"""CSS selector parser.""" from __future__ import annotations import re from functools import lru_cache from . import util from . import css_match as cm from . import css_types as ct from .util import SelectorSyntaxError import warnings from typing import Match, Any, Callable, Iterator, Iterable, Mapping, cast from dataclasses import dataclass from collections import UserDict import threading __all__ = ('SelectorSyntaxError',) RE_LOCK = threading.Lock() SEL_LOCK = threading.RLock() UNICODE_REPLACEMENT_CHAR = 0xFFFD # Non-ASCII identifier code point ranges (inclusive): https://drafts.csswg.org/css-syntax-3/#non-ascii-ident-code-point NON_ASCII_IDENT_RANGES = ( (0x00B7, 0x00B7), (0x00C0, 0x00D6), (0x00D8, 0x00F6), (0x00F8, 0x037D), (0x037F, 0x1FFF), (0x200C, 0x200D), (0x203F, 0x2040), (0x2070, 0x218F), (0x2C00, 0x2FEF), (0x3001, 0xD7FF), (0xF900, 0xFDCF), (0xFDF0, 0xFFFD), (0x10000, 0x10FFFF) ) SELECTOR_LIMIT = 8192 # Simple pseudo classes that take no parameters PSEUDO_SIMPLE = { ":any-link", ":empty", ":first-child", ":first-of-type", ":in-range", ":open", ":out-of-range", ":last-child", ":last-of-type", ":link", ":only-child", ":only-of-type", ":root", ':checked', ':default', ':disabled', ':enabled', ':indeterminate', ':optional', ':placeholder-shown', ':read-only', ':read-write', ':required', ':scope', ':defined', ':muted' } # Supported, simple pseudo classes that match nothing in the Soup Sieve environment PSEUDO_SIMPLE_NO_MATCH = { ':active', ':autofill', ':buffering', ':current', ':focus', ':focus-visible', ':focus-within', ':fullscreen', ':future', ':host', ':hover', ':local-link', ':past', ':paused', ':picture-in-picture', ':playing', ':popover-open', ':seeking', ':stalled', ':target', ':target-within', ':user-invalid', ':volume-locked', ':visited' } # Complex pseudo classes that take selector lists PSEUDO_COMPLEX = { ':contains', ':-soup-contains', ':-soup-contains-own', ':has', ':is', ':matches', ':not', ':where' } PSEUDO_COMPLEX_NO_MATCH = { ':current', ':host', ':host-context' } # Complex pseudo classes that take very specific parameters and are handled special PSEUDO_SPECIAL = { ':dir', ':lang', ':nth-child', ':nth-last-child', ':nth-last-of-type', ':nth-of-type' } PSEUDO_SUPPORTED = PSEUDO_SIMPLE | PSEUDO_SIMPLE_NO_MATCH | PSEUDO_COMPLEX | PSEUDO_COMPLEX_NO_MATCH | PSEUDO_SPECIAL # Sub-patterns parts # Whitespace NEWLINE = r'(?>\r\n|[\n\f\r])' WS = fr'(?:[ \t]|{NEWLINE})' # Comments COMMENTS = r'(?:/\*(?:[^*]|\*(?!/))*+\*/)' # Whitespace with comments included WSC = fr'(?:{WS}|{COMMENTS})' # CSS escapes CSS_ESCAPES = fr'(?:\\(?:[a-f0-9]{{1,6}}+{WS}?+|[^\r\n\f]|$))' CSS_STRING_ESCAPES = fr'(?:\\(?:[a-f0-9]{{1,6}}+{WS}?+|{NEWLINE}|[^\r\n\f]|$))' # CSS identifier code points (browsers treat all non-ASCII code points as identifier code points) IDENT_START = r'[A-Za-z_\x80-\U0010ffff]' IDENT_CHAR = r'[-0-9A-Za-z_\x80-\U0010ffff]' # CSS Identifier IDENTIFIER = fr''' (?:(?:--|-?+(?:{IDENT_START}|{CSS_ESCAPES})) (?:{IDENT_CHAR}|{CSS_ESCAPES})*+) ''' # `nth` content NTH = fr'[-+]?+(?:[0-9]++n?+|n)(?:(?<=n){WSC}*+[-+]{WSC}*+[0-9]++)?+' # Value: quoted string or identifier VALUE = fr''' (?:"(?:\\(?:{NEWLINE}|[^\r\n\f])|[^\\"\r\n\f])*+"| '(?:\\(?:{NEWLINE}|[^\r\n\f])|[^\\'\r\n\f])*+'| {IDENTIFIER}) ''' # Attribute value comparison. `!=` is handled special as it is non-standard. ATTR = fr'(?:{WSC}*+(?P[!~^|*$]?+=){WSC}*+(?P{VALUE})(?:{WSC}*+(?P[is]))?+)?+' # Selector patterns # IDs (`#id`) PAT_ID = fr'\#{IDENTIFIER}' # Classes (`.class`) PAT_CLASS = fr'\.{IDENTIFIER}' # Prefix:Tag (`prefix|tag`) PAT_TAG = fr'(?P(?:{IDENTIFIER}|\*)?+\|)?+(?P{IDENTIFIER}|\*)' # Attributes (`[attr]`, `[attr=value]`, etc.) PAT_ATTR = fr'\[{WSC}*+(?P(?:{IDENTIFIER}|\*)?+\|(?!=))?+(?P{IDENTIFIER}){ATTR}{WSC}*+\]' # Pseudo class (`:pseudo-class`, `:pseudo-class(`) PAT_PSEUDO_CLASS = fr'(?P:{IDENTIFIER})(?P\({WSC}*+)?+' # Pseudo class special patterns. Matches `:pseudo-class(` for special case pseudo classes. PAT_PSEUDO_CLASS_SPECIAL = fr'(?P:{IDENTIFIER})(?P\({WSC}*+)' # Custom pseudo class (`:--custom-pseudo`) PAT_PSEUDO_CLASS_CUSTOM = fr'(?P:(?=--){IDENTIFIER})' # Nesting ampersand selector. Matches `&` PAT_AMP = r'&' # Closing pseudo group (`)`) PAT_PSEUDO_CLOSE = fr'{WSC}*+\)' # Pseudo element (`::pseudo-element`) PAT_PSEUDO_ELEMENT = fr':{PAT_PSEUDO_CLASS}' # At rule (`@page`, etc.) (not supported) PAT_AT_RULE = fr'@P{IDENTIFIER}' # Pseudo class `nth-child` (`:nth-child(an+b [of S]?)`, `:first-child`, etc.) PAT_PSEUDO_NTH_CHILD = fr''' (?P{PAT_PSEUDO_CLASS_SPECIAL} (?P{NTH}|even|odd))(?:{WSC}*+\)|(?P{COMMENTS}*+{WS}{WSC}*+of{COMMENTS}*+{WS}{WSC}*+)) ''' # Pseudo class `nth-of-type` (`:nth-of-type(an+b)`, `:first-of-type`, etc.) PAT_PSEUDO_NTH_TYPE = fr''' (?P{PAT_PSEUDO_CLASS_SPECIAL} (?P{NTH}|even|odd)){WSC}*+\) ''' # Pseudo class language (`:lang("*-de", en)`) PAT_PSEUDO_LANG = fr'{PAT_PSEUDO_CLASS_SPECIAL}(?P{VALUE}(?:{WSC}*+,{WSC}*+{VALUE})*+){WSC}*+\)' # Pseudo class direction (`:dir(ltr)`) PAT_PSEUDO_DIR = fr'{PAT_PSEUDO_CLASS_SPECIAL}(?Pltr|rtl){WSC}*+\)' # Combining characters (`>`, `~`, ` `, `+`, `,`) PAT_COMBINE = fr'{COMMENTS}*+(?