Files
Sheerka-Old/src/parsers/BaseNodeParser.py
T

800 lines
25 KiB
Python

from collections import namedtuple
from dataclasses import dataclass
from enum import Enum
import core.utils
from core.builtin_concepts import BuiltinConcepts
from core.concept import VARIABLE_PREFIX, Concept, DEFINITION_TYPE_BNF, ConceptParts
from core.sheerka.ExecutionContext import ExecutionContext
from core.tokenizer import TokenKind, LexerError, Token
from parsers.BaseParser import Node, BaseParser, ErrorNode
DEBUG_COMPILED = True
@dataclass()
class LexerNode(Node):
start: int # starting index in the tokens list
end: int # ending index in the tokens list
tokens: list = None # tokens
source: str = None # string representation of what was parsed
def __post_init__(self):
if self.source is None:
self.source = BaseParser.get_text_from_tokens(self.tokens)
def __eq__(self, other):
if not isinstance(other, LexerNode):
return False
return self.start == other.start and \
self.end == other.end and \
self.source == other.source and \
self.tokens == other.tokens
def fix_source(self, force=True):
if force or self.source is None:
self.source = BaseParser.get_text_from_tokens(self.tokens)
return self
class UnrecognizedTokensNode(LexerNode):
def __init__(self, start, end, tokens):
super().__init__(start, end, tokens)
self.is_frozen = False
self.parenthesis_count = 0
def freeze(self):
self.is_frozen = True
def reset(self):
self.start = self.end = -1
self.tokens.clear()
self.is_frozen = False
self.parenthesis_count = 0
def has_open_paren(self):
return self.parenthesis_count > 0
def add_token(self, token, pos):
if self.is_frozen:
raise Exception("The node is frozen")
if self.end != -1 and pos == self.end + 2:
# add the missing whitespace
p = self.tokens[-1] # previous token
self.tokens.append(Token(TokenKind.WHITESPACE, " ", p.index + 1, p.line, p.column + 1))
self.tokens.append(token)
self.end = pos
if self.start == -1:
self.start = pos
if token.type == TokenKind.LPAR:
self.parenthesis_count += 1
if token.type == TokenKind.RPAR:
self.parenthesis_count -= 1
return self
def not_whitespace(self):
return not self.is_whitespace()
def is_whitespace(self):
for t in self.tokens:
if t.type not in (TokenKind.WHITESPACE, TokenKind.NEWLINE):
return False
return True
def is_empty(self):
return len(self.tokens) == 0
def __eq__(self, other):
if isinstance(other, utnode):
return self.start == other.start and \
self.end == other.end and \
self.source == other.source
if isinstance(other, UTN):
return other == self
if not isinstance(other, UnrecognizedTokensNode):
return False
return self.start == other.start and \
self.end == other.end and \
self.source == other.source
def __hash__(self):
return hash((self.start, self.end, self.source))
def __repr__(self):
return f"UnrecognizedTokensNode(start={self.start}, end={self.end}, source='{self.source}')"
def clone(self):
clone = UnrecognizedTokensNode(self.start, self.end, self.tokens[:])
clone.is_frozen = self.is_frozen
clone.parenthesis_count = self.parenthesis_count
return clone
class ConceptNode(LexerNode):
"""
Returned by the BnfNodeParser
It represents a recognized concept
"""
def __init__(self, concept, start, end, tokens=None, source=None, underlying=None):
super().__init__(start, end, tokens, source)
self.concept = concept
self.underlying = underlying
self.fix_source(False)
def __eq__(self, other):
if id(self) == id(other):
return True
if isinstance(other, (CN, CNC)):
return other == self
if isinstance(other, cnode):
return self.concept.key == other.concept_key and \
self.start == other.start and \
self.end == other.end and \
self.source == other.source
if isinstance(other, short_cnode):
return self.concept.key == other.concept_key and self.source == other.source
if not isinstance(other, ConceptNode):
return False
return self.concept == other.concept and \
self.start == other.start and \
self.end == other.end and \
self.source == other.source and \
self.underlying == other.underlying
def __hash__(self):
return hash((self.concept, self.start, self.end, self.source, self.underlying))
def __repr__(self):
text = f"ConceptNode(concept='{self.concept}', source='{self.source}', start={self.start}, end={self.end}"
if DEBUG_COMPILED:
for k, v in self.concept.compiled.items():
text += f", {k}='{v}'"
return text + ")"
def clone(self):
