def split_rule(rules, rule_name, symbols_to_extract: List[str], subrule_name: str):
"""
Let only options which are starting with symbols from symbols_to_extract.
Put the rest to a subrule.
"""
r = rule_by_name(rules, rule_name)
assert isinstance(r.body, Antlr4Selection), r
sub_options = Antlr4Selection([])
for o in r.body:
start_symbols = set()
_direct_left_corner(o, start_symbols, allow_eps_in_sel=True)
if not start_symbols.intersection(symbols_to_extract):
sub_options.append(o)
r.body = Antlr4Selection([o for o in r.body if not (o in sub_options)])
r.body.insert(0, Antlr4Symbol(subrule_name, False))
if len(r.body) == 1:
r.body = r.body[0]
assert len(sub_options) > 0
if len(sub_options) == 1:
sub_options = sub_options[0]
else:
sub_options = Antlr4Selection(sub_options)
sub_r = Antlr4Rule(subrule_name, sub_options)
rules.insert(rules.index(r), sub_r)
return sub_r
def extract_bin_ops(rules, current_expr_rule, ops_to_extrat, new_rule_name,
top_rule_name, handle_conditional_fn, handle_inside_fn):
# find option with binary op rule
# expr = rule_by_name(rules, "expression")
ops_no_special = [
o for o in ops_to_extrat if o not in [
"KW_INSIDE",
"KW_DIST",
"QUESTIONMARK",
]
]
bin_op_choices = []
if len(ops_no_special) > 0:
if len(ops_no_special) == 1:
op = Antlr4Symbol(ops_no_special[0], False)
else:
op = Antlr4Selection(
[Antlr4Symbol(o, False) for o in ops_no_special])
# expression (binary_operator ( attribute_instance )* expression)*
bin_op_choice = Antlr4Sequence([
Antlr4Symbol(current_expr_rule.name, False),
Antlr4Iteration(
Antlr4Sequence([
op,
Antlr4Iteration(Antlr4Symbol("attribute_instance", False)),
Antlr4Symbol(top_rule_name, False)
]))
])
bin_op_choices.append(bin_op_choice)
if "KW_INSIDE" in ops_to_extrat:
handle_inside_fn(bin_op_choices, current_expr_rule)
if "KW_DIST" in ops_to_extrat:
# handled differently, only allowed on specified places
pass
if "QUESTIONMARK" in ops_to_extrat:
handle_conditional_fn(bin_op_choices, current_expr_rule)
# create a new rule which contains rule for extracted binary operators
if len(bin_op_choices) > 1:
new_body = Antlr4Selection(bin_op_choices)
else:
new_body = bin_op_choices[0]
new_r = Antlr4Rule(new_rule_name, new_body)
rules.insert(rules.index(current_expr_rule), new_r)
return new_r
def fix_priority_of__class_scope__package_scope(rules):
orig = Antlr4Selection([
Antlr4Symbol("class_scope", False),
Antlr4Symbol("package_scope", False)
])
repl = Antlr4Selection([
Antlr4Symbol("package_scope", False),
Antlr4Symbol("class_scope", False)
])
def match_replace_fn(o):
if o == orig:
return deepcopy(repl)
for rule in rules:
replace_item_by_sequence(rule, match_replace_fn)
def parse_element_selection(
self, ctx: Element_selectionContext) -> Antlr4Selection:
"""
element_selection: element_sequence ('|' element_sequence)+;
"""
body = []
for c in ctx.element_sequence():
body.append(self.parse_element_sequence(c))
return Antlr4Selection(body)
def extract_option_as_rule(rules, rule_name, options_i, new_rule_name):
r = rule_by_name(rules, rule_name)
assert isinstance(r.body, Antlr4Selection)
new_body = Antlr4Selection([])
for i in options_i:
new_body.append(r.body[i])
r.body[options_i[0]] = Antlr4Sequence(
[Antlr4Symbol(new_rule_name, False),
Antlr4Newline(),
Antlr4Indent(1)])
r.body = Antlr4Selection(
[x for i, x in enumerate(r.body) if i not in options_i[1:]])