={WS}|[,+>~]){WSC}*+(?P[,+>~])?+{WSC}*+' # Extra: Contains (`:contains(text)`) PAT_PSEUDO_CONTAINS = fr'{PAT_PSEUDO_CLASS_SPECIAL}(?P{VALUE}(?:{WSC}*+,{WSC}*+{VALUE})*+){WSC}*+\)' # Regular expressions # CSS escape pattern RE_CSS_ESC = re.compile(fr'(?:(\\[a-f0-9]{{1,6}}+{WSC}?+)|(\\[^\r\n\f])|(\\$))', re.I) RE_CSS_STR_ESC = re.compile(fr'(?:(\\[a-f0-9]{{1,6}}+{WS}?+)|(\\[^\r\n\f])|(\\$)|(\\{NEWLINE}))', re.I) # Pattern to break up `nth` specifiers RE_NTH = re.compile(fr'(?P[-+])?+(?P[0-9]++n?+|n)(?:(?<=n){WSC}*+(?P[-+]){WSC}*+(?P[0-9]++))?+', re.I) # Pattern to iterate multiple values. RE_VALUES = re.compile(fr'(?:(?P{VALUE})|(?P{WSC}*+,{WSC}*+))', re.X) # Whitespace checks RE_WS = re.compile(WS) RE_WS_BEGIN = re.compile(fr'^{WSC}*+') RE_WS_END = re.compile(fr'^(?:[ \t]|(?>\n\r|[\n\f\r])|{COMMENTS})*+') RE_CUSTOM = re.compile(fr'^{PAT_PSEUDO_CLASS_CUSTOM}$', re.X) RE_PSEUDO_CLASS_SPECIAL = re.compile(PAT_PSEUDO_CLASS_SPECIAL, re.I | re.X | re.U) RE_PSEUDO_IGNORE = re.compile(PAT_PSEUDO_CLASS, re.X | re.I | re.U) QUOTED = ("'", '"') # Constants # List split token COMMA_COMBINATOR = ',' # Relation token for descendant WS_COMBINATOR = " " # Parse flags FLG_PSEUDO = 0x01 FLG_NOT = 0x02 FLG_RELATIVE = 0x04 FLG_DEFAULT = 0x08 FLG_HTML = 0x10 FLG_INDETERMINATE = 0x20 FLG_OPEN = 0x40 FLG_IN_RANGE = 0x80 FLG_OUT_OF_RANGE = 0x100 FLG_PLACEHOLDER_SHOWN = 0x200 FLG_FORGIVE = 0x400 # Maximum cached patterns to store _MAXCACHE = 500 @lru_cache(maxsize=_MAXCACHE) def _cached_css_compile( pattern: str, namespaces: ct.Namespaces | None, custom: ct.CustomSelectors | None, ignore: tuple[str] | None = None, flags: int = 0, max_selectors: int = SELECTOR_LIMIT ) -> cm.SoupSieve: """Cached CSS compile.""" return cm.SoupSieve( pattern, CSSParser( pattern, custom=custom, ignore=ignore, flags=flags, max_selectors=max_selectors ).process_selectors(), namespaces, custom, flags ) def _purge_cache() -> None: """Purge the cache.""" _cached_css_compile.cache_clear() def css_unescape(content: str) -> str: """ Unescape CSS value. Strings allow for spanning the value on multiple strings by escaping a new line. """ def replace(m: Match[str]) -> str: """Replace with the appropriate substitute.""" if m.group(1): codepoint = int(m.group(1)[1:], 16) # A surrogate, or a value past Unicode, is `U+FFFD`. Zero already is. if ( codepoint == 0 or codepoint > 0x10FFFF or 0xD800 <= codepoint <= 0xDFFF ): codepoint = UNICODE_REPLACEMENT_CHAR value = chr(codepoint) elif m.group(2): value = m.group(2)[1:] elif m.group(3): value = '\ufffd' else: value = '' return value quoted = content.startswith(QUOTED) return (RE_CSS_STR_ESC if quoted else RE_CSS_ESC).sub( replace, content[1:-1] if quoted else content ) def escape(ident: str) -> str: """Escape identifier.""" string = [] length = len(ident) start_dash = length > 0 and ident[0] == '-' if length == 1 and start_dash: # Need to escape identifier that is a single `-` with no other characters string.append(f'\\{ident}') else: for index, c in enumerate(ident): codepoint = ord(c) if codepoint == 0x00: string.append('\ufffd') elif (0x01 <= codepoint <= 0x1F) or codepoint == 0x7F: string.append(f'\\{codepoint:x} ') elif (index == 0 or (start_dash and index == 1)) and (0x30 <= codepoint <= 0x39): string.append(f'\\{codepoint:x} ') elif ( codepoint in (0x2D, 0x5F) or (0x30 <= codepoint <= 0x39) or (0x41 <= codepoint <= 0x5A) or (0x61 <= codepoint <= 0x7A) or (codepoint >= 0x80 and any(start <= codepoint <= end for start, end in NON_ASCII_IDENT_RANGES)) ): string.append(c) else: string.append(f'\\{c}') return ''.join(string) class SelectorPattern: """Selector pattern.""" def __init__(self, name: str, pattern: str) -> None: """Initialize.""" self.name = name self.pattern = pattern self._re_pattern: re.Pattern[str] | None = None @property def re_pattern(self) -> re.Pattern[str]: """Retrieve the compiled regular expression pattern.""" with RE_LOCK: if self._re_pattern is None: self._re_pattern = re.compile(self.pattern, re.I | re.X | re.U) return self._re_pattern def get_name(self) -> str: """Get name.""" return self.name def match(self, selector: str, index: int, flags: int) -> Match[str] | None: """Match the selector.""" return self.re_pattern.match(selector, index) class SpecialPseudoPattern(SelectorPattern): """Selector pattern.""" def __init__(self, patterns: tuple[tuple[str, tuple[str, ...], str, type[SelectorPattern]], ...]) -> None: """Initialize.""" self.patterns = {} for p in patterns: name = p[0] pattern = p[3](name, p[2]) for pseudo in p[1]: self.patterns[pseudo] = pattern self.matched_name: SelectorPattern | None = None def get_name(self) -> str: """Get name.""" return '' if self.matched_name is None else self.matched_name.get_name() def match(self, selector: str, index: int, flags: int) -> Match[str] | None: """Match the selector.""" pseudo = None m = RE_PSEUDO_CLASS_SPECIAL.match(selector, index) if