New Nim AST based backend - #defines working

nimterop/ast2.nim

@@ -0,0 +1,133 @@
+import macros, os, strutils, tables, times
+
+import compiler/[ast, idents, options, renderer]
+
+import "."/treesitter/api
+
+import "."/[compat, globals, getters]
+
+proc addConst(nimState: NimState, node: TSNode) =
+  echo "Const: " & nimState.getNodeVal(node)
+  if node.len() >= 2:
+    let
+      constDef = newNode(nkConstDef)
+      (name, info) = nimState.getNameInfo(node.getAtom(), nskConst)
+      ident = nimState.getIdent(name, info)
+      val = nimState.getNodeVal(node[1]).getLit()
+    if val.kind != nkNilLit:
+      constDef.add ident
+      constDef.add newNode(nkEmpty)
+      constDef.add val
+
+      nimState.constSection.add constDef
+
+proc addType(nimState: NimState, node: TSNode) =
+  echo "Type: " & nimState.getNodeVal(node)
+
+proc addEnum(nimState: NimState, node: TSNode) =
+  echo "Enum: " & nimState.getNodeVal(node)
+
+proc addProc(nimState: NimState, node: TSNode) =
+  echo "Proc: " & nimState.getNodeVal(node)
+
+proc processNode(nimState: NimState, node: TSNode): bool =
+  result = true
+
+  case node.getName()
+  of "preproc_def":
+    nimState.addConst(node)
+  of "type_definition":
+    if node.inTree("struct_specifier"):
+      nimState.addType(node)
+    elif node.inTree("enum_specifier"):
+      nimState.addEnum(node)
+    else:
+      # Unknown type
+      result = false
+  of "struct_specifier":
+    nimState.addType(node)
+  of "enum_specifier":
+    nimState.addEnum(node)
+  of "declaration":
+    nimState.addProc(node)
+  else:
+    # Unknown
+    result = false
+
+proc searchTree(nimState: NimState, root: TSNode) =
+  # Search AST generated by tree-sitter for recognized elements
+  var
+    node = root
+    nextnode: TSNode
+    depth = 0
+    processed = false
+
+  while true:
+    if not node.tsNodeIsNull() and depth > -1:
+      processed = nimState.processNode(node)
+    else:
+      break
+
+    if not processed and node.len() != 0:
+      nextnode = node[0]
+      depth += 1
+    else:
+      nextnode = node.tsNodeNextNamedSibling()
+
+    if nextnode.tsNodeIsNull():
+      while true:
+        node = node.tsNodeParent()
+        depth -= 1
+        if depth == -1:
+          break
+        if node == root:
+          break
+        if not node.tsNodeNextNamedSibling().tsNodeIsNull():
+          node = node.tsNodeNextNamedSibling()
+          break
+    else:
+      node = nextnode
+
+    if node == root:
+      break
+
+proc printNimHeader*() =
+  echo """# Generated at $1
+# Command line:
+#   $2 $3
+
+{.hint[ConvFromXtoItselfNotNeeded]: off.}
+
+import nimterop/types
+""" % [$now(), getAppFilename(), commandLineParams().join(" ")]
+
+proc printNim*(gState: State, fullpath: string, root: TSNode) =
+  var
+    nimState = new(NimState)
+    fp = fullpath.replace("\\", "/")
+
+  nimState.identifiers = newTable[string, string]()
+
+  nimState.gState = gState
+  nimState.currentHeader = getCurrentHeader(fullpath)
+  nimState.sourceFile = fullpath
+
+  # Nim compiler objects
+  nimState.identCache = newIdentCache()
+  nimState.config = newConfigRef()
+
+  nimState.constSection = newNode(nkConstSection)
+  nimState.enumSection = newNode(nkStmtList)
+  nimState.procSection = newNode(nkStmtList)
+  nimState.typeSection = newNode(nkTypeSection)
+
+  nimState.searchTree(root)
+
+  var
+    tree = newNode(nkStmtList)
+  tree.add nimState.enumSection
+  tree.add nimState.constSection
+  tree.add nimState.typeSection
+  tree.add nimState.procSection
+
+  echo tree.renderTree()

nimterop/getters.nim

@@ -2,6 +2,8 @@ import dynlib, macros, os, sequtils, sets, strformat, strutils, tables, times
 
 import regex
 
+import compiler/[ast, idents, lineinfos, msgs, pathutils]
+
 import "."/[build, compat, globals, plugin, treesitter/api]
 
 const gReserved = """
@@ -26,6 +28,8 @@ when while
 xor
 yield""".split(Whitespace).toHashSet()
 
