+} // namespace %(namespace)s
+
+'''
+
+class ISAParser(Grammar):
+ def __init__(self, output_dir, cpu_models):
+ super(ISAParser, self).__init__()
+ self.output_dir = output_dir
+
+ self.cpuModels = cpu_models
+
+ # variable to hold templates
+ self.templateMap = {}
+
+ # This dictionary maps format name strings to Format objects.
+ self.formatMap = {}
+
+ # The format stack.
+ self.formatStack = Stack(NoFormat())
+
+ # The default case stack.
+ self.defaultStack = Stack(None)
+
+ # Stack that tracks current file and line number. Each
+ # element is a tuple (filename, lineno) that records the
+ # *current* filename and the line number in the *previous*
+ # file where it was included.
+ self.fileNameStack = Stack()
+
+ symbols = ('makeList', 're', 'string')
+ self.exportContext = dict([(s, eval(s)) for s in symbols])
+
+ self.maxInstSrcRegs = 0
+ self.maxInstDestRegs = 0
+ self.maxMiscDestRegs = 0
+
+ #####################################################################
+ #
+ # Lexer
+ #
+ # The PLY lexer module takes two things as input:
+ # - A list of token names (the string list 'tokens')
+ # - A regular expression describing a match for each token. The
+ # regexp for token FOO can be provided in two ways:
+ # - as a string variable named t_FOO
+ # - as the doc string for a function named t_FOO. In this case,
+ # the function is also executed, allowing an action to be
+ # associated with each token match.
+ #
+ #####################################################################
+
+ # Reserved words. These are listed separately as they are matched
+ # using the same regexp as generic IDs, but distinguished in the
+ # t_ID() function. The PLY documentation suggests this approach.
+ reserved = (
+ 'BITFIELD', 'DECODE', 'DECODER', 'DEFAULT', 'DEF', 'EXEC', 'FORMAT',
+ 'HEADER', 'LET', 'NAMESPACE', 'OPERAND_TYPES', 'OPERANDS',
+ 'OUTPUT', 'SIGNED', 'TEMPLATE'
+ )
+
+ # List of tokens. The lex module requires this.
+ tokens = reserved + (
+ # identifier
+ 'ID',
+
+ # integer literal
+ 'INTLIT',
+
+ # string literal
+ 'STRLIT',
+
+ # code literal
+ 'CODELIT',
+
+ # ( ) [ ] { } < > , ; . : :: *
+ 'LPAREN', 'RPAREN',
+ 'LBRACKET', 'RBRACKET',
+ 'LBRACE', 'RBRACE',
+ 'LESS', 'GREATER', 'EQUALS',
+ 'COMMA', 'SEMI', 'DOT', 'COLON', 'DBLCOLON',
+ 'ASTERISK',
+
+ # C preprocessor directives
+ 'CPPDIRECTIVE'
+
+ # The following are matched but never returned. commented out to
+ # suppress PLY warning
+ # newfile directive
+ # 'NEWFILE',
+
+ # endfile directive
+ # 'ENDFILE'
+ )
+
+ # Regular expressions for token matching
+ t_LPAREN = r'\('
+ t_RPAREN = r'\)'
+ t_LBRACKET = r'\['
+ t_RBRACKET = r'\]'
+ t_LBRACE = r'\{'
+ t_RBRACE = r'\}'
+ t_LESS = r'\<'
+ t_GREATER = r'\>'
+ t_EQUALS = r'='
+ t_COMMA = r','
+ t_SEMI = r';'
+ t_DOT = r'\.'
+ t_COLON = r':'
+ t_DBLCOLON = r'::'
+ t_ASTERISK = r'\*'
+
+ # Identifiers and reserved words
+ reserved_map = { }
+ for r in reserved:
+ reserved_map[r.lower()] = r
+
+ def t_ID(self, t):
+ r'[A-Za-z_]\w*'
+ t.type = self.reserved_map.get(t.value, 'ID')
+ return t
+
+ # Integer literal
+ def t_INTLIT(self, t):
+ r'-?(0x[\da-fA-F]+)|\d+'
+ try:
+ t.value = int(t.value,0)
+ except ValueError:
+ error(t, 'Integer value "%s" too large' % t.value)
+ t.value = 0
+ return t
+
+ # String literal. Note that these use only single quotes, and
+ # can span multiple lines.