m: name = util.lower(css_unescape(m.group('name'))) pattern = self.patterns.get(name) if pattern: pseudo = pattern.match(selector, index, flags) if pseudo: self.matched_name = pattern return pseudo class _Selector: """ Intermediate selector class. This stores selector data for a compound selector as we are acquiring them. Once we are done collecting the data for a compound selector, we freeze the data in an object that can be pickled and hashed. """ def __init__(self, **kwargs: Any) -> None: """Initialize.""" self.tag: ct.SelectorTag | None = kwargs.get('tag', None) self.ids: list[str] = kwargs.get('ids', []) self.classes: list[str] = kwargs.get('classes', []) self.attributes: list[ct.SelectorAttribute] = kwargs.get('attributes', []) self.nth: list[ct.SelectorNth] = kwargs.get('nth', []) self.selectors: list[ct.SelectorList] = kwargs.get('selectors', []) self.relations: list[_Selector] = kwargs.get('relations', []) self.rel_type: str | None = kwargs.get('rel_type', None) self.contains: list[ct.SelectorContains] = kwargs.get('contains', []) self.lang: list[ct.SelectorLang] = kwargs.get('lang', []) self.flags: int = kwargs.get('flags', 0) self.no_match: bool = kwargs.get('no_match', False) def _freeze(self, relation: ct.SelectorList) -> ct.Selector | ct.SelectorNull: """Freeze self with an already frozen relation.""" if self.no_match: return ct.Null else: return ct.Selector( self.tag, tuple(self.ids), tuple(self.classes), tuple(self.attributes), tuple(self.nth), tuple(self.selectors), relation, self.rel_type, tuple(self.contains), tuple(self.lang), self.flags ) def freeze(self) -> ct.Selector | ct.SelectorNull: """ Freeze self. Relations form a chain where each selector's first relation is the next link and any remaining relations are carried down to follow it. As the far end of the chain must be frozen first, flatten the chain and then freeze it in reverse to avoid recursion. """ chain: list[_Selector] = [] pending = [self] while pending: sel = pending.pop() chain.append(sel) if sel.no_match: break pending.extend(reversed(sel.relations)) relation = ct.SelectorList() for sel in reversed(chain): frozen = sel._freeze(relation) relation = ct.SelectorList([frozen]) return frozen def __str__(self) -> str: # pragma: no cover """String representation.""" return ( f'_Selector(tag={self.tag!r}, ids={self.ids!r}, classes={self.classes!r}, attributes={self.attributes!r}, ' f'nth={self.nth!r}, selectors={self.selectors!r}, relations={self.relations!r}, ' f'rel_type={self.rel_type!r}, contains={self.contains!r}, lang={self.lang!r}, flags={self.flags!r}, ' f'no_match={self.no_match!r})' ) __repr__ = __str__ class _SelectorContext: """ Parsing state for a selector list. Context are created for each nesting level to avoid recursion. """ def __init__( self, flags: int, index: int, count: int, on_close: Callable[[ct.SelectorList], None] | None = None ) -> None: """Initialize.""" self.sel = _Selector() self.selectors: list[_Selector] = [] self.relations: list[_Selector] = [] self.has_selector = False self.rel_type = ":" + WS_COMBINATOR self.index = index self.count = count # Receives the frozen selector list when an open pseudo-class is closed. self.on_close = on_close # Setup various flags self.is_open = bool(flags & FLG_OPEN) self.is_pseudo = bool(flags & FLG_PSEUDO) self.is_relative = bool(flags & FLG_RELATIVE) self.is_not = bool(flags & FLG_NOT) self.is_html = bool(flags & FLG_HTML) self.is_default = bool(flags & FLG_DEFAULT) self.is_indeterminate = bool(flags & FLG_INDETERMINATE) self.is_in_range = bool(flags & FLG_IN_RANGE) self.is_out_of_range = bool(flags & FLG_OUT_OF_RANGE) self.is_placeholder_shown = bool(flags & FLG_PLACEHOLDER_SHOWN) self.is_forgive = bool(flags & FLG_FORGIVE) # The algorithm for relative selectors require an initial selector in the selector list if self.is_relative: self.selectors.append(_Selector()) def debug(self) -> None: # pragma: no cover """Print out useful debug stuff.""" for name in ( 'is_pseudo', 'is_open', 'is_relative', 'is_not', 'is_html', 'is_default', 'is_indeterminate', 'is_in_range', 'is_out_of_range', 'is_placeholder_shown', 'is_forgive' ): if getattr(self, name): print(f' {name}: True') @dataclass class CSSPattern: """A CSS pattern that hasn't been processed by `CSSParser` yet.""" selector: str flags: int class PseudoSelectorMap(UserDict[str, CSSPattern | ct.SelectorList]): """Pseudo selector map.""" def __setitem__(self, key: str, value: CSSPattern | ct.SelectorList) -> None: """Set item.""" self.data[key] = value def __getitem__(self, key: str) -> ct.SelectorList: """Get item.""" with SEL_LOCK: value = self.data[key] if isinstance(value, CSSPattern): value = CSSParser(value.selector).process_selectors(flags=value.flags) self.data[key] = value return value # CSS pattern for `:link` and `:any-link` CSS_LINK = CSSPattern('html|*:is(a, area)[href]', FLG_PSEUDO | FLG_HTML) # CSS pattern for `:checked` CSS_CHECKED = CSSPattern( ''' html|*:is(input[type=checkbox], input[type=radio])[checked], html|option[selected] ''', FLG_PSEUDO | FLG_HTML ) # CSS pattern for `:default` (must compile CSS_CHECKED first) CSS_DEFAULT = CSSPattern( ''' :checked, /* This pattern must be at the end. Special logic is applied to the last selector. */ html|form html|*:is(button, input)[type="submit"] ''', FLG_PSEUDO | FLG_HTML | FLG_DEFAULT ) # CSS pattern for `:indeterminate` CSS_INDETERMINATE = CSSPattern( ''' html|input[type="checkbox"][indeterminate], html|input[type="radio"]:is(:not([name]), [name=""]):not([checked]), html|progress:not([value]), /* This pattern must be at the end. Special logic is applied to the last selector. */ html|input[type="radio"][name]:not([name='']):not([checked]) ''', FLG_PSEUDO | FLG_HTML | FLG_INDETERMINATE ) # CSS pattern for `:disabled` CSS_DISABLED = CSSPattern( ''' html|*:is(input:not([type=hidden]), button, select, textarea, fieldset, optgroup, option, fieldset)[disabled], html|optgroup[disabled] > html|option, html|fieldset[disabled] > html|*:is(input:not([type=hidden]), button, select, textarea, fieldset), html|fieldset[disabled] > html|*:not(legend:nth-of-type(1)) html|*:is(input:not([type=hidden]), button, select, textarea, fieldset) ''', FLG_PSEUDO | FLG_HTML ) # CSS pattern for `:enabled` CSS_ENABLED = CSSPattern( ''' html|*:is(input:not([type=hidden]), button, select, textarea, fieldset, optgroup, option, fieldset):not(:disabled) ''', FLG_PSEUDO | FLG_HTML ) # CSS pattern for `:required` CSS_REQUIRED = CSSPattern('html|*:is(input, textarea, select)[required]', FLG_PSEUDO | FLG_HTML) # CSS pattern for `:optional` CSS_OPTIONAL = CSSPattern('html|*:is(input, textarea, select):not([required])', FLG_PSEUDO | FLG_HTML) # CSS pattern for `:placeholder-shown` CSS_PLACEHOLDER_SHOWN = CSSPattern( ''' html|input:is( :not([type]), [type=""], [type=text], [type=search], [type=url], [type=tel], [type=email], [type=password], [type=number] )[placeholder]:not([placeholder='']):is(:not([value]), [value=""]), html|textarea[placeholder]:not([placeholder='']) ''', FLG_PSEUDO | FLG_HTML | FLG_PLACEHOLDER_SHOWN ) # CSS pattern for `:read-write` (CSS_DISABLED must be compiled first) CSS_READ_WRITE = CSSPattern( ''' html|*:is( textarea, input:is( :not([type]), [type=""], [type=text], [type=search], [type=url], [type=tel], [type=email], [type=number], [type=password], [type=date], [type=datetime-local], [type=month], [type=time], [type=week] ) ):not([readonly], :disabled), html|*:is([contenteditable=""], [contenteditable="true" i]) ''', FLG_PSEUDO | FLG_HTML ) # CSS pattern for `:read-only` CSS_READ_ONLY = CSSPattern('html|*:not(:read-write)', FLG_PSEUDO | FLG_HTML) # CSS pattern for `:in-range` CSS_IN_RANGE = CSSPattern( ''' html|input:is( [type="date"], [type="month"], [type="week"], [type="time"], [type="datetime-local"], [type="number"], [type="range"] ):is( [min], [max] ) ''', FLG_PSEUDO | FLG_HTML | FLG_IN_RANGE ) # CSS pattern for `:out-of-range` CSS_OUT_OF_RANGE = CSSPattern( ''' html|input:is( [type="date"], [type="month"], [type="week"], [type="time"], [type="datetime-local"], [type="number"], [type="range"] ):is( [min], [max] ) ''', FLG_PSEUDO | FLG_HTML | FLG_OUT_OF_RANGE ) # CSS pattern for :open CSS_OPEN = CSSPattern('html|*:is(details, dialog)[open]', FLG_PSEUDO | FLG_HTML) # CSS pattern for :muted CSS_MUTED = CSSPattern('html|*:is(video, audio)[muted]', FLG_PSEUDO | FLG_HTML) # CSS pattern default for `:nth-child` "of S" feature CSS_NTH_OF_S_DEFAULT = CSSPattern("*|*", FLG_PSEUDO) class CSSParser: """Parse CSS selectors.""" CSS_TOKENS = ( SelectorPattern("pseudo_close", PAT_PSEUDO_CLOSE), SpecialPseudoPattern( ( ( "pseudo_contains", (':contains', ':-soup-contains', ':-soup-contains-own'), PAT_PSEUDO_CONTAINS, SelectorPattern ), ("pseudo_nth_child", (':nth-child', ':nth-last-child'), PAT_PSEUDO_NTH_CHILD, SelectorPattern), ("pseudo_nth_type", (':nth-of-type', ':nth-last-of-type'), PAT_PSEUDO_NTH_TYPE, SelectorPattern), ("pseudo_lang", (':lang',), PAT_PSEUDO_LANG, SelectorPattern), ("pseudo_dir", (':dir',), PAT_PSEUDO_DIR, SelectorPattern) ) ), SelectorPattern("pseudo_class_custom", PAT_PSEUDO_CLASS_CUSTOM), SelectorPattern("pseudo_class", PAT_PSEUDO_CLASS), SelectorPattern("pseudo_element", PAT_PSEUDO_ELEMENT), SelectorPattern("amp", PAT_AMP), SelectorPattern("at_rule", PAT_AT_RULE), SelectorPattern("id", PAT_ID), SelectorPattern("class", PAT_CLASS), SelectorPattern("tag", PAT_TAG), SelectorPattern("attribute", PAT_ATTR), SelectorPattern("combine", PAT_COMBINE) ) # Pseudos that expand to selectors PSEUDO_SELECTORS = PseudoSelectorMap( { ':link': CSS_LINK, ':any-link': CSS_LINK, ':checked': CSS_CHECKED, ':default': CSS_DEFAULT, ':indeterminate': CSS_INDETERMINATE, ':disabled': CSS_DISABLED, ':enabled': CSS_ENABLED, ':required': CSS_REQUIRED, ':muted': CSS_MUTED, ':open': CSS_OPEN, ':optional': CSS_OPTIONAL, ':read-only': CSS_READ_ONLY, ':read-write': CSS_READ_WRITE, ':in-range': CSS_IN_RANGE, ':out-of-range': CSS_OUT_OF_RANGE, ':placeholder-shown': CSS_PLACEHOLDER_SHOWN, '': CSS_NTH_OF_S_DEFAULT } ) def __init__( self, selector: str, *, custom: Mapping[str, ct.SelectorList] | None = None, ignore: Iterable[str] | None = None, flags: int = 0, max_selectors: int = SELECTOR_LIMIT ) -> None: """Initialize.""" self.pattern = selector.replace('\x00', '\ufffd') self.flags = flags self.debug = self.flags & util.DEBUG self.nostrict = self.flags & util.NOSTRICT self.custom = custom self.ignore = frozenset([] if ignore is None else ignore) self.count = 0 self.maxsel = max_selectors def increment_count(self, increment: int = 1) -> None: """Check the current selector count.""" self.count += increment if self.count > self.maxsel: raise ValueError(f'Selector exceeds pseudo-class nesting limit of {self.maxsel}') def parse_attribute_selector(self, sel: _Selector, m: Match[str], has_selector: bool) -> bool: """Create attribute selector from the returned regex match.""" op = m.group('cmp') case = util.lower(m.group('case')) if m.group('case') else None ns = css_unescape(m.group('attr_ns')[:-1]) if m.group('attr_ns') else '' attr = css_unescape(m.group('attr_name')) is_type = False pattern2 = None value = '' inverse = False if case: flags = (re.I if case == 'i' else 0) | re.DOTALL elif util.lower(attr) == 'type': flags = re.I | re.DOTALL is_type = True else: flags = re.DOTALL if op: value = css_unescape(m.group('value')) if not op: # Attribute name pattern = None elif op.startswith('^'): # Value start with # `^=` should match nothing if the value is empty, so use `(?!)` which cannot be matched. value = r'(?!)' if not value else re.escape(value) pattern = re.compile(r'^%s.*' % value, flags) elif op.startswith('$'): # Value ends with # `$=` should match nothing if the value is empty, so use `(?!)` which cannot be matched. value = r'(?!)' if not value else re.escape(value) pattern = re.compile(r'.*?%s$' % value, flags) elif op.startswith('*'): # Value contains # `*=` should match nothing if the value is empty, so use `(?!)` which cannot be matched. value = r'(?!)' if not value else re.escape(value) pattern = re.compile(r'.*?%s.*' % value, flags) elif op.startswith('~'): # Value contains word within space separated list # `*~` should match nothing if the value is empty, so use `(?!)` which cannot be matched. value = r'(?!)' if not value or RE_WS.search(value) else re.escape(value) pattern = re.compile(r'.*?(?:(?<=^)|(?<=[ \t\r\n\f]))%s(?=(?:[ \t\r\n\f]|$)).*' % value, flags) elif op.startswith('|'): # Value starts with word in dash separated list pattern = re.compile(r'^%s(?:-.*)?$' % re.escape(value), flags) elif op.startswith('!'): # Value does not matches pattern = re.compile(r'^(?!%s).*$' % re.escape(value), flags) inverse = True else: # Value matches pattern = re.compile(r'^%s$' % re.escape(value), flags) if is_type and pattern: pattern2 = re.compile(pattern.pattern) # Append the attribute selector sel_attr = ct.SelectorAttribute(attr, ns, pattern, pattern2, inverse) sel.attributes.append(sel_attr) has_selector = True return has_selector def parse_tag_pattern(self, sel: _Selector, m: Match[str], has_selector: bool) -> bool: """Parse tag pattern from regex match.""" prefix = css_unescape(m.group('tag_ns')[:-1]) if m.group('tag_ns') else None tag = css_unescape(m.group('tag_name')) sel.tag = ct.SelectorTag(tag, prefix) has_selector = True return has_selector def parse_pseudo_class_custom(self, sel: _Selector, m: Match[str], has_selector: bool) -> bool: """ Parse custom pseudo class alias. Compile custom selectors as we need them. When compiling a custom selector, set it to `None` in the dictionary so we can avoid an infinite loop. """ pseudo = util.lower(css_unescape(m.group('name'))) selector = self.custom.get(pseudo) if self.custom is not None else None if selector is None: raise SelectorSyntaxError( f"Undefined custom selector '{pseudo}' found at position {m.end(0)}", self.pattern, m.end(0) ) self.increment_count(selector.count) sel.selectors.append(selector) has_selector = True return has_selector def parse_pseudo_class( self, sel: _Selector, m: Match[str], has_selector: bool, is_html: bool, is_relative: bool ) -> tuple[bool, bool, _SelectorContext | None]: """ Parse pseudo class. If the pseudo-class opens a selector list, a new context is returned to parse its contents. """ context = None complex_pseudo = False pseudo = util.lower(css_unescape(m.group('name'))) if m.group('open'): complex_pseudo = True if complex_pseudo and pseudo in PSEUDO_COMPLEX: context = self.parse_pseudo_open(pseudo, m.end(0), sel.selectors.append, is_relative) has_selector = True elif not complex_pseudo and pseudo in PSEUDO_SIMPLE: if pseudo == ':root': sel.flags |= ct.SEL_ROOT elif pseudo == ':defined': sel.flags |= ct.SEL_DEFINED is_html = True elif pseudo == ':scope': sel.flags |= ct.SEL_SCOPE elif pseudo == ':empty': sel.flags |= ct.SEL_EMPTY elif pseudo in self.PSEUDO_SELECTORS: pseudo_selector = self.PSEUDO_SELECTORS[pseudo] self.increment_count(pseudo_selector.count) sel.selectors.append(pseudo_selector) elif pseudo == ':first-child': sel.nth.append(ct.SelectorNth(0, False, 1, False, False, ct.SelectorList())) elif pseudo == ':last-child': sel.nth.append(ct.SelectorNth(0, False, 1, False, True, ct.SelectorList())) elif pseudo == ':first-of-type': sel.nth.append(ct.SelectorNth(0, False, 1, True, False, ct.SelectorList())) elif pseudo == ':last-of-type': sel.nth.append(ct.SelectorNth(0, False, 1, True, True, ct.SelectorList())) elif pseudo == ':only-child': sel.nth.extend( [ ct.SelectorNth(0, False, 1, False, False, ct.SelectorList()), ct.SelectorNth(0, False, 1, False, True, ct.SelectorList()) ] ) elif pseudo == ':only-of-type': sel.nth.extend( [ ct.SelectorNth(0, False, 1, True, False, ct.SelectorList()), ct.SelectorNth(0, False, 1, True, True, ct.SelectorList()) ] ) has_selector = True elif complex_pseudo and pseudo in PSEUDO_COMPLEX_NO_MATCH: context = self.parse_pseudo_open(pseudo, m.end(0)) sel.no_match = True has_selector = True elif not complex_pseudo and pseudo in PSEUDO_SIMPLE_NO_MATCH: sel.no_match = True has_selector = True elif pseudo in PSEUDO_SUPPORTED: raise SelectorSyntaxError( f"Invalid syntax for pseudo class '{pseudo}'", self.pattern, m.start(0) ) else: raise SelectorSyntaxError( f"'{pseudo}' was detected as a pseudo-class and is either unsupported or invalid. " "If the syntax was not intended to be recognized as a pseudo-class, please escape the colon.", self.pattern, m.start(0) ) return has_selector, is_html, context def parse_pseudo_nth( self, sel: _Selector, m: Match[str], has_selector: bool ) -> tuple[bool, _SelectorContext | None]: """ Parse `nth` pseudo. If `of S` is specified, a new context is returned to parse `S`. """ context = None mdict = m.groupdict() if mdict.get('pseudo_nth_child'): postfix = '_child' else: postfix = '_type' mdict['name'] = util.lower(css_unescape(mdict['name'])) content = util.lower(mdict.get('nth' + postfix)) if content == 'even': # 2n s1 = 2 s2 = 0 var = True elif content == 'odd': # 2n+1 s1 = 2 s2 = 1 var = True else: nth_parts = cast(Match[str], RE_NTH.match(content)) _s1 = '-' if nth_parts.group('s1') and nth_parts.group('s1') == '-' else '' _s2 = '' a = nth_parts.group('a') var = a.endswith('n') if a.startswith('n'): _s1 += '1' elif var: _s1 += a[:-1] else: _s2 = _s1 + a _s1 = '' if nth_parts.group('s2') and nth_parts.group('s2') == '-': _s2 = '-' if nth_parts.group('b'): _s2 += nth_parts.group('b') s1 = int(_s1, 10) if _s1 else 0 s2 = int(_s2, 10) if _s2 else 0 pseudo_sel = mdict['name'] if postfix == '_child': last = pseudo_sel == ':nth-last-child' def append_nth(nth_sel: ct.SelectorList) -> None: """Append to this selector's `nth` attribute.""" sel.nth.append(ct.SelectorNth(s1, var, s2, False, last, nth_sel)) if m.group('of'): # Parse the rest of `of S`. context = self.parse_pseudo_open(pseudo_sel, m.end(0), append_nth) else: # Use default `*|*` for `of S`. nth_sel = self.PSEUDO_SELECTORS[''] self.increment_count(nth_sel.count) append_nth(nth_sel) else: if pseudo_sel == ':nth-of-type': sel.nth.append(ct.SelectorNth(s1, var, s2, True, False, ct.SelectorList())) elif pseudo_sel == ':nth-last-of-type': sel.nth.append(ct.SelectorNth(s1, var, s2, True, True, ct.SelectorList())) has_selector = True return has_selector, context def parse_pseudo_open( self, name: str, index: int, on_close: Callable[[ct.SelectorList], None] | None = None, is_relative: bool = False ) -> _SelectorContext: """Create a new context for nested pseudo-class selectors.""" flags = FLG_PSEUDO | FLG_OPEN if name == ':not': flags |= FLG_NOT elif name == ':has': if is_relative and not self.nostrict: raise SelectorSyntaxError( f":has() was found to be nested under :has() at position {index}", self.pattern, index ) flags |= FLG_RELATIVE elif name in (':where', ':is'): flags |= FLG_FORGIVE return self.create_context(flags, index, on_close) def create_context( self, flags: int, index: int, on_close: Callable[[ct.SelectorList], None] | None = None ) -> _SelectorContext: """Create a new selector context.""" context = _SelectorContext(flags, index, self.count, on_close) if self.debug: # pragma: no cover context.debug() return context def parse_has_combinator( self, sel: _Selector, m: Match[str], has_selector: bool, selectors: list[_Selector], rel_type: str, index: int ) -> tuple[bool, _Selector, str]: """Parse combinator tokens.""" combinator = m.group('relation') if not combinator: combinator = WS_COMBINATOR if combinator == COMMA_COMBINATOR: if not has_selector: raise SelectorSyntaxError( f"The combinator '{combinator}' at position {index}, must have a selector before it", self.pattern, index ) sel.rel_type = rel_type selectors[-1].relations.append(sel) rel_type = ":" + WS_COMBINATOR selectors.append(_Selector()) else: if has_selector and self.nostrict: # End the current selector and associate the leading combinator with this selector. sel.rel_type = rel_type selectors[-1].relations.append(sel) elif has_selector: raise SelectorSyntaxError( f'Using strict rules as proposed by CSS, the complex selectors at postition {index} is