+# Types related
+
 const gTypeMap* = {
   # char
   "char": "cchar",
@@ -91,6 +95,21 @@ proc getType*(str: string): string =
   if gTypeMap.hasKey(result):
     result = gTypeMap[result]
 
+proc getPtrType*(str: string): string =
+  result = case str:
+    of "ptr cchar":
+      "cstring"
+    of "ptr ptr cchar":
+      "ptr cstring"
+    of "ptr object":
+      "pointer"
+    of "ptr ptr object":
+      "ptr pointer"
+    else:
+      str
+
+# Identifier related
+
 proc checkIdentifier(name, kind, parent, origName: string) =
   let
     parentStr = if parent.nBl: parent & ":" else: ""
@@ -165,6 +184,8 @@ proc addNewIdentifer*(nimState: NimState, name: string, override = false): bool
       nimState.identifiers[nimName] = name
       result = true
 
+# Overrides related
+
 proc getOverride*(nimState: NimState, name: string, kind: NimSymKind): string =
   doAssert name.nBl, "Blank identifier error"
 
@@ -189,34 +210,96 @@ proc getOverrideFinal*(nimState: NimState, kind: NimSymKind): string =
     for i in nimState.gState.onSymbolOverrideFinal(typ):
       result &= "\n" & nimState.getOverride(i, kind)
 
-proc getPtrType*(str: string): string =
-  result = case str:
-    of "ptr cchar":
-      "cstring"
-    of "ptr ptr cchar":
-      "ptr cstring"
-    of "ptr object":
-      "pointer"
-    of "ptr ptr object":
-      "ptr pointer"
-    else:
-      str
-
-proc getLit*(str: string): string =
+proc getLit*(str: string): PNode =
   # Used to convert #define literals into const
   let
     str = str.replace(re"/[/*].*?(?:\*/)?$", "").strip()
 
-  if str.contains(re"^[\-]?[\d]*[.]?[\d]+$") or # decimal
-    str.contains(re"^0x[\da-fA-F]+$") or        # hexadecimal
-    str.contains(re"^'[[:ascii:]]'$") or        # char
-    str.contains(re"""^"[[:ascii:]]+"$"""):     # char *
-    return str
+  if str.contains(re"^[\-]?[\d]+$"):              # decimal
+    result = newIntNode(nkIntLit, parseInt(str))
+
+  elif str.contains(re"^[\-]?[\d]*[.]?[\d]+$"):   # float
+    result = newFloatNode(nkFloatLit, parseFloat(str))
+
+  elif str.contains(re"^0x[\da-fA-F]+$"):         # hexadecimal
+    result = newIntNode(nkIntLit, parseHexInt(str))
+
+  elif str.contains(re"^'[[:ascii:]]'$"):         # char
+    result = newNode(nkCharLit)
+    result.intVal = str[1].int64
+
+  elif str.contains(re"""^"[[:ascii:]]+"$"""):    # char *
+    result = newStrNode(nkStrLit, str[1 .. ^2])
+
+  else:
+    result = newNode(nkNilLit)
+
+# TSNode shortcuts
+
+proc len*(node: TSNode): uint =
+  if not node.tsNodeIsNull:
+    result = node.tsNodeNamedChildCount().uint
+
+proc `[]`*(node: TSNode, i: BiggestUInt): TSNode =
+  if i < node.len():
+    result = node.tsNodeNamedChild(i.uint32)
+
+proc getName*(node: TSNode): string {.inline.} =
+  if not node.tsNodeIsNull:
+    return $node.tsNodeType()
 