# do we need to clone the concept as well ?
clone = ConceptNode(self.concept, self.start, self.end, self.tokens, self.source, self.underlying)
return clone
class SourceCodeNode(LexerNode):
"""
Returned when some source code (like Python source code is recognized)
"""
def __init__(self, node, start, end, tokens=None, source=None, return_value=None):
super().__init__(start, end, tokens, source)
self.node = node # The PythonNode (or whatever language node) that is found
self.return_value = return_value # original result of the parsing
def __eq__(self, other):
if isinstance(other, scnode):
return self.start == other.start and \
self.end == other.end and \
self.source == other.source
if isinstance(other, SCN):
return other == self
if not isinstance(other, SourceCodeNode):
return False
return self.node == other.node and \
self.start == other.start and \
self.end == other.end and \
self.source == other.source
def __hash__(self):
return hash((self.start, self.end, self.source))
def __repr__(self):
return f"SourceCodeNode(start={self.start}, end={self.end}, source='{self.source}')"
class SourceCodeWithConceptNode(LexerNode):
"""
Kind of temporary version for SourceCodeNode
I know that there is some code,
I know that there are some concepts
I just don't want to make the glue yet
So I push all the nodes into one big bag
"""
def __init__(self, first_node, last_node, content_nodes=None):
super().__init__(9999, -1, None) # why not sys.maxint ?
self.first = first_node
self.last = last_node
self.nodes = content_nodes or []
self.has_unrecognized = False
self.fix_all_pos()
def add_node(self, node):
self.nodes.append(node)
self.fix_pos(node)
return self
def __eq__(self, other):
if id(self) == id(other):
return True
if not isinstance(other, SourceCodeWithConceptNode):
return False
if self.start != other.start or self.end != other.end:
return False
if self.first != other.first:
return False
if self.last != other.last:
return False
if len(self.nodes) != len(other.nodes):
return False
for self_node, other_node in zip(self.nodes, other.nodes):
if self_node != other_node:
return False
# at last
return True
def __hash__(self):
return hash((self.first, self.last, len(self.nodes)))
def __repr__(self):
return f"SourceCodeWithConceptNode(start={self.start}, end={self.end}, source='{self.source}')"
def fix_all_pos(self):
for n in [self.first, self.last] + self.nodes:
self.fix_pos(n)
def fix_pos(self, node):
if hasattr(node, "start") and node.start is not None:
if node.start < self.start:
self.start = node.start
if hasattr(node, "end") and node.end is not None:
if node.end > self.end:
self.end = node.end
return self
def pseudo_fix_source(self):
self.source = self.first.source
for n in self.nodes:
self.source += " "
if hasattr(n, "source"):
self.source += n.source
elif hasattr(n, "concept"):
self.source += str(n.concept)
else:
self.source += " unknown"
self.source += self.last.source
return self
def clone(self):
clone = SourceCodeWithConceptNode(self.first, self.last, self.nodes)
return clone
@dataclass()
class GrammarErrorNode(ErrorNode):
message: str
class SyaAssociativity(Enum):
Left = "left"
Right = "right"
No = "No"
def __repr__(self):
return self.value
cnode = namedtuple("ConceptNode", "concept_key start end source")
short_cnode = namedtuple("ConceptNode", "concept_key source")
utnode = namedtuple("utnode", "start end source")
scnode = namedtuple("scnode", "start end source")
@dataclass(init=False)
class SCWC:
"""
SourceNodeWithConcept tester class
It matches with a SourceNodeWithConcept
but it's easier to instantiate during the tests
"""
first: LexerNode
last: LexerNode
content: tuple
def __init__(self, first, last, *args):
self.first = first
self.last = last
self.content = args
class HelperWithPos:
def __init__(self, start=None, end=None):
self.start = start
self.end = end
self.start_is_fixed = start is not None
self.end_is_fixed = end is not None
def fix_pos(self, node):
if not self.start_is_fixed:
start = node.start if hasattr(node, "start") else \
node[0] if isinstance(node, tuple) else None
if start is not None and (self.start is None or start < self.start):
self.start = start
if not self.end_is_fixed:
end = node.end if hasattr(node, "end") else \
node[1] if isinstance(node, tuple) else None