+ def t_STRLIT(self, t):
+ r"(?m)'([^'])+'"
+ # strip off quotes
+ t.value = t.value[1:-1]
+ t.lexer.lineno += t.value.count('\n')
+ return t
+
+
+ # "Code literal"... like a string literal, but delimiters are
+ # '{{' and '}}' so they get formatted nicely under emacs c-mode
+ def t_CODELIT(self, t):
+ r"(?m)\{\{([^\}]|}(?!\}))+\}\}"
+ # strip off {{ & }}
+ t.value = t.value[2:-2]
+ t.lexer.lineno += t.value.count('\n')
+ return t
+
+ def t_CPPDIRECTIVE(self, t):
+ r'^\#[^\#].*\n'
+ t.lexer.lineno += t.value.count('\n')
+ return t
+
+ def t_NEWFILE(self, t):
+ r'^\#\#newfile\s+"[^"]*"'
+ self.fileNameStack.push((t.value[11:-1], t.lexer.lineno))
+ t.lexer.lineno = 0
+
+ def t_ENDFILE(self, t):
+ r'^\#\#endfile'
+ (old_filename, t.lexer.lineno) = self.fileNameStack.pop()
+
+ #
+ # The functions t_NEWLINE, t_ignore, and t_error are
+ # special for the lex module.
+ #
+
+ # Newlines
+ def t_NEWLINE(self, t):
+ r'\n+'
+ t.lexer.lineno += t.value.count('\n')
+
+ # Comments
+ def t_comment(self, t):
+ r'//.*'
+
+ # Completely ignored characters
+ t_ignore = ' \t\x0c'
+
+ # Error handler
+ def t_error(self, t):
+ error(t, "illegal character '%s'" % t.value[0])
+ t.skip(1)
+
+ #####################################################################
+ #
+ # Parser
+ #
+ # Every function whose name starts with 'p_' defines a grammar
+ # rule. The rule is encoded in the function's doc string, while
+ # the function body provides the action taken when the rule is
+ # matched. The argument to each function is a list of the values
+ # of the rule's symbols: t[0] for the LHS, and t[1..n] for the
+ # symbols on the RHS. For tokens, the value is copied from the
+ # t.value attribute provided by the lexer. For non-terminals, the
+ # value is assigned by the producing rule; i.e., the job of the
+ # grammar rule function is to set the value for the non-terminal
+ # on the LHS (by assigning to t[0]).
+ #####################################################################
+
+ # The LHS of the first grammar rule is used as the start symbol
+ # (in this case, 'specification'). Note that this rule enforces
+ # that there will be exactly one namespace declaration, with 0 or
+ # more global defs/decls before and after it. The defs & decls
+ # before the namespace decl will be outside the namespace; those
+ # after will be inside. The decoder function is always inside the
+ # namespace.
+ def p_specification(self, t):
+ 'specification : opt_defs_and_outputs name_decl opt_defs_and_outputs decode_block'
+ global_code = t[1]
+ isa_name = t[2]
+ namespace = isa_name + "Inst"
+ # wrap the decode block as a function definition
+ t[4].wrap_decode_block('''
+StaticInstPtr
+%(isa_name)s::Decoder::decodeInst(%(isa_name)s::ExtMachInst machInst)
+{
+ using namespace %(namespace)s;
+''' % vars(), '}')
+ # both the latter output blocks and the decode block are in
+ # the namespace
+ namespace_code = t[3] + t[4]
+ # pass it all back to the caller of yacc.parse()
+ t[0] = (isa_name, namespace, global_code, namespace_code)
+
+ # ISA name declaration looks like "namespace <foo>;"
+ def p_name_decl(self, t):
+ 'name_decl : NAMESPACE ID SEMI'
+ t[0] = t[2]
+
+ # 'opt_defs_and_outputs' is a possibly empty sequence of
+ # def and/or output statements.
