From d60e690445ab702b570e15ca7a23e37a4882fd15 Mon Sep 17 00:00:00 2001
From: Reid 'arrdem' McKenzie <me@arrdem.com>
Date: Sun, 14 Jun 2020 12:33:26 -0600
Subject: [PATCH] Starting to stub out a syntax analyzer

---
 src/python/flowmetal/parser.py          |   7 +-
 src/python/flowmetal/syntax_analyzer.py | 146 ++++++++++++++++++++++++
 2 files changed, 149 insertions(+), 4 deletions(-)
 create mode 100644 src/python/flowmetal/syntax_analyzer.py

diff --git a/src/python/flowmetal/parser.py b/src/python/flowmetal/parser.py
index 850c733..eda3494 100644
--- a/src/python/flowmetal/parser.py
+++ b/src/python/flowmetal/parser.py
@@ -80,7 +80,6 @@ class SymbolToken(NamedTuple, TokenBase):
     raw: str
     pos: Position
 
-
 class KeywordToken(NamedTuple, TokenBase):
     """A read keyword."""
     data: str
@@ -335,7 +334,7 @@ class Parser(SexpParser):
             buff = str(rtb)
             return CommentToken(buff, buff, pos)
 
-## Parsing
+## Parsing interface
 def parses(buff: str,
            parser: SexpParser = Parser,
            source_name=None):
@@ -365,7 +364,7 @@ def parse(file: IO,
     )
 
 
-## Loading
+## Loading interface
 def loads(buff: str,
           parser: SexpParser = Parser,
           source_name=None):
@@ -395,7 +394,7 @@ def load(file: IO,
     )
 
 
-## Dumping
+## Dumping interface
 def dump(file: IO, obj):
     """Given an object, dump its s-expression coding to the given file-like object."""
 
diff --git a/src/python/flowmetal/syntax_analyzer.py b/src/python/flowmetal/syntax_analyzer.py
new file mode 100644
index 0000000..98c32eb
--- /dev/null
+++ b/src/python/flowmetal/syntax_analyzer.py
@@ -0,0 +1,146 @@
+"""
+The parser just parses and tokenizes.
+
+The [syntax] analyzer interprets a parse sequence into a syntax tree which can be checked, type inferred and compiled.
+"""
+
+from abc import ABC, abstractmethod
+from io import StringIO, IO
+from typing import NamedTuple
+
+import flowmetal.parser as p
+
+### Types
+## We are not, in fact, sponsored by Typelevel LLC.
+class TypeLevelExpr(ABC):
+    """A base class for type-level expressions."""
+    pass
+
+
+class GenericExpr(TypeLevelExpr, NamedTuple):
+    """'invocation' (application) of a generic type to Type[Level]Exprs."""
+    pass
+
+
+class TypeExpr(TypeLevelExpr, NamedTuple):
+    """A bound (or yet to be bound) type level symbol."""
+
+
+## Now down to reality
+class ValueLevelExpr(ABC):
+    """A base class for value-level expressions."""
+
+
+class AscribeExpr(ValueLevelExpr, NamedTuple):
+    """Ascribe a type (via type-level expression) to a value-level expression."""
+
+
+class InvokeExpr(ValueLevelExpr, NamedTuple):
+    """(a ⊢ (fn A ⊢ B) [...] ⊢ A) ⊢ B"""
+
+
+class IfExpr(ValueLevelExpr, NamedTuple):
+    """(if test a ⊢ A b ⊢ B) ⊢ (Variant A B)."""
+
+
+class LetExpr(ValueLevelExpr, NamedTuple):
+    """Let a single binding and wrap a body. Yes one. N-ary let is an abstraction."""
+
+
+class DoExpr(ValueError, NamedTuple):
+    """do a procedure ahem sequence of things.
+
+    (do a b c ... ω ⊢ Ω) ⊢ Ω
+    """
+
+ProcExpr = DoExpr  # ain't broke don't fix it
+
+class MappingExpr(ValueLevelExpr, NamedTuple):
+    """Mappings require their own constructor expression due to local/symbol references."""
+
+
+class SetExpr(ValueLevelExpr, NamedTuple):
+    """Sets require their own constructor expression due to local/symbol references."""
+
+
+class ListExpr(ValueLevelExpr, NamedTuple):
+    """While round () lists are generally InvokeExprs, [] lists are constructors like sets and maps."""
+
+
+## Reader implementation
+class SexpAnalyzer(ABC):
+    """A base class for Analyzers."""
+
+
+class Analyzer(SexpAnalyzer):
+    """A reference Analyzer implementation.
+
+    Walks a parsed token tree, building up a syntax tree.
+    """
+
+    @classmethod
+    def read(cls, token: p.TokenBase):
+        if isinstance(token, p.WhitespaceToken):
+            ## Whitespace tokens are discarded when considering syntax
+            pass
+
+        elif isinstance(token, (p.StringToken, p.KeywordToken,
+                                p.IntegerToken, p.RationalToken, p.FloatToken)):
+            ## These are atoms we don't do much with
+            pass
+
+        elif isinstance(token, p.SetToken):
+            ## Set tokens have their own syntax object to allow for lexical sets
+            pass
+
+        elif isinstance(token, p.MappingToken):
+            ## As with sets, mappings have their own syntax object
+            pass
+
+        elif isinstance(token, p.ListToken):
+            ## This is the fun one because it's where most of the notation is implemented
+            pass
+
+    @classmethod
+    def read_symexpr(cls, token: p.SymbolToken):
+        """Emit a representation of using a binding."""
+
+    @classmethod
+    def read_setexpr(cls, token: p.SetToken):
+        """Emit a SetExpr """
+
+    @classmethod
+    def
+
+
+## Analysis interface
+def reads(buff: str,
+          reader: SexpReader = Reader,
+          parser: p.SexpParser = p.Parser,
+          source_name=None):
+    """Parse a single s-expression from a string, returning its token tree."""
+
+    return read(StringIO(buff), parser, source_name or f"<string {id(buff):x}>")
+
+
+def readf(path: str,
+          reader: SexpReader = Reader,
+          parser: p.SexpParser = p.Parser):
+    """Parse a single s-expression from the file named by a string, returning its token tree."""
+
+    with open(path, "r") as f:
+        return read(f, parser, path)
+
+
+def read(file: IO,
+         reader: SexpReader = Reader,
+         parser: p.SexpParser = p.Parser,
+         source_name=None):
+    """Parse a single sexpression from a file-like object, returning its token tree."""
+
+    return parser.parse(
+        PosTrackingBufferedReader(
+            file,
+            source_name=source_name
+        )
+    )