Source code for Parser.RulesCodeDOM

# EMACS settings: -*-	tab-width: 2; indent-tabs-mode: t; python-indent-offset: 2 -*-
# vim: tabstop=2:shiftwidth=2:noexpandtab
# kate: tab-width 2; replace-tabs off; indent-width 2;
#
# ==============================================================================
# Authors:          Patrick Lehmann
#
# Python Module:    TODO
#
# License:
# ==============================================================================
# Copyright 2007-2016 Technische Universitaet Dresden - Germany
#                     Chair of VLSI-Design, Diagnostics and Architecture
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
#
# load dependencies
from lib.Parser     import MismatchingParserResult, MatchingParserResult, EmptyChoiseParserResult, StartOfDocumentToken
from lib.Parser     import SpaceToken, CharacterToken, StringToken
from lib.CodeDOM    import EmptyLine, CommentLine, BlockedStatement as BlockStatementBase, StringLiteral
from lib.CodeDOM    import Statement, BlockStatement


__api__ = [
	'InFileStatements',
	'PreProcessStatements',
	'PostProcessStatements',
	'DocumentStatements',
	'CopyStatement',
	'DeleteStatement',
	'ReplaceStatement',
	'AppendLineStatement',
	'FileStatement',
	'ProcessRulesBlockStatement',
	'PreProcessRulesStatement',
	'PostProcessRulesStatement',
	'Document'
]
__all__ = __api__

# ==============================================================================
# Blocked Statements (Forward declaration)
# ==============================================================================
[docs]class InFileStatements(BlockStatementBase): _allowedStatements = []
[docs]class PreProcessStatements(BlockStatementBase): _allowedStatements = []
[docs]class PostProcessStatements(BlockStatementBase): _allowedStatements = []
[docs]class DocumentStatements(BlockStatementBase): _allowedStatements = []
# ============================================================================== # File Reference Statements # ==============================================================================
[docs]class CopyStatement(Statement): def __init__(self, source, destination, commentText): super().__init__() self._sourcePath = source self._destinationPath = destination self._commentText = commentText @property def SourcePath(self): return self._sourcePath @property def DestinationPath(self): return self._destinationPath @classmethod # mccabe:disable=MC0001
[docs] def GetParser(cls): # match for optional whitespacex token = yield if isinstance(token, SpaceToken): token = yield # match for COPY keyword if (not isinstance(token, StringToken)): raise MismatchingParserResult("CopyParser: Expected COPY keyword.") if (token.Value.lower() != "copy"): raise MismatchingParserResult("CopyParser: Expected COPY keyword.") # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("CopyParser: Expected whitespace before source filename.") # match for string: sourceFile; use a StringLiteralParser to parse the pattern parser = StringLiteral.GetParser() parser.send(None) sourceFile = None try: while True: token = yield parser.send(token) except MatchingParserResult as ex: sourceFile = ex.value.Value # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("CopyParser: Expected whitespace before TO keyword.") # match for TO keyword token = yield if (not isinstance(token, StringToken)): raise MismatchingParserResult("CopyParser: Expected TO keyword.") if (token.Value.lower() != "to"): raise MismatchingParserResult("CopyParser: Expected TO keyword.") # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("CopyParser: Expected whitespace before destination directory.") # match for string: destinationDirectory; use a StringLiteralParser to parse the pattern parser = StringLiteral.GetParser() parser.send(None) destinationDirectory = None try: while True: token = yield parser.send(token) except MatchingParserResult as ex: destinationDirectory = ex.value.Value # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for delimiter sign: \n commentText = "" if (not isinstance(token, CharacterToken)): raise MismatchingParserResult("CopyParser: Expected end of line or comment") if (token.Value == "\n"): pass elif (token.Value == "#"): # match for any until line end while True: token = yield if (isinstance(token, CharacterToken) and (token.Value == "\n")): break commentText += token.Value else: raise MismatchingParserResult("CopyParser: Expected end of line or comment") # construct result result = cls(sourceFile, destinationDirectory, commentText) raise MatchingParserResult(result)
def __str__(self, indent=0): if (self._commentText != ""): return "{0}Copy \"{1!s}\" To \"{2!s}\" # {3}".format((" " * indent), self._sourcePath, self._destinationPath, self._commentText) else: return "{0}Copy \"{1!s}\" To \"{2!s}\"".format((" " * indent), self._sourcePath, self._destinationPath)
[docs]class DeleteStatement(Statement): def __init__(self, file, commentText): super().__init__() self._filePath = file self._commentText = commentText @property def FilePath(self): return self._filePath @classmethod