if end is not None and (self.end is None or end > self.end):
self.end = end
return self
class SCN(HelperWithPos):
"""
SourceCodeNode tester class
It matches with SourceCodeNode but with less constraints
SCN == SourceCodeNode if source, start, end (start and end are not validated when None)
"""
def __init__(self, source, start=None, end=None):
super().__init__(start, end)
self.source = source
def __eq__(self, other):
if id(self) == id(other):
return True
if isinstance(other, SourceCodeNode):
if self.source != other.source:
return False
if self.start is not None and self.start != other.start:
return False
if self.end is not None and self.end != other.end:
return False
return True
if not isinstance(other, CN):
return False
return self.source == other.source and \
self.start == other.start and \
self.end == other.end
def __hash__(self):
return hash((self.source, self.start, self.end))
def __repr__(self):
txt = f"SCN(source='{self.source}'"
if self.start is not None:
txt += f", start={self.start}"
if self.end is not None:
txt += f", end={self.end}"
return txt + ")"
class CN(HelperWithPos):
"""
ConceptNode tester class
It matches with ConceptNode but with less constraints
CNC == ConceptNode if concept key, start, end and source are the same
"""
def __init__(self, concept, start=None, end=None, source=None):
"""
:param concept: Concept or concept_key (only the key is used anyway)
:param start:
:param end:
:param source:
"""
super().__init__(start, end)
self.concept_key = concept.key if isinstance(concept, Concept) else concept
self.source = source
self.concept = concept if isinstance(concept, Concept) else None
def fix_source(self, str_tokens):
self.source = "".join(str_tokens)
return self
def __eq__(self, other):
if id(self) == id(other):
return True
if isinstance(other, ConceptNode):
if other.concept is None:
return False
if other.concept.key != self.concept_key:
return False
if self.start is not None and self.start != other.start:
return False
if self.end is not None and self.end != other.end:
return False
if self.source is not None and self.source != other.source:
return False
return True
if not isinstance(other, CN):
return False
return self.concept_key == other.concept_key and \
self.start == other.start and \
self.end == other.end and \
self.source == other.source
def __hash__(self):
return hash((self.concept_key, self.start, self.end, self.source))
def __repr__(self):
if self.concept:
txt = f"CN(concept='{self.concept}'"
else:
txt = f"CN(concept_key='{self.concept_key}'"
txt += f", source='{self.source}'"
if self.start is not None:
txt += f", start={self.start}"
if self.end is not None:
txt += f", end={self.end}"
return txt + ")"
class CNC(CN):
"""
ConceptNode for Compiled tester class
It matches with ConceptNode
But focuses on the 'compiled' property of the concept
CNC == ConceptNode if CNC.compiled == ConceptNode.concept.compiled
"""
def __init__(self, concept_key, start=None, end=None, source=None, exclude_body=False, **kwargs):
super().__init__(concept_key, start, end, source)
self.compiled = kwargs
self.exclude_body = exclude_body
def __eq__(self, other):
if id(self) == id(other):
return True
if isinstance(other, ConceptNode):
if other.concept is None:
return False
if other.concept.key != self.concept_key:
return False
if self.start is not None and self.start != other.start:
return False
if self.end is not None and self.end != other.end:
return False
if self.source is not None and self.source != other.source:
return False
if self.exclude_body:
to_compare = {k: v for k, v in other.concept.compiled.items() if k != ConceptParts.BODY}
else:
to_compare = other.concept.compiled
return self.compiled == to_compare
if not isinstance(other, CNC):
return False
return self.concept_key == other.concept_key and \
self.start == other.start and \
self.end == other.end and \
self.source == other.source and \
self.compiled == other.compiled
def __repr__(self):
if self.concept:
txt = f"CNC(concept='{self.concept}'"
else:
txt = f"CNC(concept_key='{self.concept_key}'"
txt += f", source='{self.source}'"
if self.start is not None:
txt += f", start={self.start}"
if self.end is not None:
txt += f", end={self.end}"
for k, v in self.compiled.items():
txt += f", {k}='{v}'"
return txt + ")"
class UTN(HelperWithPos):
"""
Tester class for UnrecognizedTokenNode