+ def p_opt_defs_and_outputs_0(self, t):
+ 'opt_defs_and_outputs : empty'
+ t[0] = GenCode(self)
+
+ def p_opt_defs_and_outputs_1(self, t):
+ 'opt_defs_and_outputs : defs_and_outputs'
+ t[0] = t[1]
+
+ def p_defs_and_outputs_0(self, t):
+ 'defs_and_outputs : def_or_output'
+ t[0] = t[1]
+
+ def p_defs_and_outputs_1(self, t):
+ 'defs_and_outputs : defs_and_outputs def_or_output'
+ t[0] = t[1] + t[2]
+
+ # The list of possible definition/output statements.
+ def p_def_or_output(self, t):
+ '''def_or_output : def_format
+ | def_bitfield
+ | def_bitfield_struct
+ | def_template
+ | def_operand_types
+ | def_operands
+ | output_header
+ | output_decoder
+ | output_exec
+ | global_let'''
+ t[0] = t[1]
+
+ # Output blocks 'output <foo> {{...}}' (C++ code blocks) are copied
+ # directly to the appropriate output section.
+
+ # Massage output block by substituting in template definitions and
+ # bit operators. We handle '%'s embedded in the string that don't
+ # indicate template substitutions (or CPU-specific symbols, which
+ # get handled in GenCode) by doubling them first so that the
+ # format operation will reduce them back to single '%'s.
+ def process_output(self, s):
+ s = self.protectNonSubstPercents(s)
+ # protects cpu-specific symbols too
+ s = self.protectCpuSymbols(s)
+ return substBitOps(s % self.templateMap)
+
+ def p_output_header(self, t):
+ 'output_header : OUTPUT HEADER CODELIT SEMI'
+ t[0] = GenCode(self, header_output = self.process_output(t[3]))
+
+ def p_output_decoder(self, t):
+ 'output_decoder : OUTPUT DECODER CODELIT SEMI'
+ t[0] = GenCode(self, decoder_output = self.process_output(t[3]))
+
+ def p_output_exec(self, t):
+ 'output_exec : OUTPUT EXEC CODELIT SEMI'
+ t[0] = GenCode(self, exec_output = self.process_output(t[3]))
+
+ # global let blocks 'let {{...}}' (Python code blocks) are
+ # executed directly when seen. Note that these execute in a
+ # special variable context 'exportContext' to prevent the code
+ # from polluting this script's namespace.
+ def p_global_let(self, t):
+ 'global_let : LET CODELIT SEMI'
+ self.updateExportContext()
+ self.exportContext["header_output"] = ''
+ self.exportContext["decoder_output"] = ''
+ self.exportContext["exec_output"] = ''
+ self.exportContext["decode_block"] = ''
+ try:
+ exec fixPythonIndentation(t[2]) in self.exportContext
+ except Exception, exc:
+ if debug:
+ raise
+ error(t, 'error: %s in global let block "%s".' % (exc, t[2]))
+ t[0] = GenCode(self,
+ header_output=self.exportContext["header_output"],
+ decoder_output=self.exportContext["decoder_output"],
+ exec_output=self.exportContext["exec_output"],
+ decode_block=self.exportContext["decode_block"])
+
+ # Define the mapping from operand type extensions to C++ types and
+ # bit widths (stored in operandTypeMap).
+ def p_def_operand_types(self, t):
+ 'def_operand_types : DEF OPERAND_TYPES CODELIT SEMI'
+ try:
+ self.operandTypeMap = eval('{' + t[3] + '}')
+ except Exception, exc:
+ if debug:
+ raise
+ error(t,
+ 'error: %s in def operand_types block "%s".' % (exc, t[3]))
+ t[0] = GenCode(self) # contributes nothing to the output C++ file
+
+ # Define the mapping from operand names to operand classes and
+ # other traits. Stored in operandNameMap.