[docs] def GetParser(cls): # match for optional whitespacex token = yield if isinstance(token, SpaceToken): token = yield # match for DELETE keyword if (not isinstance(token, StringToken)): raise MismatchingParserResult("DeleteParser: Expected DELETE keyword.") if (token.Value.lower() != "delete"): raise MismatchingParserResult("DeleteParser: Expected DELETE keyword.") # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("DeleteParser: Expected whitespace before filename.") # match for string: file; use a StringLiteralParser to parse the pattern parser = StringLiteral.GetParser() parser.send(None) file = None try: while True: token = yield parser.send(token) except MatchingParserResult as ex: file = ex.value.Value # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for delimiter sign: \n commentText = "" if (not isinstance(token, CharacterToken)): raise MismatchingParserResult("DeleteParser: Expected end of line or comment") if (token.Value == "\n"): pass elif (token.Value == "#"): # match for any until line end while True: token = yield if (isinstance(token, CharacterToken) and (token.Value == "\n")): break commentText += token.Value else: raise MismatchingParserResult("DeleteParser: Expected end of line or comment") # construct result result = cls(file, commentText) raise MatchingParserResult(result)
def __str__(self, indent=0): if (self._commentText != ""): return "{0}Delete \"{1!s}\" # {2}".format((" " * indent), self._filePath, self._commentText) else: return "{0}Delete \"{1!s}\"".format((" " * indent), self._filePath)
[docs]class ReplaceStatement(Statement): def __init__(self, searchPattern, replacePattern, caseInsensitive, multiLine, dotAll, commentText): super().__init__() self._searchPattern = searchPattern self._replacePattern = replacePattern self._caseInsensitive = caseInsensitive self._multiLine = multiLine self._dotAll = dotAll self._commentText = commentText @property def SearchPattern(self): return self._searchPattern @property def ReplacePattern(self): return self._replacePattern @property def CaseInsensitive(self): return self._caseInsensitive @property def MultiLine(self): return self._multiLine @property def DotAll(self): return self._dotAll @classmethod # mccabe:disable=MC0001
[docs] def GetParser(cls): multiLine = False dotAll = False caseInsensitive = False # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for keyword: REPLACE if (not isinstance(token, StringToken)): raise MismatchingParserResult("ReplaceParser: Expected REPLACE keyword.") if (token.Value.lower() != "replace"): raise MismatchingParserResult("ReplaceParser: Expected REPLACE keyword.") # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("ReplaceParser: Expected whitespace before search pattern.") # match for string: searchPattern; use a StringLiteralParser to parse the pattern parser = StringLiteral.GetParser() parser.send(None) searchPattern = None try: while True: token = yield parser.send(token) except MatchingParserResult as ex: searchPattern = ex.value.Value # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("ReplaceParser: Expected whitespace before WITH keyword.") # match for WITH keyword token = yield if (not isinstance(token, StringToken)): raise MismatchingParserResult("ReplaceParser: Expected WITH keyword.") if (token.Value.lower() != "with"): raise MismatchingParserResult("ReplaceParser: Expected WITH keyword.") # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("ReplaceParser: Expected whitespace before replace pattern.") # match for string: replacePattern; use a StringLiteralParser to parse the pattern parser = StringLiteral.GetParser() parser.send(None) replacePattern = None try: while True: token = yield parser.send(token) except MatchingParserResult as ex: replacePattern = ex.value.Value # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for line end, comment or OPTIONS keyword if isinstance(token, StringToken): if (token.Value.lower() == "options"): # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("ReplaceParser: Expected whitespace before MULTILINE, DOTALL or CASEINSENSITIVE keyword.") token = yield # first keyword for _ in range(3): # match for MULTILINE, DOTALL or CASEINSENSITIVE keyword if (not isinstance(token, StringToken)): raise MismatchingParserResult("ReplaceParser: Expected MULTILINE, DOTALL or CASEINSENSITIVE keyword.") if (token.Value.lower() == "multiline"): multiLine = True elif (token.Value.lower() == "dotall"): dotAll = True elif (token.Value.lower() == "caseinsensitive"): caseInsensitive = True else: raise MismatchingParserResult("ReplaceParser: Expected MULTILINE, DOTALL or CASEINSENSITIVE keyword.") # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield if (not isinstance(token, CharacterToken)): raise MismatchingParserResult("ReplaceParser: Expected more options, end of line or comment.") if (token.Value == ","): # match for optional whitespace, before going into the next iteration token = yield if isinstance(token, SpaceToken): token = yield continue # with 'token' set to next keyword else: break # match for delimiter sign: \n or # commentText = "" if (not isinstance(token, CharacterToken)): raise MismatchingParserResult("ReplaceParser: Expected end of line or comment") if (token.Value == "\n"): pass elif (token.Value == "#"): # match for any until line end while True: token = yield if (isinstance(token, CharacterToken) and (token.Value == "\n")): break commentText += token.Value else: raise MismatchingParserResult("ReplaceParser: Expected end of line or comment") # construct result result = cls(searchPattern, replacePattern, caseInsensitive, multiLine, dotAll, commentText) raise MatchingParserResult(result)