compare the source, and start, end if defined
"""
def __init__(self, source, start=None, end=None):
"""
:param concept: Concept or concept_key (only the key is used anyway)
:param start:
:param end:
:param source:
"""
super().__init__(start, end)
self.source = source
def __eq__(self, other):
if id(self) == id(other):
return True
if isinstance(other, UnrecognizedTokensNode):
return self.start == other.start and \
self.end == other.end and \
self.source == other.source
if not isinstance(other, UTN):
return False
return self.start == other.start and \
self.end == other.end and \
self.source == other.source
def __hash__(self):
return hash((self.source, self.start, self.end))
def __repr__(self):
txt = f"UTN( source='{self.source}'"
if self.start is not None:
txt += f", start={self.start}"
if self.end is not None:
txt += f", end={self.end}"
return txt + ")"
class BaseNodeParser(BaseParser):
def __init__(self, name, priority, **kwargs):
super().__init__(name, priority)
if 'sheerka' in kwargs:
sheerka = kwargs.get("sheerka")
self.concepts_by_first_keyword = sheerka.resolved_concepts_by_first_keyword
else:
self.concepts_by_first_keyword = None
self.token = None
self.pos = -1
self.tokens = None
self.context: ExecutionContext = None
self.text = None
self.sheerka = None
def init_from_concepts(self, context, concepts, **kwargs):
"""
Initialize the parser with a list of concepts
For unit tests convenience
:param context
:param concepts
:return:
"""
concepts_by_first_keyword = self.get_concepts_by_first_keyword(context, concepts).body
self.concepts_by_first_keyword = self.resolve_concepts_by_first_keyword(context, concepts_by_first_keyword).body
def reset_parser(self, context, text):
self.context = context
self.sheerka = context.sheerka
self.text = text
try:
self.tokens = list(self.get_input_as_tokens(text))
except LexerError as e:
self.add_error(self.sheerka.new(BuiltinConcepts.ERROR, body=e), False)
return False
self.token = None
self.pos = -1
return True
def add_error(self, error, next_token=True):
self.error_sink.append(error)
if next_token:
self.next_token()
return error
def get_token(self) -> Token:
return self.token
def next_token(self, skip_whitespace=True):
if self.token and self.token.type == TokenKind.EOF:
return False
self.pos += 1
self.token = self.tokens[self.pos]
if skip_whitespace:
while self.token.type == TokenKind.WHITESPACE or self.token.type == TokenKind.NEWLINE:
self.pos += 1
self.token = self.tokens[self.pos]
return self.token.type != TokenKind.EOF
def get_concepts(self, token, to_keep, custom=None, to_map=None, strip_quotes=False):
"""
Tries to find if there are concepts that match the value of the token
:param token:
:param to_keep: predicate to tell if the concept is eligible
:param custom: lambda name -> List[Concepts] that gives extra concepts, according to the name
:param to_map:
:param strip_quotes: Remove quotes from strings
:return:
"""
if token.type == TokenKind.WHITESPACE:
return None
if token.type == TokenKind.STRING:
name = token.value[1:-1] if strip_quotes else token.value
elif token.type == TokenKind.KEYWORD:
name = token.value.value
else:
name = token.value
custom_concepts = custom(name) if custom else []
result = []
if name in self.concepts_by_first_keyword:
for concept_id in self.concepts_by_first_keyword.get(name):
concept = self.sheerka.get_by_id(concept_id)
if not to_keep(concept):
continue
concept = to_map(self, concept) if to_map else concept
result.append(concept)
return result + custom_concepts
return custom_concepts if custom else None
@staticmethod
def get_token_value(token):
if token.type == TokenKind.STRING:
return token.value[1:-1]
elif token.type == TokenKind.KEYWORD:
return token.value.value
else:
return token.value
@staticmethod
def get_concepts_by_first_keyword(context, concepts, use_sheerka=False):
"""
Create the map describing the first token expected by a concept
:param context:
:param concepts: lists of concepts to parse
:param use_sheerka: if True, update concepts_by_first_keyword from sheerka
:return:
"""
sheerka = context.sheerka
res = sheerka.cache_manager.copy(sheerka.CONCEPTS_BY_FIRST_KEYWORD_ENTRY) if use_sheerka else {}
for concept in concepts:
keywords = BaseNodeParser.get_first_tokens(sheerka, concept)
if keywords is None:
