%PDF- %PDF-
Direktori : /opt/cloudlinux/venv/lib64/python3.11/site-packages/tomlkit/ |
Current File : //opt/cloudlinux/venv/lib64/python3.11/site-packages/tomlkit/parser.py |
from __future__ import annotations import datetime import re import string from tomlkit._compat import decode from tomlkit._utils import RFC_3339_LOOSE from tomlkit._utils import _escaped from tomlkit._utils import parse_rfc3339 from tomlkit.container import Container from tomlkit.exceptions import EmptyKeyError from tomlkit.exceptions import EmptyTableNameError from tomlkit.exceptions import InternalParserError from tomlkit.exceptions import InvalidCharInStringError from tomlkit.exceptions import InvalidControlChar from tomlkit.exceptions import InvalidDateError from tomlkit.exceptions import InvalidDateTimeError from tomlkit.exceptions import InvalidNumberError from tomlkit.exceptions import InvalidTimeError from tomlkit.exceptions import InvalidUnicodeValueError from tomlkit.exceptions import ParseError from tomlkit.exceptions import UnexpectedCharError from tomlkit.exceptions import UnexpectedEofError from tomlkit.items import AoT from tomlkit.items import Array from tomlkit.items import Bool from tomlkit.items import BoolType from tomlkit.items import Comment from tomlkit.items import Date from tomlkit.items import DateTime from tomlkit.items import Float from tomlkit.items import InlineTable from tomlkit.items import Integer from tomlkit.items import Item from tomlkit.items import Key from tomlkit.items import KeyType from tomlkit.items import Null from tomlkit.items import SingleKey from tomlkit.items import String from tomlkit.items import StringType from tomlkit.items import Table from tomlkit.items import Time from tomlkit.items import Trivia from tomlkit.items import Whitespace from tomlkit.source import Source from tomlkit.toml_char import TOMLChar from tomlkit.toml_document import TOMLDocument CTRL_I = 0x09 # Tab CTRL_J = 0x0A # Line feed CTRL_M = 0x0D # Carriage return CTRL_CHAR_LIMIT = 0x1F CHR_DEL = 0x7F class Parser: """ Parser for TOML documents. """ def __init__(self, string: str) -> None: # Input to parse self._src = Source(decode(string)) self._aot_stack: list[Key] = [] @property def _state(self): return self._src.state @property def _idx(self): return self._src.idx @property def _current(self): return self._src.current @property def _marker(self): return self._src.marker def extract(self) -> str: """ Extracts the value between marker and index """ return self._src.extract() def inc(self, exception: type[ParseError] | None = None) -> bool: """ Increments the parser if the end of the input has not been reached. Returns whether or not it was able to advance. """ return self._src.inc(exception=exception) def inc_n(self, n: int, exception: type[ParseError] | None = None) -> bool: """ Increments the parser by n characters if the end of the input has not been reached. """ return self._src.inc_n(n=n, exception=exception) def consume(self, chars, min=0, max=-1): """ Consume chars until min/max is satisfied is valid. """ return self._src.consume(chars=chars, min=min, max=max) def end(self) -> bool: """ Returns True if the parser has reached the end of the input. """ return self._src.end() def mark(self) -> None: """ Sets the marker to the index's current position """ self._src.mark() def parse_error(self, exception=ParseError, *args, **kwargs): """ Creates a generic "parse error" at the current position. """ return self._src.parse_error(exception, *args, **kwargs) def parse(self) -> TOMLDocument: body = TOMLDocument(True) # Take all keyvals outside of tables/AoT's. while not self.end(): # Break out if a table is found if self._current == "[": break # Otherwise, take and append one KV item = self._parse_item() if not item: break key, value = item if (key is not None and key.is_multi()) or not self._merge_ws(value, body): # We actually have a table try: body.append(key, value) except Exception as e: raise self.parse_error(ParseError, str(e)) from e self.mark() while not self.end(): key, value = self._parse_table() if isinstance(value, Table) and value.is_aot_element(): # This is just the first table in an AoT. Parse the rest of the array # along with it. value = self._parse_aot(value, key) try: body.append(key, value) except Exception as e: raise self.parse_error(ParseError, str(e)) from