 proc getNodeVal*(nimState: NimState, node: TSNode): string =
-  return nimState.gState.code[node.tsNodeStartByte() .. node.tsNodeEndByte()-1].strip()
+  if not node.tsNodeIsNull:
+    return nimState.gState.code[node.tsNodeStartByte() .. node.tsNodeEndByte()-1].strip()
+
+proc getAtom*(node: TSNode): TSNode =
+  if not node.tsNodeIsNull:
+    # Get child node which is topmost atom
+    if node.getName() in gAtoms:
+      return node
+    elif node.len() != 0:
+      return node[0].getAtom()
+
+proc getPtrCount*(node: TSNode): string =
+  if not node.tsNodeIsNull:
+    # Get number of ptr nodes in tree
+    var
+      cnode = node
+    while "pointer_declarator" in cnode.getName():
+      result &= "ptr "
+      if cnode.len() != 0:
+        cnode = cnode[0]
+      else:
+        break
+
+proc getDeclarator*(node: TSNode): TSNode =
+  if not node.tsNodeIsNull:
+    # Return if child is a function or array declarator
+    if node.getName() in ["function_declarator", "array_declarator"]:
+      return node
+    elif node.len() != 0:
+      return node[0].getDeclarator()
+
+proc inTree*(node: TSNode, ntype: string): bool =
+  # Search for node type in tree - first children
+  result = false
+  var
+    cnode = node
+  while not cnode.tsNodeIsNull:
+    if cnode.getName() == ntype:
+      return true
+    cnode = cnode[0]
+
+proc inChildren*(node: TSNode, ntype: string): bool =
+  # Search for node type in immediate children
+  result = false
+  for i in 0 ..< node.len():
+    if (node[i]).getName() == ntype:
+      result = true
+      break
 
 proc getLineCol*(gState: State, node: TSNode): tuple[line, col: int] =
+  # Get line number and column info for node
   result.line = 1
   result.col = 1
   for i in 0 .. node.tsNodeStartByte().int-1:
@@ -225,6 +308,54 @@ proc getLineCol*(gState: State, node: TSNode): tuple[line, col: int] =
       result.line += 1
     result.col += 1
 
+proc getTSNodeNamedChildCountSansComments*(node: TSNode): int =
+  for i in 0 ..< node.len():
+    if node.getName() != "comment":
+      result += 1
+
+proc getPxName*(node: TSNode, offset: int): string =
+  # Get the xth (grand)parent of the node
+  var
+    np = node
+    count = 0
+
+  while not np.tsNodeIsNull() and count < offset:
+    np = np.tsNodeParent()
+    count += 1
+
+  if count == offset and not np.tsNodeIsNull():
+    return np.getName()
+
+# Compiler shortcuts
+
+proc getLineInfo*(nimState: NimState, node: TSNode): TLineInfo =
+  # Get Nim equivalent line:col info from node
+  let
+    (line, col) = nimState.gState.getLineCol(node)
+
+  result = newLineInfo(nimState.config, nimState.sourceFile.AbsoluteFile, line, col)
+
+proc getIdent*(nimState: NimState, name: string, info: TLineInfo, exported = true): PNode =
+  # Get ident PNode for name + info
+  let
+    exp = getIdent(nimState.identCache, "*")
+    ident = getIdent(nimState.identCache, name)
+
+  if exported:
+    result = newNode(nkPostfix)
+    result.add newIdentNode(exp, info)
+    result.add newIdentNode(ident, info)
+  else:
+    result = newIdentNode(ident, info)
+
+proc getNameInfo*(nimState: NimState, node: TSNode, kind: NimSymKind, parent = ""):
+  tuple[name: string, info: TLineInfo] =
+  # Shortcut to get identifier name and info (node value and line:col)
+  let
+    name = nimState.getNodeVal(node)
+  result.name = nimState.getIdentifier(name, kind, parent)
+  result.info = nimState.getLineInfo(node)
+
 proc getCurrentHeader*(fullpath: string): string =
   ("header" & fullpath.splitFile().name.multiReplace([(".", ""), ("-", "")]))
 
@@ -291,91 +422,6 @@ proc getPreprocessor*(gState: State, fullpath: string, mode = "cpp"): string =
     replace(re"__attribute__[ ]*\(\(.*?\)\)([ ,;])", "$1").
     removeStatic()
 