+ def p_def_operands(self, t):
+ 'def_operands : DEF OPERANDS CODELIT SEMI'
+ if not hasattr(self, 'operandTypeMap'):
+ error(t, 'error: operand types must be defined before operands')
+ try:
+ user_dict = eval('{' + t[3] + '}', self.exportContext)
+ except Exception, exc:
+ if debug:
+ raise
+ error(t, 'error: %s in def operands block "%s".' % (exc, t[3]))
+ self.buildOperandNameMap(user_dict, t.lexer.lineno)
+ t[0] = GenCode(self) # contributes nothing to the output C++ file
+
+ # A bitfield definition looks like:
+ # 'def [signed] bitfield <ID> [<first>:<last>]'
+ # This generates a preprocessor macro in the output file.
+ def p_def_bitfield_0(self, t):
+ 'def_bitfield : DEF opt_signed BITFIELD ID LESS INTLIT COLON INTLIT GREATER SEMI'
+ expr = 'bits(machInst, %2d, %2d)' % (t[6], t[8])
+ if (t[2] == 'signed'):
+ expr = 'sext<%d>(%s)' % (t[6] - t[8] + 1, expr)
+ hash_define = '#undef %s\n#define %s\t%s\n' % (t[4], t[4], expr)
+ t[0] = GenCode(self, header_output=hash_define)
+
+ # alternate form for single bit: 'def [signed] bitfield <ID> [<bit>]'
+ def p_def_bitfield_1(self, t):
+ 'def_bitfield : DEF opt_signed BITFIELD ID LESS INTLIT GREATER SEMI'
+ expr = 'bits(machInst, %2d, %2d)' % (t[6], t[6])
+ if (t[2] == 'signed'):
+ expr = 'sext<%d>(%s)' % (1, expr)
+ hash_define = '#undef %s\n#define %s\t%s\n' % (t[4], t[4], expr)
+ t[0] = GenCode(self, header_output=hash_define)
+
+ # alternate form for structure member: 'def bitfield <ID> <ID>'
+ def p_def_bitfield_struct(self, t):
+ 'def_bitfield_struct : DEF opt_signed BITFIELD ID id_with_dot SEMI'
+ if (t[2] != ''):
+ error(t, 'error: structure bitfields are always unsigned.')
+ expr = 'machInst.%s' % t[5]
+ hash_define = '#undef %s\n#define %s\t%s\n' % (t[4], t[4], expr)
+ t[0] = GenCode(self, header_output=hash_define)
+
+ def p_id_with_dot_0(self, t):
+ 'id_with_dot : ID'
+ t[0] = t[1]
+
+ def p_id_with_dot_1(self, t):
+ 'id_with_dot : ID DOT id_with_dot'
+ t[0] = t[1] + t[2] + t[3]
+
+ def p_opt_signed_0(self, t):
+ 'opt_signed : SIGNED'
+ t[0] = t[1]
+
+ def p_opt_signed_1(self, t):
+ 'opt_signed : empty'
+ t[0] = ''
+
+ def p_def_template(self, t):
+ 'def_template : DEF TEMPLATE ID CODELIT SEMI'
+ self.templateMap[t[3]] = Template(self, t[4])
+ t[0] = GenCode(self)
+
+ # An instruction format definition looks like
+ # "def format <fmt>(<params>) {{...}};"
+ def p_def_format(self, t):
+ 'def_format : DEF FORMAT ID LPAREN param_list RPAREN CODELIT SEMI'
+ (id, params, code) = (t[3], t[5], t[7])
+ self.defFormat(id, params, code, t.lexer.lineno)
+ t[0] = GenCode(self)
+
+ # The formal parameter list for an instruction format is a
+ # possibly empty list of comma-separated parameters. Positional
+ # (standard, non-keyword) parameters must come first, followed by
+ # keyword parameters, followed by a '*foo' parameter that gets
+ # excess positional arguments (as in Python). Each of these three
+ # parameter categories is optional.
