• All Add_* operations always test their conditions, if value is None() then it is treated as empty string:

  weidu/src/dc.ml

  Lines 844 to 846 in fc5a92f

  	match trans.Dlg.action with
  	| None -> if passes d_when "" then trans.Dlg.action <- Some(action_text)
  	| Some(str) -> if passes d_when str then

Now, there is a symmetry within Replace_* and Add_* operation groups, but when combined with serialization behavior it creates an interesting issue when operation behaves differently depending on whether its target was declared locally or read from disk:

Declaration	Semantics	Add_Trans_Action	Replace_Trans_Action	Same behavior
Just declared IF ~~ DO ~~ EXIT	Can match with condition IF ~^$~	Runs	Runs	✅
Load old IF ~~ DO ~~ EXIT	Can match with condition IF ~^$~	Runs	Skips	❌
Just declared IF ~~ DO ~~ EXIT	Update value	Runs	Runs	✅
Load old IF ~~ DO ~~ EXIT	Update value	Runs	Skips	❌

Proposal

Add the same "None() is treated as empty string" behavior for all *_Action operations.
While that would break symmetry among Replace_* operations - how axis, it would introduce symmetry for what axis among *_Action operations.

Counter argument:
This breaks compatibility.

Counter counter arguments:

• I could not find a single use of REPLACE_* with condition. This does not mean that nobody uses that, but points to Replace semantics being naturally similar to IF condition and also might indicate that people already found out that using UNLESS is unreliable so they opted for different solutions.
• Technically it would improve compatibility between mods, because there would be no risk of making transition action "unreplacable", like in this synthetic example:

IF ~~ DO ~initialAction()~ EXIT
...
REPLACE_TRANS_ACTION foo 0 0  ~initialAction()~ ~~
...
REPLACE_TRANS_ACTION foo 0 0  ~^$~ ~anotherAction()~ // this will not run, because at this point action is None()

[podcherklife]

Oh, also a disclaimer:

This is not a real issue that is based on some user feedback.
I am reimplementing a compiler for .d files and this is a part of a test set that I am working on , me thinking it could be fixed in weidu rather than simply copied as is.
If this direction deemed acceptable, I can provide code for that.