not allowed', self.pattern, index ) elif rel_type[1:] != WS_COMBINATOR: # It's impossible to have two whitespace combinators after each other as the patterns # will gobble up trailing whitespace. It is also impossible to have a whitespace # combinator after any other kind for the same reason. But we could have # multiple non-whitespace combinators. So if the current combinator is not a whitespace, # then we've hit the multiple combinator case, so we should fail. raise SelectorSyntaxError( f'The multiple combinators at position {index}', self.pattern, index ) # Set the leading combinator for the next selector. rel_type = ':' + combinator sel = _Selector() has_selector = False return has_selector, sel, rel_type def parse_combinator( self, sel: _Selector, m: Match[str], has_selector: bool, selectors: list[_Selector], relations: list[_Selector], is_pseudo: bool, is_forgive: bool, index: int ) -> tuple[bool, _Selector]: """Parse combinator tokens.""" combinator = m.group('relation') if not combinator: combinator = WS_COMBINATOR if not has_selector: if not is_forgive or combinator != COMMA_COMBINATOR: raise SelectorSyntaxError( f"The combinator '{combinator}' at position {index}, must have a selector before it", self.pattern, index ) # If we are in a forgiving pseudo class, just make the selector a "no match" if combinator == COMMA_COMBINATOR: self.increment_count() del relations[:] else: if combinator == COMMA_COMBINATOR: if not sel.tag and not is_pseudo: # Implied `*` sel.tag = ct.SelectorTag('*', None) sel.relations.extend(relations) selectors.append(sel) del relations[:] else: sel.relations.extend(relations) sel.rel_type = combinator del relations[:] relations.append(sel) sel = _Selector() has_selector = False return has_selector, sel def parse_class_id(self, sel: _Selector, m: Match[str], has_selector: bool) -> bool: """Parse HTML classes and ids.""" selector = m.group(0) if selector.startswith('.'): sel.classes.append(css_unescape(selector[1:])) else: sel.ids.append(css_unescape(selector[1:])) has_selector = True return has_selector def parse_pseudo_contains(self, sel: _Selector, m: Match[str], has_selector: bool) -> bool: """Parse contains.""" pseudo = util.lower(css_unescape(m.group('name'))) if pseudo == ":contains": warnings.warn( # noqa: B028 "The pseudo class ':contains' is deprecated, ':-soup-contains' should be used moving forward.", FutureWarning ) contains_own = pseudo == ":-soup-contains-own" patterns = [ css_unescape(token.group('value')) for token in RE_VALUES.finditer(m.group('values')) if not token.group('split') ] sel.contains.append(ct.SelectorContains(patterns, contains_own)) has_selector = True return has_selector def parse_pseudo_lang(self, sel: _Selector, m: Match[str], has_selector: bool) -> bool: """Parse pseudo language.""" patterns = [ css_unescape(token.group('value')) for token in RE_VALUES.finditer(m.group('values')) if not token.group('split') ] sel.lang.append(ct.SelectorLang(patterns)) has_selector = True return has_selector def parse_pseudo_dir(self, sel: _Selector, m: Match[str], has_selector: bool) -> bool: """Parse pseudo direction.""" value = ct.SEL_DIR_LTR if util.lower(m.group('dir')) == 'ltr' else ct.SEL_DIR_RTL sel.flags |= value has_selector = True return has_selector def parse_selectors( self, iselector: Iterator[tuple[str, Match[str]]], index: int = 0, flags: int = 0 ) -> ct.SelectorList: """ Parse selectors. Pseudo-classes that open a nested selector list push a new context onto a stack, and closing them pops back to the parent context, so nesting does not recurse. """ ctx = self.create_context(flags, index) stack: list[_SelectorContext] = [] for key, m in iselector: child: _SelectorContext | None = None if key not in ('combine', 'pseudo_close'): self.increment_count() # Handle parts if self.ignore: mi = RE_PSEUDO_IGNORE.match(m.group(0)) if mi is not None and mi.group('name').lower() in self.ignore: raise SelectorSyntaxError( f"The selector '{mi.group('name')}' at position {m.start(0)}, has been disallowed", self.pattern, m.start(0) ) if key == "at_rule": raise NotImplementedError(f"At-rules found at position {m.start(0)}") elif key == "amp": ctx.sel.flags |= ct.SEL_SCOPE ctx.has_selector = True elif key == 'pseudo_class_custom': ctx.has_selector = self.parse_pseudo_class_custom(ctx.sel, m, ctx.has_selector) elif key == 'pseudo_class': ctx.has_selector, ctx.is_html, child = self.parse_pseudo_class( ctx.sel, m, ctx.has_selector, ctx.is_html, ctx.is_relative ) elif key == 'pseudo_element': raise NotImplementedError(f"Pseudo-element found at position {m.start(0)}") elif key == 'pseudo_contains': ctx.has_selector = self.parse_pseudo_contains(ctx.sel, m, ctx.has_selector) elif key in ('pseudo_nth_type', 'pseudo_nth_child'): ctx.has_selector, child = self.parse_pseudo_nth(ctx.sel, m, ctx.has_selector) elif key == 'pseudo_lang': ctx.has_selector = self.parse_pseudo_lang(ctx.sel, m, ctx.has_selector) elif key == 'pseudo_dir': ctx.has_selector = self.parse_pseudo_dir(ctx.sel, m, ctx.has_selector) # Currently only supports HTML ctx.is_html = True elif key == 'pseudo_close': if not ctx.has_selector: if not ctx.is_forgive: raise SelectorSyntaxError( f"Expected a selector at position {m.start(0)}", self.pattern, m.start(0) ) ctx.sel.no_match = True if not ctx.is_open: raise SelectorSyntaxError( f"Unmatched pseudo-class close at position {m.start(0)}", self.pattern, m.start(0) ) # Hand the completed selector list to the parent and resume parsing the parent. selectors = self.finalize_selectors(ctx) if ctx.on_close is not None: ctx.on_close(selectors) ctx = stack.pop() continue elif key == 'combine': if ctx.is_relative: ctx.has_selector, ctx.sel, ctx.rel_type = self.parse_has_combinator( ctx.sel, m, ctx.has_selector, ctx.selectors, ctx.rel_type, ctx.index ) else: ctx.has_selector, ctx.sel = self.parse_combinator( ctx.sel, m, ctx.has_selector, ctx.selectors, ctx.relations, ctx.is_pseudo, ctx.is_forgive, ctx.index ) elif key == 'attribute': ctx.has_selector = self.parse_attribute_selector(ctx.sel, m, ctx.has_selector) elif key == 'tag': if ctx.has_selector: raise SelectorSyntaxError( f"Tag name found at position {m.start(0)} instead of at the start", self.pattern, m.start(0) ) ctx.has_selector = self.parse_tag_pattern(ctx.sel, m, ctx.has_selector) elif key in ('class', 'id'): ctx.has_selector = self.parse_class_id(ctx.sel, m, ctx.has_selector) ctx.index = m.end(0) # A pseudo-class opened a nested selector list, parse it before resuming this one. if child is not None: stack.append(ctx) ctx = child # Handle selectors that are not closed if ctx.is_open: raise SelectorSyntaxError( f"Unclosed pseudo-class at position {ctx.index}", self.pattern, ctx.index ) return self.finalize_selectors(ctx) def finalize_selectors(self, ctx: _SelectorContext) -> ct.SelectorList: """Complete the selector list of a context and freeze it.""" sel = ctx.sel selectors = ctx.selectors relations = ctx.relations has_selector = ctx.has_selector # Cleanup completed selector piece if has_selector: if not sel.tag and not ctx.is_pseudo: # Implied `*` sel.tag = ct.SelectorTag('*', None) if ctx.is_relative: sel.rel_type = ctx.rel_type selectors[-1].relations.append(sel) else: sel.relations.extend(relations) del relations[:] selectors.append(sel) # Forgive empty slots in pseudo-classes that have lists (and are forgiving) elif ctx.is_forgive and (not selectors or not relations): # Handle normal pseudo-classes with empty slots like `:is()` etc. self.increment_count() sel.no_match = True del relations[:] selectors.append(sel) has_selector = True if not has_selector: # We will always need to finish a selector when `:has()` is used as it leads with combining. # May apply to others as well. raise SelectorSyntaxError( f'Expected a selector at position {ctx.index}', self.pattern, ctx.index ) # Some patterns require additional logic, such as default. We try to make these the # last pattern, and append the appropriate flag to that selector which communicates # to the matcher what additional logic is required. if ctx.is_default: selectors[-1].flags = ct.SEL_DEFAULT if ctx.is_indeterminate: selectors[-1].flags = ct.SEL_INDETERMINATE if ctx.is_in_range: selectors[-1].flags = ct.SEL_IN_RANGE if ctx.is_out_of_range: selectors[-1].flags = ct.SEL_OUT_OF_RANGE if ctx.is_placeholder_shown: selectors[-1].flags = ct.SEL_PLACEHOLDER_SHOWN # Return selector list return ct.SelectorList([s.freeze() for s in selectors], ctx.is_not, ctx.is_html, self.count - ctx.count) def selector_iter(self, pattern: str) -> Iterator[tuple[str, Match[str]]]: """Iterate selector tokens.""" # Ignore whitespace and comments at start and end of pattern m = RE_WS_BEGIN.search(pattern) index = m.end(0) if m else 0 m = RE_WS_END.search(pattern[::-1]) offset = m.end(0) if m else 0 end = len(pattern) - (1 + offset) if self.debug: # pragma: no cover print(f'## PARSING: {pattern!r}') while index <= end: m = None for v in self.CSS_TOKENS: m = v.match(pattern, index, self.flags) if m: name = v.get_name() if self.debug: # pragma: no cover print(f"TOKEN: '{name}' --> {m.group(0)!r} at position {m.start(0)}") index = m.end(0) yield name, m break if m is None: c = pattern[index] # If the character represents the start of one of the known selector types, # throw an exception mentioning that the known selector type is in error; # otherwise, report the invalid character. if c == '[': msg = f"Malformed attribute selector at position {index}" elif c == '.': msg = f"Malformed class selector at position {index}" elif c == '#': msg = f"Malformed id selector at position {index}" elif c == ':': msg = f"Malformed pseudo-class selector at position {index}" else: msg = f"Invalid character {c!r} position {index}" raise SelectorSyntaxError(msg, self.pattern, index) if self.debug: # pragma: no cover print('## END PARSING') def process_selectors(self, index: int = 0, flags: int = 0) -> ct.SelectorList: """Process selectors.""" return self.parse_selectors(self.selector_iter(self.pattern), index, flags)