+ #
+ # Note that we do not support the '**foo' parameter for collecting
+ # otherwise undefined keyword args. Otherwise the parameter list
+ # is (I believe) identical to what is supported in Python.
+ #
+ # The param list generates a tuple, where the first element is a
+ # list of the positional params and the second element is a dict
+ # containing the keyword params.
+ def p_param_list_0(self, t):
+ 'param_list : positional_param_list COMMA nonpositional_param_list'
+ t[0] = t[1] + t[3]
+
+ def p_param_list_1(self, t):
+ '''param_list : positional_param_list
+ | nonpositional_param_list'''
+ t[0] = t[1]
+
+ def p_positional_param_list_0(self, t):
+ 'positional_param_list : empty'
+ t[0] = []
+
+ def p_positional_param_list_1(self, t):
+ 'positional_param_list : ID'
+ t[0] = [t[1]]
+
+ def p_positional_param_list_2(self, t):
+ 'positional_param_list : positional_param_list COMMA ID'
+ t[0] = t[1] + [t[3]]
+
+ def p_nonpositional_param_list_0(self, t):
+ 'nonpositional_param_list : keyword_param_list COMMA excess_args_param'
+ t[0] = t[1] + t[3]
+
+ def p_nonpositional_param_list_1(self, t):
+ '''nonpositional_param_list : keyword_param_list
+ | excess_args_param'''
+ t[0] = t[1]
+
+ def p_keyword_param_list_0(self, t):
+ 'keyword_param_list : keyword_param'
+ t[0] = [t[1]]
+
+ def p_keyword_param_list_1(self, t):
+ 'keyword_param_list : keyword_param_list COMMA keyword_param'
+ t[0] = t[1] + [t[3]]
+
+ def p_keyword_param(self, t):
+ 'keyword_param : ID EQUALS expr'
+ t[0] = t[1] + ' = ' + t[3].__repr__()
+
+ def p_excess_args_param(self, t):
+ 'excess_args_param : ASTERISK ID'
+ # Just concatenate them: '*ID'. Wrap in list to be consistent
+ # with positional_param_list and keyword_param_list.
+ t[0] = [t[1] + t[2]]
+
+ # End of format definition-related rules.
+ ##############
+
+ #
+ # A decode block looks like:
+ # decode <field1> [, <field2>]* [default <inst>] { ... }
+ #
+ def p_decode_block(self, t):
+ 'decode_block : DECODE ID opt_default LBRACE decode_stmt_list RBRACE'
+ default_defaults = self.defaultStack.pop()
+ codeObj = t[5]
+ # use the "default defaults" only if there was no explicit
+ # default statement in decode_stmt_list
+ if not codeObj.has_decode_default:
+ codeObj += default_defaults
+ codeObj.wrap_decode_block('switch (%s) {\n' % t[2], '}\n')
+ t[0] = codeObj
+
+ # The opt_default statement serves only to push the "default
+ # defaults" onto defaultStack. This value will be used by nested
+ # decode blocks, and used and popped off when the current
+ # decode_block is processed (in p_decode_block() above).
+ def p_opt_default_0(self, t):
+ 'opt_default : empty'
+ # no default specified: reuse the one currently at the top of
+ # the stack
+ self.defaultStack.push(self.defaultStack.top())
+ # no meaningful value returned
+ t[0] = None
+
+ def p_opt_default_1(self, t):
+ 'opt_default : DEFAULT inst'
+ # push the new default
+ codeObj = t[2]
+ codeObj.wrap_decode_block('\ndefault:\n', 'break;\n')
+ self.defaultStack.push(codeObj)
+ # no meaningful value returned
+ t[0] = None
+
+ def p_decode_stmt_list_0(self, t):
+ 'decode_stmt_list : decode_stmt'
+ t[0] = t[1]
+
+ def p_decode_stmt_list_1(self, t):
+ 'decode_stmt_list : decode_stmt decode_stmt_list'
+ if (t[1].has_decode_default and t[2].has_decode_default):
+ error(t, 'Two default cases in decode block')
+ t[0] = t[1] + t[2]
+
+ #
+ # Decode statement rules
+ #
+ # There are four types of statements allowed in a decode block:
+ # 1. Format blocks 'format <foo> { ... }'
+ # 2. Nested decode blocks
+ # 3. Instruction definitions.