e body.parsing(False) return body def _merge_ws(self, item: Item, container: Container) -> bool: """ Merges the given Item with the last one currently in the given Container if both are whitespace items. Returns True if the items were merged. """ last = container.last_item() if not last: return False if not isinstance(item, Whitespace) or not isinstance(last, Whitespace): return False start = self._idx - (len(last.s) + len(item.s)) container.body[-1] = ( container.body[-1][0], Whitespace(self._src[start : self._idx]), ) return True def _is_child(self, parent: Key, child: Key) -> bool: """ Returns whether a key is strictly a child of another key. AoT siblings are not considered children of one another. """ parent_parts = tuple(parent) child_parts = tuple(child) if parent_parts == child_parts: return False return parent_parts == child_parts[: len(parent_parts)] def _parse_item(self) -> tuple[Key | None, Item] | None: """ Attempts to parse the next item and returns it, along with its key if the item is value-like. """ self.mark() with self._state as state: while True: c = self._current if c == "\n": # Found a newline; Return all whitespace found up to this point. self.inc() return None, Whitespace(self.extract()) elif c in " \t\r": # Skip whitespace. if not self.inc(): return None, Whitespace(self.extract()) elif c == "#": # Found a comment, parse it indent = self.extract() cws, comment, trail = self._parse_comment_trail() return None, Comment(Trivia(indent, cws, comment, trail)) elif c == "[": # Found a table, delegate to the calling function. return else: # Beginning of a KV pair. # Return to beginning of whitespace so it gets included # as indentation for the KV about to be parsed. state.restore = True break return self._parse_key_value(True) def _parse_comment_trail(self, parse_trail: bool = True) -> tuple[str, str, str]: """ Returns (comment_ws, comment, trail) If there is no comment, comment_ws and comment will simply be empty. """ if self.end(): return "", "", "" comment = "" comment_ws = "" self.mark() while True: c = self._current if c == "\n": break elif c == "#": comment_ws = self.extract() self.mark() self.inc() # Skip # # The comment itself while not self.end() and not self._current.is_nl(): code = ord(self._current) if code == CHR_DEL or code <= CTRL_CHAR_LIMIT and code != CTRL_I: raise self.parse_error(InvalidControlChar, code, "comments") if not self.inc(): break comment = self.extract() self.mark() break elif c in " \t\r": self.inc() else: raise self.parse_error(UnexpectedCharError, c) if self.end(): break trail = "" if parse_trail: while self._current.is_spaces() and self.inc(): pass if self._current == "\r": self.inc() if self._current == "\n": self.inc() if self._idx != self._marker or self._current.is_ws(): trail = self.extract() return comment_ws, comment, trail def _parse_key_value(self, parse_comment: bool = False) -> tuple[Key, Item]: # Leading indent self.mark() while self._current.is_spaces() and self.inc(): pass indent = self.extract() # Key key = self._parse_key() self.mark() found_equals = self._current == "=" while self._current.is_kv_sep() and self.inc(): if self._current == "=": if found_equals: raise self.parse_error(UnexpectedCharError, "=") else: found_equals = True if not found_equals: raise self.parse_error(UnexpectedCharError, self._current) if not key.sep: key.sep = self.extract() else: key.sep += self.extract() # Value val = self._parse_value() # Comment if parse_comment: cws, comment, trail = self._parse_comment_trail() meta = val.trivia if not meta.comment_ws: meta.comment_ws = cws meta.comment = comment meta.trail = trail else: val.trivia.trail = "" val.trivia.indent = indent return key, val def _parse_key(self) -> Key: """ Parses a Key at the current position; WS before the key must be exhausted first at the callsite. """ self.mark() while self._current.is_spaces() and self.inc(): # Skip any leading whitespace pass if self._current in "\"'": return self._parse_quoted_key() else: return self._parse_bare_key() def _parse_quoted_key(self) -> Key: """ Parses a key enclosed in either single or double quotes. """ # Extract the leading whitespace original = self.extract() quote_style = self._current key_type = next((t for t in KeyType if t.value == quote_style), None) if key_type is None: raise RuntimeError("Should not