def __str__(self, indent=0): return "{0}Replace {1} by {2}".format(" " * indent, self._searchPattern, self._replacePattern)
[docs]class AppendLineStatement(Statement): def __init__(self, appendPattern, commentText): super().__init__() self._appendPattern = appendPattern self._commentText = commentText @property def AppendPattern(self): return self._appendPattern @classmethod
[docs] def GetParser(cls): # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for keyword: APPENDLINE if (not isinstance(token, StringToken)): raise MismatchingParserResult("AppendLineParser: Expected APPENDLINE keyword.") if (token.Value.lower() != "appendline"): raise MismatchingParserResult("AppendLineParser: Expected APPENDLINE keyword.") # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("AppendLineParser: Expected whitespace before append pattern.") # match for string: appendPattern; use a StringLiteralParser to parse the pattern parser = StringLiteral.GetParser() parser.send(None) appendPattern = None try: while True: token = yield parser.send(token) except MatchingParserResult as ex: appendPattern = ex.value.Value # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for delimiter sign: \n commentText = "" if (not isinstance(token, CharacterToken)): raise MismatchingParserResult("AppendLineParser: Expected end of line or comment") if (token.Value == "\n"): pass elif (token.Value == "#"): # match for any until line end while True: token = yield if (isinstance(token, CharacterToken) and (token.Value == "\n")): break commentText += token.Value else: raise MismatchingParserResult("AppendLineParser: Expected end of line or comment") # construct result result = cls(appendPattern, commentText) raise MatchingParserResult(result)
def __str__(self, indent=0): return "{0}AppendLine {1}".format(" " * indent, self._appendPattern)
# ============================================================================== # Block Statements # ==============================================================================
[docs]class FileStatement(BlockStatement): def __init__(self, file, commentText): super().__init__() self._filePath = file self._commentText = commentText @property def FilePath(self): return self._filePath @classmethod # mccabe:disable=MC0001
[docs] def GetParser(cls): # match for IN ... FILE clause # ========================================================================== # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for keyword: FILE if (not isinstance(token, StringToken)): raise MismatchingParserResult("FileParser: Expected FILE keyword.") if (token.Value.lower() != "file"): raise MismatchingParserResult("FileParser: Expected FILE keyword.") # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("FileParser: Expected whitespace before filename.") # match for string: replaceFilename; use a StringLiteralParser to parse the pattern parser = StringLiteral.GetParser() parser.send(None) replaceFilename = None try: while True: token = yield parser.send(token) except MatchingParserResult as ex: replaceFilename = ex.value.Value # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for delimiter sign: \n commentText = "" if (not isinstance(token, CharacterToken)): raise MismatchingParserResult("FileParser: Expected end of line or comment") if (token.Value == "\n"): pass elif (token.Value == "#"): # match for any until line end while True: token = yield if (isinstance(token, CharacterToken) and (token.Value == "\n")): break commentText += token.Value else: raise MismatchingParserResult("FileParser: Expected end of line or comment") # match for inner statements # ========================================================================== # construct result result = cls(replaceFilename, commentText) parser = cls.GetRepeatParser(result.AddStatement, InFileStatements.GetParser) parser.send(None) try: while True: token = yield parser.send(token) except EmptyChoiseParserResult: print("ERROR in *.rules file -> fix me") except MatchingParserResult: pass # match for END FILE clause # ========================================================================== # match for optional whitespace if isinstance(token, SpaceToken): token = yield # match for keyword: END if (not isinstance(token, StringToken)): raise MismatchingParserResult("FileParser: Expected END keyword.") if (token.Value.lower() != "end"): raise MismatchingParserResult("FileParser: Expected END keyword.") # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult("FileParser: Expected whitespace before FILE keyword.") # match for keyword: FILE token = yield if (not isinstance(token, StringToken)): raise MismatchingParserResult("FileParser: Expected FILE keyword.") if (token.Value.lower() != "file"): raise MismatchingParserResult("FileParser: Expected FILE keyword.") # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for delimiter sign: \n if (not isinstance(token, CharacterToken)): raise MismatchingParserResult("FileParser: Expected end of line or comment") if (token.Value == "\n"): pass elif (token.Value == "#"): # match for any until line end while True: token = yield if (isinstance(token, CharacterToken) and (token.Value == "\n")): break else: raise MismatchingParserResult("FileParser: Expected end of line or comment") result.CommentText = commentText raise MatchingParserResult(result)