+ # 4. C preprocessor directives.
+
+
+ # Preprocessor directives found in a decode statement list are
+ # passed through to the output, replicated to all of the output
+ # code streams. This works well for ifdefs, so we can ifdef out
+ # both the declarations and the decode cases generated by an
+ # instruction definition. Handling them as part of the grammar
+ # makes it easy to keep them in the right place with respect to
+ # the code generated by the other statements.
+ def p_decode_stmt_cpp(self, t):
+ 'decode_stmt : CPPDIRECTIVE'
+ t[0] = GenCode(self, t[1], t[1], t[1], t[1])
+
+ # A format block 'format <foo> { ... }' sets the default
+ # instruction format used to handle instruction definitions inside
+ # the block. This format can be overridden by using an explicit
+ # format on the instruction definition or with a nested format
+ # block.
+ def p_decode_stmt_format(self, t):
+ 'decode_stmt : FORMAT push_format_id LBRACE decode_stmt_list RBRACE'
+ # The format will be pushed on the stack when 'push_format_id'
+ # is processed (see below). Once the parser has recognized
+ # the full production (though the right brace), we're done
+ # with the format, so now we can pop it.
+ self.formatStack.pop()
+ t[0] = t[4]
+
+ # This rule exists so we can set the current format (& push the
+ # stack) when we recognize the format name part of the format
+ # block.
+ def p_push_format_id(self, t):
+ 'push_format_id : ID'
+ try:
+ self.formatStack.push(self.formatMap[t[1]])
+ t[0] = ('', '// format %s' % t[1])
+ except KeyError:
+ error(t, 'instruction format "%s" not defined.' % t[1])
+
+ # Nested decode block: if the value of the current field matches
+ # the specified constant, do a nested decode on some other field.
+ def p_decode_stmt_decode(self, t):
+ 'decode_stmt : case_label COLON decode_block'
+ label = t[1]
+ codeObj = t[3]
+ # just wrap the decoding code from the block as a case in the
+ # outer switch statement.
+ codeObj.wrap_decode_block('\n%s:\n' % label)
+ codeObj.has_decode_default = (label == 'default')
+ t[0] = codeObj
+
+ # Instruction definition (finally!).
+ def p_decode_stmt_inst(self, t):
+ 'decode_stmt : case_label COLON inst SEMI'
+ label = t[1]
+ codeObj = t[3]
+ codeObj.wrap_decode_block('\n%s:' % label, 'break;\n')
+ codeObj.has_decode_default = (label == 'default')
+ t[0] = codeObj
+
+ # The case label is either a list of one or more constants or
+ # 'default'
+ def p_case_label_0(self, t):
+ 'case_label : intlit_list'
+ def make_case(intlit):
+ if intlit >= 2**32:
+ return 'case ULL(%#x)' % intlit
+ else:
+ return 'case %#x' % intlit
+ t[0] = ': '.join(map(make_case, t[1]))
+
+ def p_case_label_1(self, t):
+ 'case_label : DEFAULT'
+ t[0] = 'default'
+
+ #
+ # The constant list for a decode case label must be non-empty, but
+ # may have one or more comma-separated integer literals in it.
+ #
+ def p_intlit_list_0(self, t):
+ 'intlit_list : INTLIT'
+ t[0] = [t[1]]
+
+ def p_intlit_list_1(self, t):
+ 'intlit_list : intlit_list COMMA INTLIT'
+ t[0] = t[1]
+ t[0].append(t[3])
+
+ # Define an instruction using the current instruction format
+ # (specified by an enclosing format block).
+ # "<mnemonic>(<args>)"
+ def p_inst_0(self, t):
+ 'inst : ID LPAREN arg_list RPAREN'
+ # Pass the ID and arg list to the current format class to deal with.