have entered _parse_quoted_key()") key_str = self._parse_string( StringType.SLB if key_type == KeyType.Basic else StringType.SLL ) if key_str._t.is_multiline(): raise self.parse_error(UnexpectedCharError, key_str._t.value) original += key_str.as_string() self.mark() while self._current.is_spaces() and self.inc(): pass original += self.extract() key = SingleKey(str(key_str), t=key_type, sep="", original=original) if self._current == ".": self.inc() key = key.concat(self._parse_key()) return key def _parse_bare_key(self) -> Key: """ Parses a bare key. """ while ( self._current.is_bare_key_char() or self._current.is_spaces() ) and self.inc(): pass original = self.extract() key = original.strip() if not key: # Empty key raise self.parse_error(EmptyKeyError) if " " in key: # Bare key with spaces in it raise self.parse_error(ParseError, f'Invalid key "{key}"') key = SingleKey(key, KeyType.Bare, "", original) if self._current == ".": self.inc() key = key.concat(self._parse_key()) return key def _parse_value(self) -> Item: """ Attempts to parse a value at the current position. """ self.mark() c = self._current trivia = Trivia() if c == StringType.SLB.value: return self._parse_basic_string() elif c == StringType.SLL.value: return self._parse_literal_string() elif c == BoolType.TRUE.value[0]: return self._parse_true() elif c == BoolType.FALSE.value[0]: return self._parse_false() elif c == "[": return self._parse_array() elif c == "{": return self._parse_inline_table() elif c in "+-" or self._peek(4) in { "+inf", "-inf", "inf", "+nan", "-nan", "nan", }: # Number while self._current not in " \t\n\r#,]}" and self.inc(): pass raw = self.extract() item = self._parse_number(raw, trivia) if item is not None: return item raise self.parse_error(InvalidNumberError) elif c in string.digits: # Integer, Float, Date, Time or DateTime while self._current not in " \t\n\r#,]}" and self.inc(): pass raw = self.extract() m = RFC_3339_LOOSE.match(raw) if m: if m.group(1) and m.group(5): # datetime try: dt = parse_rfc3339(raw) assert isinstance(dt, datetime.datetime) return DateTime( dt.year, dt.month, dt.day, dt.hour, dt.minute, dt.second, dt.microsecond, dt.tzinfo, trivia, raw, ) except ValueError: raise self.parse_error(InvalidDateTimeError) if m.group(1): try: dt = parse_rfc3339(raw) assert isinstance(dt, datetime.date) date = Date(dt.year, dt.month, dt.day, trivia, raw) self.mark() while self._current not in "\t\n\r#,]}" and self.inc(): pass time_raw = self.extract() time_part = time_raw.rstrip() trivia.comment_ws = time_raw[len(time_part) :] if not time_part: return date dt = parse_rfc3339(raw + time_part) assert isinstance(dt, datetime.datetime) return DateTime( dt.year, dt.month, dt.day, dt.hour, dt.minute, dt.second, dt.microsecond, dt.tzinfo, trivia, raw + time_part, ) except ValueError: raise self.parse_error(InvalidDateError) if m.group(5): try: t = parse_rfc3339(raw) assert isinstance(t, datetime.time) return Time( t.hour, t.minute, t.second, t.microsecond, t.tzinfo, trivia, raw, ) except ValueError: raise self.parse_error(InvalidTimeError) item = self._parse_number(raw, trivia) if item is not None: return item raise self.parse_error(InvalidNumberError) else: raise self.parse_error(UnexpectedCharError, c) def _parse_true(self): return self._parse_bool(BoolType.TRUE) def _parse_false(self): return self._parse_bool(BoolType.FALSE) def _parse_bool(self, style: BoolType) -> Bool: with self._state: style = BoolType(style) # only keep parsing for bool if the characters match the style # try consuming rest of chars in style for c in style: self.consume(c, min=1, max=1) return Bool(style, Trivia()) def _parse_array(self) -> Array: # Consume opening bracket, EOF here is an issue (middle of array) self.inc(exception=UnexpectedEofError) elems: list[Item] = [] prev_value = None while True: # consume whitespace mark = self._idx self.consume(TOMLChar.SPACES + TOMLChar.NL) indent = self._src[mark : self._idx] newline = set(TOMLChar.NL) & set(indent) if newline: elems.append(Whitespace(indent)) continue # consume comment if self._current == "#": cws, comment, trail = self._parse_comment_trail(parse_trail=False) elems.append(Comment(Trivia(indent, cws, comment, trail))) continue # consume indent if indent: elems.append(Whitespace(indent)) continue # consume value if not prev_value: try: elems.append(self._parse_value()) prev_value = True continue except UnexpectedCharError: pass # consume comma if prev_value and self._current == ",": self.inc(exception=UnexpectedEofError) elems.append(Whitespace(",")) prev_value = False continue # consume closing bracket if self._current == "]": # consume closing bracket, EOF here doesn't matter self.inc() break raise self.parse_error(UnexpectedCharError, self._current) try: res = Array(elems, Trivia()) except ValueError: pass else: return res def _parse_inline_table(self) -> InlineTable: # consume opening bracket, EOF here is an issue (middle of array) self.inc(exception=UnexpectedEofError) elems = Container(True) trailing_comma = None while True: # consume leading whitespace mark = self._idx self.consume(TOMLChar.SPACES) raw = self._src[mark : self._idx] if raw: elems.add(Whitespace(raw)) if not trailing_comma: # None: empty inline table # False: previous key-value pair was not followed by a comma if self._current == "}": # consume closing bracket, EOF here doesn't matter self.inc() break if ( trailing_comma is False or trailing_comma is None and self._current == "," ): # Either the previous key-value pair was not followed by a comma # or the table has an unexpected leading comma. raise self.parse_error(UnexpectedCharError, self._current) else: # True: previous key-value pair was followed by a comma if self._current == "}" or self._current == ",": raise self.parse_error(UnexpectedCharError, self._current) key, val = self._parse_key_value(False) elems.add(key, val) # consume trailing whitespace mark = self._idx self.consume(TOMLChar.SPACES) raw = self._src[mark : self._idx] if raw: elems.add(Whitespace(raw)) # consume trailing comma trailing_comma = self._current == "," if trailing_comma: # consume closing bracket, EOF here is an issue (middle of inline table) self.inc(exception=UnexpectedEofError) return InlineTable(elems, Trivia()) def _parse_number(self, raw: str, trivia: Trivia) -> Item | None: # Leading zeros are not allowed sign = "" if raw.startswith(("+", "-")): sign = raw[0] raw = raw[1:] if len(raw) > 1 and ( raw.startswith("0") and not raw.startswith(("0.", "0o", "0x", "0b", "0e")) or sign and raw.startswith(".") ): return None if raw.startswith(("0o", "0x", "0b")) and sign: return None digits = "[0-9]" base = 10 if raw.startswith("0b"): digits = "[01]" base = 2 elif raw.startswith("0o"): digits = "[0-7]" base = 8 elif raw.startswith("0x"): digits = "[0-9a-f]" base = 16 # Underscores should be surrounded by digits clean = re.sub(f"(?i)(?<={digits})_(?={digits})", "", raw).lower() if "_" in clean: return None if ( clean.endswith(".") or not clean.startswith("0x") and clean.split("e", 1)[0].endswith(".") ): return None try: return Integer(int(sign + clean, base), trivia, sign + raw) except ValueError: try: return Float(float(sign + clean), trivia, sign + raw) except ValueError: return None def _parse_literal_string(self) -> String: with self._state: return self._parse_string(StringType.SLL) def _parse_basic_string(self) -> String: with self._state: return self._parse_string(StringType.SLB) def _parse_escaped_char(self, multiline): if multiline and self._current.is_ws(): # When the last non-whitespace character on a line is # a \, it will be trimmed along with all whitespace # (including newlines) up to the next non-whitespace # character or closing delimiter. # """\ # hello \ # world""" tmp = "" while self._current.is_ws(): tmp += self._current # consume the whitespace, EOF here is an issue # (middle of string) self.inc(exception=UnexpectedEofError) continue # the escape followed by whitespace must have a newline # before any other chars if "\n" not in tmp: raise self.parse_error(InvalidCharInStringError, self._current) return "" if self._current in _escaped: c = _escaped[self._current] # consume this char, EOF here is an issue (middle of string) self.inc(exception=UnexpectedEofError) return c if self._current in {"u", "U"}: # this needs to be a unicode u, ue = self._peek_unicode(self._current == "U") if u is not None: # consume the U char and the unicode value self.inc_n(len(ue) + 1) return u raise self.parse_error(InvalidUnicodeValueError) raise self.parse_error(InvalidCharInStringError, self._current) def _parse_string(self, delim: StringType) -> String: # only keep parsing for string if the current character