-converter toString*(kind: Kind): string =
-  return case kind:
-    of exactlyOne:
-      ""
-    of oneOrMore:
-      "+"
-    of zeroOrMore:
-      "*"
-    of zeroOrOne:
-      "?"
-    of orWithNext:
-      "!"
-
-converter toKind*(kind: string): Kind =
-  return case kind:
-    of "+":
-      oneOrMore
-    of "*":
-      zeroOrMore
-    of "?":
-      zeroOrOne
-    of "!":
-      orWithNext
-    else:
-      exactlyOne
-
-proc getNameKind*(name: string): tuple[name: string, kind: Kind, recursive: bool] =
-  if name[0] == '^':
-    result.recursive = true
-    result.name = name[1 .. ^1]
-  else:
-    result.name = name
-  result.kind = $name[^1]
-
-  if result.kind != exactlyOne:
-    result.name = result.name[0 .. ^2]
-
-proc getTSNodeNamedChildCountSansComments*(node: TSNode): int =
-  if node.tsNodeNamedChildCount() != 0:
-    for i in 0 .. node.tsNodeNamedChildCount()-1:
-      if $node.tsNodeType() != "comment":
-        result += 1
-
-proc getTSNodeNamedChildNames*(node: TSNode): seq[string] =
-  if node.tsNodeNamedChildCount() != 0:
-    for i in 0 .. node.tsNodeNamedChildCount()-1:
-      let
-        name = $node.tsNodeNamedChild(i).tsNodeType()
-
-      if name != "comment":
-        result.add(name)
-
-proc getRegexForAstChildren*(ast: ref Ast): string =
-  result = "^"
-  for i in 0 .. ast.children.len-1:
-    let
-      kind: string = ast.children[i].kind
-      begin = if result[^1] == '|': "" else: "(?:"
-    case kind:
-      of "!":
-        result &= &"{begin}{ast.children[i].name}|"
-      else:
-        result &= &"{begin}{ast.children[i].name}){kind}"
-  result &= "$"
-
-proc getAstChildByName*(ast: ref Ast, name: string): ref Ast =
-  for i in 0 .. ast.children.len-1:
-    if name in ast.children[i].name.split("|"):
-      return ast.children[i]
-
-  if ast.children.len == 1 and ast.children[0].name == ".":
-    return ast.children[0]
-
-proc getPxName*(node: TSNode, offset: int): string =
-  var
-    np = node
-    count = 0
-
-  while not np.tsNodeIsNull() and count < offset:
-    np = np.tsNodeParent()
-    count += 1
-
-  if count == offset and not np.tsNodeIsNull():
-    return $np.tsNodeType()
-
 proc getNimExpression*(nimState: NimState, expr: string): string =
   var
     clean = expr.multiReplace([("\n", " "), ("\r", "")])
@@ -435,42 +481,6 @@ proc getSplitComma*(joined: seq[string]): seq[string] =
   for i in joined:
     result = result.concat(i.split(","))
 
-proc getHeader*(nimState: NimState): string =
-  result =
-    if nimState.gState.dynlib.Bl:
-      &", header: {nimState.currentHeader}"
-    else:
-      ""
-
-proc getDynlib*(nimState: NimState): string =
-  result =
-    if nimState.gState.dynlib.nBl:
-      &", dynlib: {nimState.gState.dynlib}"
-    else:
-      ""
-
-proc getImportC*(nimState: NimState, origName, nimName: string): string =
-  if nimName != origName:
-    result = &"importc: \"{origName}\"{nimState.getHeader()}"
-  else:
-    result = nimState.impShort
-
-proc getPragma*(nimState: NimState, pragmas: varargs[string]): string =
-  result = ""
-  for pragma in pragmas.items():
-    if pragma.nBl:
-      result &= pragma & ", "
-  if result.nBl:
-    result = " {." & result[0 .. ^3] & ".}"
-
-  result = result.replace(nimState.impShort & ", cdecl", nimState.impShort & "C")
-
-  let
-    dy = nimState.getDynlib()
-
-  if ", cdecl" in result and dy.nBl:
-    result = result.replace(".}", dy & ".}")
-
 proc getComments*(nimState: NimState, strip = false): string =
   if not nimState.gState.nocomments and nimState.commentStr.nBl:
     result = "\n" & nimState.commentStr

nimterop/globals.nim

@@ -1,6 +1,6 @@
 import sequtils, sets, tables
 
-import regex
+import compiler/[ast, idents, options]
 
 import "."/plugin
 
@@ -34,24 +34,6 @@ const
   ].concat(toSeq(gExpressions.items))
 
