From a02ba006de66f86ae766eb24f4158c7614bf3b5f Mon Sep 17 00:00:00 2001
From: Yehowshua Immanuel <yehowshua@yehowshuaimmanuel.com>
Date: Tue, 10 Dec 2024 13:35:31 -0500
Subject: [PATCH] stopping point8 before repairing parser

---
 TODO.md                   | 53 ++++++++-------------------------------
 src/RTLILParser/Parser.hs |  6 ++---
 2 files changed, 13 insertions(+), 46 deletions(-)

diff --git a/TODO.md b/TODO.md
index 4bb47ca..f07f6d7 100644
--- a/TODO.md
+++ b/TODO.md
@@ -12,64 +12,29 @@
  - [ ] add validation section to README
  - [ ] just support Cell memV2
  - [ ] for value, what happens if we have 0 binary digits
- - [ ] modify parser to ignore comments... - should probably wait until 
-       parser is finished before asking
  - [ ] modify AST to support src tracking - needed to allow for human 
        readable and correctable validation errors
  - [ ] when writing simulator, must specify directions on cell ports
- - [ ] in the <process>, why are we allowed to have <assign-stmt> before
-       and after the optional <switch> stmt? If the ordering here is
-       semantically significant, then modify AST to only have single
-       `[AssignStmt]` field and update parser behavior accordingly.
- - [ ] inspect Chris's mini-RTLIL
- - [ ] add RST grammar file to repo
- - [ ] name parsers so that that we know where failures occured
+ - [ ] track RTLIL EBNF spec vs RTLIL Lex/Yacc implementation
+       [deviation issue][deviation-issue]
+ - [ ] Update issue above with fact that `<memwr>` is also missing
  - [ ] may want to also derive equality statements
- - [x] replace both `pEol *> pMaybeWs` and `pEol <* pMaybeWs` 
-       with `pEolAndAdvanceToNextNonWs`
  - [x] Check inline sequencing of whitespace parsers in do blocks.
        Terminating instances of `pWs` should be preceeded by `<*`
- - [ ] Verify no backtracking needed when sequencing `many` parsers.
-       Basically, we want to make sure that the argument of the `many`
-       parser doesn't conflict(exhibit a partial early match) with 
-       the argument of the parser after the argument of the `many` parser.
 
 # Parser Development
- - [x] Sync
- - [ ] Process
- - [x] Finish `pCell` with `pCellEndStmt`
- - [ ] Rewrite `pWireStmt` and `pMemoryStmt` using do-notation...
- - [x] Remove all instances of `_ <-`
- - [ ] Module
  - [ ] Remove weird GHC imports
+ - [ ] Test `fsm1.il` which has recursive slices on line 180
  - [ ] Embed locs in AST
- - [ ] Scrap `pEolAndAdvanceToNextNonWs` and use `tok`
- - [ ] Remove `preProcessDiscardComments` from exports
- - [ ] Install README in dir containing `Parser.hs`
- - [ ] Discuss deviations of parser against Yosys behaviors
- - [x] Are the `try` statements in `pWireOption` correctly constructed?
- - [ ] Consider the very weird case where the process body has nothing,
-       thus, `pEolAndAdvanceToNextNonWs` may never get invoked in any of
-       the sub-parsers encapsulated in `pProcessBody`. Do we need to 
-       advance whitespaces so we can hit `<proc-end-stmt>`?
-      - [ ] What are the implications for other parsers?
-
-            I think that in this case we're OK as `<proc-stmt>` necessarily
-            precedes `<process-body>` and `<proc-stmt>` terminates in an EOL
-            parser that advances to the next non-whitespace.
-
-            I still need to verify how other parsers behave. For example, what
-            happens if we have a cell with no `<cell-body-stmt>`
+ - [ ] Name new parsers
+ - [ ] Remove `a` and `val` from exports
 
 # Parser Verification
  - [ ] I think only EOL terminated parsers should be responsible 
-       for advancing the Parsec scanner to the next non-space...
+       for advancing the Parsec scanner to the next token
  - [ ] lift grammar into prover and show that all EOL terminated parsers
        are either followed by EOF or a keyword such "module", "autoidx",
        etc
- - [ ] first, manually inspect switch parser and try and see if it
-       can infinitely recurse... Then empirically validate against
-       corpus
 
 # Simulation Behavior
  - [ ] Figure out the computational semantics of what it means to
@@ -82,4 +47,6 @@
  - [ ] Subsequent IR passes may emerge natrually
 
 # Optimizations
- - [ ] squeeze out recursive slices
\ No newline at end of file
+ - [ ] squeeze out recursive slices
+
+[deviation-issue]: https://github.com/YosysHQ/yosys/issues/4811
\ No newline at end of file
diff --git a/src/RTLILParser/Parser.hs b/src/RTLILParser/Parser.hs
index 38b705a..29de085 100644
--- a/src/RTLILParser/Parser.hs
+++ b/src/RTLILParser/Parser.hs
@@ -67,11 +67,9 @@ import RTLILParser.Primitives(
 -- taken from: https://yosyshq.readthedocs.io/projects/yosys/en/0.47/appendix/rtlil_text.html
 -- parsers below are split int sections from the above link
 
-runParser str = case parse pFile "pFile" preProcessedFile of
--- runParser str = case parse pCell "pCell" preProcessedFile of
+runParser str = case parse pFile "pFile" str of
     Left err -> error $ show err
     Right val -> val
-    where preProcessedFile = str
 
 -- identifiers
 pId :: Parser Id
@@ -305,6 +303,8 @@ pWireOption :: Parser WireOption
 pWireOption = p <?> name where
     name = "WireOption"
     p =
+        -- We technically don't need the first `try` below so this is a
+        -- stylistic choice. The other `try` statements are needed.
         try (WireOptionWidth  <$> (string "width"  *> pWs *> pInteger)) <|>
         try (WireOptionOffset <$> (string "offset" *> pWs *> pInteger)) <|>
         try (WireOptionInput  <$> (string "input"  *> pWs *> pInteger)) <|>