matches the delim if self._current != delim.unit: raise self.parse_error( InternalParserError, f"Invalid character for string type {delim}", ) # consume the opening/first delim, EOF here is an issue # (middle of string or middle of delim) self.inc(exception=UnexpectedEofError) if self._current == delim.unit: # consume the closing/second delim, we do not care if EOF occurs as # that would simply imply an empty single line string if not self.inc() or self._current != delim.unit: # Empty string return String(delim, "", "", Trivia()) # consume the third delim, EOF here is an issue (middle of string) self.inc(exception=UnexpectedEofError) delim = delim.toggle() # convert delim to multi delim self.mark() # to extract the original string with whitespace and all value = "" # A newline immediately following the opening delimiter will be trimmed. if delim.is_multiline(): if self._current == "\n": # consume the newline, EOF here is an issue (middle of string) self.inc(exception=UnexpectedEofError) else: cur = self._current with self._state(restore=True): if self.inc(): cur += self._current if cur == "\r\n": self.inc_n(2, exception=UnexpectedEofError) escaped = False # whether the previous key was ESCAPE while True: code = ord(self._current) if ( delim.is_singleline() and not escaped and (code == CHR_DEL or code <= CTRL_CHAR_LIMIT and code != CTRL_I) ) or ( delim.is_multiline() and not escaped and ( code == CHR_DEL or code <= CTRL_CHAR_LIMIT and code not in [CTRL_I, CTRL_J, CTRL_M] ) ): raise self.parse_error(InvalidControlChar, code, "strings") elif not escaped and self._current == delim.unit: # try to process current as a closing delim original = self.extract() close = "" if delim.is_multiline(): # Consume the delimiters to see if we are at the end of the string close = "" while self._current == delim.unit: close += self._current self.inc() if len(close) < 3: # Not a triple quote, leave in result as-is. # Adding back the characters we already consumed value += close continue if len(close) == 3: # We are at the end of the string return String(delim, value, original, Trivia()) if len(close) >= 6: raise self.parse_error(InvalidCharInStringError, self._current) value += close[:-3] original += close[:-3] return String(delim, value, original, Trivia()) else: # consume the closing delim, we do not care if EOF occurs as # that would simply imply the end of self._src self.inc() return String(delim, value, original, Trivia()) elif delim.is_basic() and escaped: # attempt to parse the current char as an escaped value, an exception # is raised if this fails value += self._parse_escaped_char(delim.is_multiline()) # no longer escaped escaped = False elif delim.is_basic() and self._current == "\\": # the next char is being escaped escaped = True # consume this char, EOF here is an issue (middle of string) self.inc(exception=UnexpectedEofError) else: # this is either a literal string where we keep everything as is, # or this is not a special escaped char in a basic string value += self._current # consume this char, EOF here is an issue (middle of string) self.inc(exception=UnexpectedEofError) def _parse_table( self, parent_name: Key | None = None, parent: Table | None = None ) -> tuple[Key, Table | AoT]: """ Parses a table element. """ if self._current != "[": raise self.parse_error( InternalParserError, "_parse_table() called on non-bracket character." ) indent = self.extract() self.inc() # Skip opening bracket if self.end(): raise self.parse_error(UnexpectedEofError) is_aot = False if self._current == "[": if not self.inc(): raise self.parse_error(UnexpectedEofError) is_aot = True try: key = self._parse_key() except EmptyKeyError: raise self.parse_error(EmptyTableNameError) from None if self.end(): raise self.parse_error(UnexpectedEofError) elif self._current != "]": raise self.parse_error(UnexpectedCharError, self._current) key.sep = "" full_key = key name_parts = tuple(key) if any(" " in part.key.strip() and part.is_bare() for part in name_parts): raise self.parse_error( ParseError, f'Invalid table name "{full_key.as_string()}"' ) missing_table = False if parent_name: parent_name_parts = tuple(parent_name) else: parent_name_parts = () if len(name_parts) > len(parent_name_parts) + 1: missing_table = True name_parts = name_parts[len(parent_name_parts) :] values = Container(True) self.inc() # Skip