 type
-  Kind = enum
-    exactlyOne
-    oneOrMore     # +
-    zeroOrMore    # *
-    zeroOrOne     # ?
-    orWithNext    # !
-
-  Ast = object
-    name*: string
-    kind*: Kind
-    recursive*: bool
-    children*: seq[ref Ast]
-    when not declared(CIMPORT):
-      tonim*: proc (ast: ref Ast, node: TSNode, nimState: NimState)
-    regex*: Regex
-
-  AstTable {.used.} = TableRef[string, seq[ref Ast]]
-
   State = ref object
     compile*, defines*, headers*, includeDirs*, searchDirs*, prefix*, suffix*, symOverride*: seq[string]
 
@@ -67,7 +49,12 @@ type
   NimState {.used.} = ref object
     identifiers*: TableRef[string, string]
 
-    commentStr*, constStr*, debugStr*, enumStr*, procStr*, skipStr*, typeStr*: string
+    commentStr*, debugStr*, skipStr*: string
+
+    # Nim compiler objects
+    constSection*, enumSection*, procSection*, typeSection*: PNode
+    identCache*: IdentCache
+    config*: ConfigRef
 
     gState*: State
 
@@ -94,7 +81,7 @@ type CompileMode = enum
 const modeDefault {.used.} = $cpp # TODO: USE this everywhere relevant
 
 when not declared(CIMPORT):
-  export gAtoms, gExpressions, gEnumVals, Kind, Ast, AstTable, State, NimState,
+  export gAtoms, gExpressions, gEnumVals, State, NimState,
     nBl, Bl, CompileMode, modeDefault
 
   # Redirect output to file when required

nimterop/nim.cfg

@@ -0,0 +1 @@
+--path:"$nim"

nimterop/toast.nim

@@ -2,7 +2,7 @@ import os, osproc, strformat, strutils, times
 
 import "."/treesitter/[api, c, cpp]
 
-import "."/[ast, compat, globals, getters, grammar]
+import "."/[ast2, compat, globals, getters]
 
 proc printLisp(gState: State, root: TSNode) =
   var
@@ -53,7 +53,7 @@ proc printLisp(gState: State, root: TSNode) =
     if node == root:
       break
 
-proc process(gState: State, path: string, astTable: AstTable) =
+proc process(gState: State, path: string) =
   doAssert existsFile(path), &"Invalid path {path}"
 
   var
@@ -93,7 +93,7 @@ proc process(gState: State, path: string, astTable: AstTable) =
   if gState.past:
     gState.printLisp(root)
   elif gState.pnim:
-    gState.printNim(path, root, astTable)
+    gState.printNim(path, root)
   elif gState.preprocess:
     gecho gState.code
 
@@ -109,7 +109,6 @@ proc main(
     nocomments = false,
     output = "",
     past = false,
-    pgrammar = false,
     pluginSourcePath: string = "",
     pnim = false,
     prefix: seq[string] = @[],
@@ -167,19 +166,12 @@ proc main(
     doAssert gState.outputHandle.open(outputFile, fmWrite),
       &"Failed to write to {outputFile}"
 
-  # Process grammar into AST
-  let
-    astTable = parseGrammar()
-
-  if pgrammar:
-    # Print AST of grammar
-    gState.printGrammar(astTable)
-  elif source.nBl:
+  if source.nBl:
     # Print source after preprocess or Nim output
     if gState.pnim:
       gState.printNimHeader()
     for src in source:
-      gState.process(src.expandSymlinkAbs(), astTable)
+      gState.process(src.expandSymlinkAbs())
 
   # Close outputFile
   if outputFile.len != 0:
@@ -241,7 +233,6 @@ when isMainModule:
     "nocomments": "exclude top-level comments from output",
     "output": "file to output content - default stdout",
     "past": "print AST output",
-    "pgrammar": "print grammar",
     "pluginSourcePath": "Nim file to build and load as a plugin",
     "pnim": "print Nim output",
     "preprocess": "run preprocessor on header",
@@ -260,7 +251,6 @@ when isMainModule:
     "nocomments": 'c',
     "output": 'o',
     "past": 'a',
-    "pgrammar": 'g',
     "pnim": 'n',
     "prefix": 'E',
     "preprocess": 'p',