# no first token found for a concept ?
return sheerka.ret(sheerka.name, False, concept)
for keyword in keywords:
res.setdefault(keyword, []).append(concept.id)
return sheerka.ret("BaseNodeParser", True, res)
@staticmethod
def resolve_concepts_by_first_keyword(context, concepts_by_first_keyword):
sheerka = context.sheerka
def _make_unique(elements):
keys = {}
for e in elements:
keys[e] = 1
return list(keys.keys())
def _resolve_concepts(concept_str):
resolved = []
to_resolve = []
concept = sheerka.get_by_id(core.utils.unstr_concept(concept_str)[1])
if sheerka.isaset(context, concept):
concepts = sheerka.get_set_elements(context, concept)
else:
concepts = [concept]
for concept in concepts:
BaseNodeParser.ensure_bnf(context, concept) # need to make sure that it cannot fail
keywords = BaseNodeParser.get_first_tokens(sheerka, concept)
for keyword in keywords:
(to_resolve if keyword.startswith("c:|") else resolved).append(keyword)
for concept_to_resolve_str in to_resolve:
resolved += _resolve_concepts(concept_to_resolve_str)
return resolved
res = {}
for k, v in concepts_by_first_keyword.items():
if k.startswith("c:|"):
resolved_keywords = _resolve_concepts(k)
for resolved in resolved_keywords:
res.setdefault(resolved, []).extend(v)
else:
res.setdefault(k, []).extend(v)
# 'uniquify' the lists
for k, v in res.items():
res[k] = _make_unique(v)
return sheerka.ret("BaseNodeParser", True, res)
@staticmethod
def resolve_sya_associativity_and_precedence(context, sya):
pass
@staticmethod
def get_first_tokens(sheerka, concept):
"""
:param sheerka:
:param concept:
:return:
"""
if concept.bnf:
from parsers.BnfNodeParser import BnfNodeFirstTokenVisitor
bnf_visitor = BnfNodeFirstTokenVisitor(sheerka)
bnf_visitor.visit(concept.bnf)
return bnf_visitor.first_tokens
else:
keywords = concept.key.split()
for keyword in keywords:
if keyword.startswith(VARIABLE_PREFIX):
continue
return [keyword]
return None
@staticmethod
def ensure_bnf(context, concept, parser_name="BaseNodeParser"):
if concept.metadata.definition_type == DEFINITION_TYPE_BNF and not concept.bnf:
from parsers.BnfParser import BnfParser
regex_parser = BnfParser()
desc = f"Resolving BNF {concept.metadata.definition}"
with context.push(parser_name, obj=concept, desc=desc) as sub_context:
sub_context.add_inputs(parser_input=concept.metadata.definition)
bnf_parsing_ret_val = regex_parser.parse(sub_context, concept.metadata.definition)
sub_context.add_values(return_values=bnf_parsing_ret_val)
if not bnf_parsing_ret_val.status:
raise Exception(bnf_parsing_ret_val.value)
concept.bnf = bnf_parsing_ret_val.body.body
if concept.id:
context.sheerka.get_by_id(concept.id).bnf = concept.bnf # update bnf in cache