if len(new_body) == 1:
new_body = new_body[0]
new_r = Antlr4Rule(new_rule_name, new_body)
rules.insert(rules.index(r), new_r)
return new_r
def direct_left_recurse_rm(rules, rule_name):
r = rule_by_name(rules, rule_name)
if isinstance(r.body, Antlr4Selection):
choices = r.body
elif isinstance(r.body, Antlr4Sequence):
choices = [r.body, ]
else:
raise NotImplementedError()
# find choices which starts with this rule non terminal
lr_choices = []
for c in choices:
if isinstance(c, Antlr4Sequence):
first = next(iter_non_visuals(c))
if isinstance(first, Antlr4Symbol) and first.symbol == rule_name:
lr_choices.append(c)
else:
raise NotImplementedError()
# remove choices which are causing left recursion
assert len(lr_choices) >= 1, rule_name
for lr_choice in lr_choices:
choices.remove(lr_choice)
if len(choices) == 0:
raise NotImplementedError()
elif len(choices) == 1:
r.body = choices[0]
# renaame this rule to rule_item
r_base_name = r.name + "_item"
for _r in rules:
assert r.name != r_base_name, r_base_name
r.name = r_base_name
# create new rule which will implement removed choices and also expands to rule_item
choices_new = Antlr4Selection([])
for lr_choice in lr_choices:
first = next(iter_non_visuals(lr_choice))
assert isinstance(first, Antlr4Symbol) and first.symbol == rule_name
repl = Antlr4Symbol(r_base_name, False)
_iterate_everything_except_first_and_replace_first(lr_choice, repl)
if not choices_new:
lr_choice.insert(0, Antlr4Newline())
lr_choice.insert(1, Antlr4Indent(1))
choices_new.append(lr_choice)
body_new = choices_new[0] if len(choices_new) == 1 else choices_new
r_new = Antlr4Rule(rule_name, body_new)
rules.insert(rules.index(r), r_new)
def add_string_literal_rules(p):
string_char = Antlr4Rule("ANY_ASCII_CHARACTERS", Antlr4Selection([
Antlr4Symbol('~["\\\\\\r\\n]', True, True),
Antlr4Symbol('\\\n', True),
Antlr4Symbol('\\\r\n', True),
Antlr4Symbol('\t', True),
Antlr4Symbol('\\\\', True),
Antlr4Symbol('\v', True),
Antlr4Symbol('\f', True),
Antlr4Symbol('\a', True),
Antlr4Symbol("'\\\\' [0-9] [0-9]? [0-9]?", True, True),
Antlr4Symbol("'\\\\' 'x' [0-9A-Fa-f] [0-9A-Fa-f]?", True, True),
]), is_fragment=True)
p.rules.append(string_char)
any_printable_ASCII_character_except_white_space = Antlr4Rule(
"ANY_PRINTABLE_ASCII_CHARACTER_EXCEPT_WHITE_SPACE",
Antlr4Symbol("'\\u0021'..'\\u007E'", True, True),
is_fragment=True)
p.rules.append(any_printable_ASCII_character_except_white_space)
and not isinstance(s2, Antlr4Indent)):
all_to_remove = False
if _s and all_to_remove:
s.pop()
continue
break
if r.name == "signature":
# rm ()? as it is in ()? every where it is used
a, b = r.body[0].body
a = a.body
b = b.body
# ( ( type_mark ( COMMA type_mark )* )? ( RETURN type_mark )? )?
r.body = Antlr4Selection([
Antlr4Sequence([a, Antlr4Newline(),
Antlr4Indent(1)]),
Antlr4Sequence([a, b,
Antlr4Newline(),
Antlr4Indent(1)]),
Antlr4Sequence([b, Antlr4Newline()]),
])
HEADER = """/*
* Grammar extracted from the VHDL 1993, 2002, 2008, 2018 standard and then merged together
* (the standard is selected by parser property)
*/
grammar vhdl;
"""
with open("vhdl.g4", "w") as f:
f.write("\n\n")
f.write(HEADER)
for kw in keywords:
def _optimise_selections(elm: iAntlr4GramElem):
"""
Reduce selection options which differ only in single item
to a sequence with selection of different items.