closing bracket if is_aot: # TODO: Verify close bracket self.inc() cws, comment, trail = self._parse_comment_trail() result = Null() table = Table( values, Trivia(indent, cws, comment, trail), is_aot, name=name_parts[0].key if name_parts else key.key, display_name=full_key.as_string(), is_super_table=False, ) if len(name_parts) > 1: if missing_table: # Missing super table # i.e. a table initialized like this: [foo.bar] # without initializing [foo] # # So we have to create the parent tables table = Table( Container(True), Trivia("", cws, comment, trail), is_aot and name_parts[0] in self._aot_stack, is_super_table=True, name=name_parts[0].key, ) result = table key = name_parts[0] for i, _name in enumerate(name_parts[1:]): child = table.get( _name, Table( Container(True), Trivia(indent, cws, comment, trail), is_aot and i == len(name_parts) - 2, is_super_table=i < len(name_parts) - 2, name=_name.key, display_name=full_key.as_string() if i == len(name_parts) - 2 else None, ), ) if is_aot and i == len(name_parts) - 2: table.raw_append(_name, AoT([child], name=table.name, parsed=True)) else: table.raw_append(_name, child) table = child values = table.value else: if name_parts: key = name_parts[0] while not self.end(): item = self._parse_item() if item: _key, item = item if not self._merge_ws(item, values): table.raw_append(_key, item) else: if self._current == "[": _, key_next = self._peek_table() if self._is_child(full_key, key_next): key_next, table_next = self._parse_table(full_key, table) table.raw_append(key_next, table_next) # Picking up any sibling while not self.end(): _, key_next = self._peek_table() if not self._is_child(full_key, key_next): break key_next, table_next = self._parse_table(full_key, table) table.raw_append(key_next, table_next) break else: raise self.parse_error( InternalParserError, "_parse_item() returned None on a non-bracket character.", ) if isinstance(result, Null): result = table if is_aot and (not self._aot_stack or full_key != self._aot_stack[-1]): result = self._parse_aot(result, full_key) return key, result def _peek_table(self) -> tuple[bool, Key]: """ Peeks ahead non-intrusively by cloning then restoring the initial state of the parser. Returns the name of the table about to be parsed, as well as whether it is part of an AoT. """ # we always want to restore after exiting this scope with self._state(save_marker=True, restore=True): if self._current != "[": raise self.parse_error( InternalParserError, "_peek_table() entered on non-bracket character", ) # AoT self.inc() is_aot = False if self._current == "[": self.inc() is_aot = True try: return is_aot, self._parse_key() except EmptyKeyError: raise self.parse_error(EmptyTableNameError) from None def _parse_aot(self, first: Table, name_first: Key) -> AoT: """ Parses all siblings of the provided table first and bundles them into an AoT. """ payload = [first] self._aot_stack.append(name_first) while not self.end(): is_aot_next, name_next = self._peek_table() if is_aot_next and name_next == name_first: _, table = self._parse_table(name_first) payload.append(table) else: break self._aot_stack.pop() return AoT(payload, parsed=True) def _peek(self, n: int) -> str: """ Peeks ahead n characters. n is the max number of characters that will be peeked. """ # we always want to restore after exiting this scope with self._state(restore=True): buf = "" for _ in range(n): if self._current not in " \t\n\r#,]}" + self._src.EOF: buf += self._current self.inc() continue break return buf def _peek_unicode(self, is_long: bool) -> tuple[str | None, str | None]: """ Peeks ahead non-intrusively by cloning then restoring the initial state of the parser. Returns the unicode value is it's a valid one else None. """ # we always want to restore after exiting this scope with self._state(save_marker=True, restore=True): if self._current not in {"u", "U"}: raise self.parse_error( InternalParserError, "_peek_unicode() entered on non-unicode value" ) self.inc() # Dropping prefix self.mark() if is_long: chars = 8 else: chars = 4 if not self.inc_n(chars): value, extracted = None, None else: extracted = self.extract() if extracted[0].lower() == "d" and extracted[1].strip("01234567"): return None, None try: value = chr(int(extracted, 16)) except (ValueError, OverflowError): value = None return value, extracted