+ currentFormat = self.formatStack.top()
+ codeObj = currentFormat.defineInst(self, t[1], t[3], t.lexer.lineno)
+ args = ','.join(map(str, t[3]))
+ args = re.sub('(?m)^', '//', args)
+ args = re.sub('^//', '', args)
+ comment = '\n// %s::%s(%s)\n' % (currentFormat.id, t[1], args)
+ codeObj.prepend_all(comment)
+ t[0] = codeObj
+
+ # Define an instruction using an explicitly specified format:
+ # "<fmt>::<mnemonic>(<args>)"
+ def p_inst_1(self, t):
+ 'inst : ID DBLCOLON ID LPAREN arg_list RPAREN'
+ try:
+ format = self.formatMap[t[1]]
+ except KeyError:
+ error(t, 'instruction format "%s" not defined.' % t[1])
+
+ codeObj = format.defineInst(self, t[3], t[5], t.lexer.lineno)
+ comment = '\n// %s::%s(%s)\n' % (t[1], t[3], t[5])
+ codeObj.prepend_all(comment)
+ t[0] = codeObj
+
+ # The arg list generates a tuple, where the first element is a
+ # list of the positional args and the second element is a dict
+ # containing the keyword args.
+ def p_arg_list_0(self, t):
+ 'arg_list : positional_arg_list COMMA keyword_arg_list'
+ t[0] = ( t[1], t[3] )
+
+ def p_arg_list_1(self, t):
+ 'arg_list : positional_arg_list'
+ t[0] = ( t[1], {} )
+
+ def p_arg_list_2(self, t):
+ 'arg_list : keyword_arg_list'
+ t[0] = ( [], t[1] )
+
+ def p_positional_arg_list_0(self, t):
+ 'positional_arg_list : empty'
+ t[0] = []
+
+ def p_positional_arg_list_1(self, t):
+ 'positional_arg_list : expr'
+ t[0] = [t[1]]
+
+ def p_positional_arg_list_2(self, t):
+ 'positional_arg_list : positional_arg_list COMMA expr'
+ t[0] = t[1] + [t[3]]
+
+ def p_keyword_arg_list_0(self, t):
+ 'keyword_arg_list : keyword_arg'
+ t[0] = t[1]
+
+ def p_keyword_arg_list_1(self, t):
+ 'keyword_arg_list : keyword_arg_list COMMA keyword_arg'
+ t[0] = t[1]
+ t[0].update(t[3])
+
+ def p_keyword_arg(self, t):
+ 'keyword_arg : ID EQUALS expr'
+ t[0] = { t[1] : t[3] }
+
+ #
+ # Basic expressions. These constitute the argument values of
+ # "function calls" (i.e. instruction definitions in the decode
+ # block) and default values for formal parameters of format
+ # functions.
+ #
+ # Right now, these are either strings, integers, or (recursively)
+ # lists of exprs (using Python square-bracket list syntax). Note
+ # that bare identifiers are trated as string constants here (since
+ # there isn't really a variable namespace to refer to).
+ #
+ def p_expr_0(self, t):
+ '''expr : ID
+ | INTLIT
+ | STRLIT
+ | CODELIT'''
+ t[0] = t[1]
+
+ def p_expr_1(self, t):
+ '''expr : LBRACKET list_expr RBRACKET'''
+ t[0] = t[2]
+
+ def p_list_expr_0(self, t):
+ 'list_expr : expr'
+ t[0] = [t[1]]
+
+ def p_list_expr_1(self, t):
+ 'list_expr : list_expr COMMA expr'
+ t[0] = t[1] + [t[3]]
+
+ def p_list_expr_2(self, t):
+ 'list_expr : empty'
+ t[0] = []
+
+ #
+ # Empty production... use in other rules for readability.
+ #
+ def p_empty(self, t):
+ 'empty :'
+ pass
+
+ # Parse error handler. Note that the argument here is the
+ # offending *token*, not a grammar symbol (hence the need to use
+ # t.value)
+ def p_error(self, t):
+ if t:
+ error(t, "syntax error at '%s'" % t.value)