Example:
a:
b c d
| b e d;
to a: b (c | e) d;
:note: ignores visuals
:note: similar sequences have to be directly after each other
because if they were not the priority of choices would
be changed
"""
if isinstance(elm, Antlr4Sequence):
modified = False
for e in elm:
modified = modified or _optimise_selections(e)
return modified
elif isinstance(elm, Antlr4Selection):
# List[Tuple[index of different item,
# List[Tuple[index in choices, selection options to replace]]]]
to_reduce = []
# tuple (index in choices, value)
similar_choices = []
diff_in = None
for c_i, choice in enumerate(elm):
if not similar_choices:
if isinstance(
choice,
Antlr4Sequence) and len_without_visuals(choice) > 1:
similar_choices.append((c_i, choice))
continue
else:
_, prev = similar_choices[0]
compatible = True
if (isinstance(prev, Antlr4Sequence)
and isinstance(choice, Antlr4Sequence)
and len_without_visuals(prev)
== len_without_visuals(choice)):
# check if differs in a single item
for i, (prev_item, current_item) in enumerate(
zip(iter_non_visuals(prev),
iter_non_visuals(choice))):
if prev_item != current_item:
if diff_in == i or diff_in is None:
diff_in = i
else:
compatible = False
break
if compatible:
similar_choices.append((c_i, choice))
else:
compatible = False
if not compatible:
if len(similar_choices) > 1:
to_reduce.append((diff_in, similar_choices))
# reset search
if isinstance(choice, Antlr4Sequence
) and len_without_visuals(choice) > 1:
similar_choices = [(c_i, choice)]
else:
similar_choices = []
diff_in = None
if len(similar_choices) > 1:
to_reduce.append((diff_in, similar_choices))
offset = 0
for diff_in, _choices in to_reduce:
choices = [c[1] for c in _choices]
start_i = _choices[0][0] + offset
assert len(_choices) > 1
try:
assert elm[start_i] is choices[0]
except AssertionError:
raise
diff_item_substitution = Antlr4Selection(
[index_non_visual(c, diff_in) for c in choices])
part_to_exclude = index_non_visual(choices[0], diff_in)
new_choice = Antlr4Sequence([
(e if e is not part_to_exclude else diff_item_substitution)
for e in choices[0]
])
elm[start_i] = new_choice
del elm[start_i + 1:start_i + len(choices)]
offset -= len(choices) - 1
return len(to_reduce)
return False
def match_replace_fn(o):
if o == cls:
return Antlr4Selection([o, deepcopy(intf)])
def remove_useless_and_normalize_names(p):
renames = {}
for k, v in SvRule2Antlr4Rule.SPEC_SYMB.items():
renames[k] = v
# rm_newline_from_simple_rules(p.rules)
# nts = get_used_non_terminals(p.rules)
# def_nts = get_defined_non_terminals(p.rules)
# overspecified
# finish_number 0 - 2
replace_rule("finish_number", "UNSIGNED_NUMBER", renames, p)
# scalar_constant 1b number
replace_rule("scalar_constant", "integral_number", renames, p)
# init_val 1b value
replace_rule("init_val", "integral_number", renames, p)
# edge_descriptor 2 tristate digits
# edge_descriptor: '01' | '10' | Z_OR_X ZERO_OR_ONE | ZERO_OR_ONE Z_OR_X;
# dpi_spec_string two concrete strings
replace_rule("dpi_spec_string", "STRING_LITERAL", renames, p)
# #0 -> # UNSIGNED_NUMBER
primitive_delay = Antlr4Rule(
"primitive_delay",
Antlr4Sequence([
Antlr4Symbol("HASH", False),
Antlr4Symbol("UNSIGNED_NUMBER", False),
]))
p.rules.append(primitive_delay)
replace_rule("#0", "primitive_delay", renames, p)
# all same
ps_identifier_rules = [
"ps_class_identifier",
"ps_covergroup_identifier",
"ps_checker_identifier",
]
for name in ps_identifier_rules:
replace_rule(name, "ps_identifier", renames, p)
ps_or_hierarchical_id_rules = [
"ps_or_hierarchical_net_identifier",
"ps_or_hierarchical_property_identifier",
"ps_or_hierarchical_sequence_identifier",
"ps_or_hierarchical_tf_identifier",
]
ps_or_hierarchical_identifier = Antlr4Rule(
"ps_or_hierarchical_identifier",
Antlr4Selection([
Antlr4Sequence([
Antlr4Option(Antlr4Symbol("package_scope", False)),
Antlr4Symbol("identifier", False)
]),
Antlr4Symbol("hierarchical_identifier", False),
]))
p.rules.append(ps_or_hierarchical_identifier)
for name in ps_or_hierarchical_id_rules:
replace_rule(name, "ps_or_hierarchical_identifier", renames, p)
to_lexer = [
"c_identifier",
"unsigned_number",
"simple_identifier",
"system_tf_identifier",
"unsigned_number",
"string_literal",
"binary_number",
"octal_number",
"hex_number",
"octal_number",
"hex_number",
"fixed_point_number",
"escaped_identifier",
"unbased_unsized_literal",
"time_literal",