def __str__(self, indent=0): _indent = " " * indent buffer = _indent + "FileParser" for stmt in self._statements: buffer += "\n{0}{1}".format(_indent, stmt.__str__(indent + 1)) return buffer
[docs]class ProcessRulesBlockStatement(BlockStatement): __PARSER_NAME__ = None __PARSER_BLOCK_NAME__ = None __PARSER_STATEMENTS__ = None def __init__(self, commentText): super().__init__() self._commentText = commentText @classmethod # mccabe:disable=MC0001
[docs] def GetParser(cls): # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for keyword: __PARSER_BLOCK_NAME__ if (not isinstance(token, StringToken)): raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected " + cls.__PARSER_BLOCK_NAME__ + " keyword.") if (token.Value.lower() != cls.__PARSER_BLOCK_NAME__): raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected " + cls.__PARSER_BLOCK_NAME__ + " keyword.") # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for delimiter sign: \n commentText = "" if (not isinstance(token, CharacterToken)): raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected end of line or comment") if (token.Value == "\n"): pass elif (token.Value == "#"): # match for any until line end while True: token = yield if (isinstance(token, CharacterToken) and (token.Value == "\n")): break commentText += token.Value else: raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected end of line or comment") # match for inner statements # ========================================================================== # construct result result = cls(commentText) parser = cls.GetRepeatParser(result.AddStatement, cls.__PARSER_STATEMENTS__.GetParser) parser.send(None) try: while True: token = yield parser.send(token) except MatchingParserResult: pass # match for END __PARSER_BLOCK_NAME__ # ========================================================================== # match for optional whitespace if isinstance(token, SpaceToken): token = yield # match for keyword: END if (not isinstance(token, StringToken)): raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected END keyword.") if (token.Value.lower() != "end"): raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected END keyword.") # match for whitespace token = yield if (not isinstance(token, SpaceToken)): raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected whitespace before " + cls.__PARSER_BLOCK_NAME__ + " keyword.") # match for keyword: __PARSER_BLOCK_NAME__ token = yield if (not isinstance(token, StringToken)): raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected " + cls.__PARSER_BLOCK_NAME__ + " keyword.") if (token.Value.lower() != cls.__PARSER_BLOCK_NAME__): raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected " + cls.__PARSER_BLOCK_NAME__ + " keyword.") # match for optional whitespace token = yield if isinstance(token, SpaceToken): token = yield # match for delimiter sign: \n # commentText = "" if (not isinstance(token, CharacterToken)): raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected end of line or comment") if (token.Value == "\n"): pass elif (token.Value == "#"): # match for any until line end while True: token = yield if (isinstance(token, CharacterToken) and (token.Value == "\n")): break # commentText += token.Value else: raise MismatchingParserResult(cls.__PARSER_NAME__ + ": Expected end of line or comment") raise MatchingParserResult(result)
def __str__(self, indent=0): _indent = " " * indent buffer = _indent + self.__PARSER_NAME__ for stmt in self._statements: buffer += "\n{0}{1}".format(_indent, stmt.__str__(indent + 1)) return buffer
[docs]class PreProcessRulesStatement(ProcessRulesBlockStatement): __PARSER_NAME__ = "PreProcessRulesParser" __PARSER_BLOCK_NAME__ = "preprocessrules" __PARSER_STATEMENTS__ = PreProcessStatements
[docs]class PostProcessRulesStatement(ProcessRulesBlockStatement): __PARSER_NAME__ = "PostProcessRulesParser" __PARSER_BLOCK_NAME__ = "postprocessrules" __PARSER_STATEMENTS__ = PostProcessStatements
[docs]class Document(BlockStatement): @classmethod
[docs] def GetParser(cls): result = cls() parser = cls.GetRepeatParser(result.AddStatement, DocumentStatements.GetParser) parser.send(None) token = yield if (not isinstance(token, StartOfDocumentToken)): raise MismatchingParserResult("Expected a StartOfDocumentToken, got {0!s}.".format(token)) try: while True: token = yield parser.send(token) except MatchingParserResult: raise MatchingParserResult(result)
def __str__(self, indent=0): buffer = " " * indent + "Document" for stmt in self._statements: buffer += "\n{0}".format(stmt.__str__(indent + 1)) return buffer
InFileStatements.AddChoice(ReplaceStatement) InFileStatements.AddChoice(AppendLineStatement) InFileStatements.AddChoice(CommentLine) InFileStatements.AddChoice(EmptyLine) PreProcessStatements.AddChoice(CopyStatement) PreProcessStatements.AddChoice(FileStatement) PreProcessStatements.AddChoice(CommentLine) PreProcessStatements.AddChoice(EmptyLine) PostProcessStatements.AddChoice(CopyStatement) PostProcessStatements.AddChoice(DeleteStatement) PostProcessStatements.AddChoice(FileStatement) PostProcessStatements.AddChoice(CommentLine) PostProcessStatements.AddChoice(EmptyLine) DocumentStatements.AddChoice(PreProcessRulesStatement) DocumentStatements.AddChoice(PostProcessRulesStatement) DocumentStatements.AddChoice(CommentLine) DocumentStatements.AddChoice(EmptyLine)