# because it is very hard to switch mode to parse
# edge_descriptor and it is easy to just parse coma separated list of 2 chars
"edge_control_specifier",
"level_symbol",
"output_symbol",
"edge_symbol",
"file_path_spec",
]
for tl in to_lexer:
renames[tl] = tl.upper()
fragments = {
"binary_value", "octal_value", "hex_value", "decimal_base",
"binary_base", "octal_base", "hex_base", "non_zero_unsigned_number",
"size", "sign", "edge_descriptor", "non_zero_decimal_digit",
"decimal_digit", "binary_digit", "octal_digit", "hex_digit", "x_digit",
"z_digit", "exp", 'white_space', 'zero_or_one', 'z_or_x',
'Any_ASCII_Characters',
"any_printable_ASCII_character_except_white_space", "time_unit"
}
for r in p.rules:
if r.name.startswith("$"):
renames[r.name] = r.name.replace("$", "dolar_")
for fr in fragments:
if r.name in fragments:
r.is_fragment = True
renames[fr] = fr.upper()
for r in p.rules:
rm_redunt_whitespaces_on_end(r)
identifier_rule_equivalents = {
r.name
for r in collect_simple_rules(p.rules, "identifier")
}
hierarchical_identifier_rule_equivalents = {
r.name
for r in collect_simple_rules(p.rules, "hierarchical_identifier")
}
to_remove = {
"comment",
"one_line_comment",
"block_comment",
"comment_text",
"white_space",
}
to_remove.update(identifier_rule_equivalents)
to_remove.update(hierarchical_identifier_rule_equivalents)
simple_rules_to_remove = [
"default_clause", # default kw
"variable_port_type",
"limit_value", # used only in more specific limit values
"dpi_function_proto", # used only in dpi block so we already know
"dpi_task_proto", # used only in dpi block so we already know
"property_lvar_port_direction", # used only in property so we already know
# "consecutive_repetition", # useless
"trans_item",
"ordered_parameter_assignment",
"function_statement",
"case_expression",
"case_item_expression",
"open_value_range", # used only in open_range_list so we already know
"constant_assignment_pattern_expression", # parser do not see the difference between const/non const
"clockvar", # used only in clockvar_expression
"path_delay_expression", # used only in more specific rules
"constant_function_call", # parser do not see the difference between const/non const
"function_subroutine_call",
"constant_let_expression", # parser do not see the difference between const/non const
"attr_name", # used only in attr_spec
"array_identifier", # never used
"checker_identifier", # used only in rule with same name
"class_identifier",
"class_variable_identifier",
"clocking_identifier",
"config_identifier",
"const_identifier",
"constraint_identifier",
"covergroup_identifier",
"covergroup_variable_identifier",
"cover_point_identifier",
"cross_identifier",
"enum_identifier",
"formal_identifier",
"function_identifier",
"generate_block_identifier",
"genvar_identifier",
"hierarchical_array_identifier",
"hierarchical_block_identifier",
"hierarchical_event_identifier",
"hierarchical_net_identifier",
"hierarchical_parameter_identifier",
"hierarchical_property_identifier",
"hierarchical_sequence_identifier",
"hierarchical_task_identifier",
"hierarchical_tf_identifier",
"hierarchical_variable_identifier",
"index_variable_identifier",
"interface_identifier",
"interface_instance_identifier",
# "inout_port_identifier",
# "input_port_identifier",
"instance_identifier",
"member_identifier",
"method_identifier",
"modport_identifier",
"module_identifier",
"net_identifier",
# "output_port_identifier"
"package_identifier",
"parameter_identifier",
"port_identifier",
"production_identifier",
"program_identifier",
"property_identifier",
"sequence_identifier",
"signal_identifier",
"specparam_identifier",
"task_identifier",
"tf_identifier",
"terminal_identifier",
"topmodule_identifier",
"udp_identifier",
"variable_identifier",
]
for sr in simple_rules_to_remove:
remove_simple_rule(sr, p)
p.rules = [r for r in p.rules if r.name not in to_remove]
for idname in identifier_rule_equivalents:
renames[idname] = "identifier"
for idname in hierarchical_identifier_rule_equivalents:
renames[idname] = "hierarchical_identifier"
apply_rename = generate_renamer(renames, True)
for r in p.rules:
r.walk(apply_rename)
r.walk(mark_regex)
for k, v in SvRule2Antlr4Rule.SPEC_SYMB.items():
body = Antlr4Symbol(k, True)
r = Antlr4Rule(v, body)
p.rules.append(r)
# because C_IDENTIFIER is just normal identifier without $ and can match identifiers
for r in p.rules:
if r.name == "identifier":
r.body.insert(0, Antlr4Symbol("C_IDENTIFIER", False))