ccxt 4.2.76__py2.py3-none-any.whl → 4.4.48__py2.py3-none-any.whl

ccxt/static_dependencies/lark/parsers/cyk.py

@@ -0,0 +1,340 @@
+"""This module implements a CYK parser."""
+
+# Author: https://github.com/ehudt (2018)
+#
+# Adapted by Erez
+
+
+from collections import defaultdict
+import itertools
+
+from ..exceptions import ParseError
+from ..lexer import Token
+from ..tree import Tree
+from ..grammar import Terminal as T, NonTerminal as NT, Symbol
+
+def match(t, s):
+    assert isinstance(t, T)
+    return t.name == s.type
+
+
+class Rule:
+    """Context-free grammar rule."""
+
+    def __init__(self, lhs, rhs, weight, alias):
+        super(Rule, self).__init__()
+        assert isinstance(lhs, NT), lhs
+        assert all(isinstance(x, NT) or isinstance(x, T) for x in rhs), rhs
+        self.lhs = lhs
+        self.rhs = rhs
+        self.weight = weight
+        self.alias = alias
+
+    def __str__(self):
+        return '%s -> %s' % (str(self.lhs), ' '.join(str(x) for x in self.rhs))
+
+    def __repr__(self):
+        return str(self)
+
+    def __hash__(self):
+        return hash((self.lhs, tuple(self.rhs)))
+
+    def __eq__(self, other):
+        return self.lhs == other.lhs and self.rhs == other.rhs
+
+    def __ne__(self, other):
+        return not (self == other)
+
+
+class Grammar:
+    """Context-free grammar."""
+
+    def __init__(self, rules):
+        self.rules = frozenset(rules)
+
+    def __eq__(self, other):
+        return self.rules == other.rules
+
+    def __str__(self):
+        return '\n' + '\n'.join(sorted(repr(x) for x in self.rules)) + '\n'
+
+    def __repr__(self):
+        return str(self)
+
+
+# Parse tree data structures
+class RuleNode:
+    """A node in the parse tree, which also contains the full rhs rule."""
+
+    def __init__(self, rule, children, weight=0):
+        self.rule = rule
+        self.children = children
+        self.weight = weight
+
+    def __repr__(self):
+        return 'RuleNode(%s, [%s])' % (repr(self.rule.lhs), ', '.join(str(x) for x in self.children))
+
+
+
+class Parser:
+    """Parser wrapper."""
+
+    def __init__(self, rules):
+        super(Parser, self).__init__()
+        self.orig_rules = {rule: rule for rule in rules}
+        rules = [self._to_rule(rule) for rule in rules]
+        self.grammar = to_cnf(Grammar(rules))
+
+    def _to_rule(self, lark_rule):
+        """Converts a lark rule, (lhs, rhs, callback, options), to a Rule."""
+        assert isinstance(lark_rule.origin, NT)
+        assert all(isinstance(x, Symbol) for x in lark_rule.expansion)
+        return Rule(
+            lark_rule.origin, lark_rule.expansion,
+            weight=lark_rule.options.priority if lark_rule.options.priority else 0,
+            alias=lark_rule)
+
+    def parse(self, tokenized, start):  # pylint: disable=invalid-name
+        """Parses input, which is a list of tokens."""
+        assert start
+        start = NT(start)
+
+        table, trees = _parse(tokenized, self.grammar)
+        # Check if the parse succeeded.
+        if all(r.lhs != start for r in table[(0, len(tokenized) - 1)]):
+            raise ParseError('Parsing failed.')
+        parse = trees[(0, len(tokenized) - 1)][start]
+        return self._to_tree(revert_cnf(parse))
+
+    def _to_tree(self, rule_node):
+        """Converts a RuleNode parse tree to a lark Tree."""
+        orig_rule = self.orig_rules[rule_node.rule.alias]
+        children = []
+        for child in rule_node.children:
+            if isinstance(child, RuleNode):
+                children.append(self._to_tree(child))
+            else:
+                assert isinstance(child.name, Token)
+                children.append(child.name)
+        t = Tree(orig_rule.origin, children)
+        t.rule=orig_rule
+        return t
+
+
+def print_parse(node, indent=0):
+    if isinstance(node, RuleNode):
+        print(' ' * (indent * 2) + str(node.rule.lhs))
+        for child in node.children:
+            print_parse(child, indent + 1)
+    else:
+        print(' ' * (indent * 2) + str(node.s))
+
+
+def _parse(s, g):
+    """Parses sentence 's' using CNF grammar 'g'."""
+    # The CYK table. Indexed with a 2-tuple: (start pos, end pos)
+    table = defaultdict(set)
+    # Top-level structure is similar to the CYK table. Each cell is a dict from
+    # rule name to the best (lightest) tree for that rule.
+    trees = defaultdict(dict)
+    # Populate base case with existing terminal production rules
+    for i, w in enumerate(s):
+        for terminal, rules in g.terminal_rules.items():
+            if match(terminal, w):
+                for rule in rules:
+                    table[(i, i)].add(rule)
+                    if (rule.lhs not in trees[(i, i)] or
+                        rule.weight < trees[(i, i)][rule.lhs].weight):
+                        trees[(i, i)][rule.lhs] = RuleNode(rule, [T(w)], weight=rule.weight)
+
+    # Iterate over lengths of sub-sentences
+    for l in range(2, len(s) + 1):
+        # Iterate over sub-sentences with the given length
+        for i in range(len(s) - l + 1):
+            # Choose partition of the sub-sentence in [1, l)
+            for p in range(i + 1, i + l):
+                span1 = (i, p - 1)
+                span2 = (p, i + l - 1)
+                for r1, r2 in itertools.product(table[span1], table[span2]):
+                    for rule in g.nonterminal_rules.get((r1.lhs, r2.lhs), []):
+                        table[(i, i + l - 1)].add(rule)
+                        r1_tree = trees[span1][r1.lhs]
+                        r2_tree = trees[span2][r2.lhs]
+                        rule_total_weight = rule.weight + r1_tree.weight + r2_tree.weight
+                        if (rule.lhs not in trees[(i, i + l - 1)]
+                            or rule_total_weight < trees[(i, i + l - 1)][rule.lhs].weight):
+                            trees[(i, i + l - 1)][rule.lhs] = RuleNode(rule, [r1_tree, r2_tree], weight=rule_total_weight)
+    return table, trees
+
+
+# This section implements context-free grammar converter to Chomsky normal form.
+# It also implements a conversion of parse trees from its CNF to the original
+# grammar.
+# Overview:
+# Applies the following operations in this order:
+# * TERM: Eliminates non-solitary terminals from all rules
+# * BIN: Eliminates rules with more than 2 symbols on their right-hand-side.
+# * UNIT: Eliminates non-terminal unit rules
+#
+# The following grammar characteristics aren't featured:
+# * Start symbol appears on RHS
+# * Empty rules (epsilon rules)
+
+
+class CnfWrapper:
+    """CNF wrapper for grammar.
+
+  Validates that the input grammar is CNF and provides helper data structures.
+  """
+
+    def __init__(self, grammar):
+        super(CnfWrapper, self).__init__()
+        self.grammar = grammar
+        self.rules = grammar.rules
+        self.terminal_rules = defaultdict(list)
+        self.nonterminal_rules = defaultdict(list)
+        for r in self.rules:
+            # Validate that the grammar is CNF and populate auxiliary data structures.
+            assert isinstance(r.lhs, NT), r
+            if len(r.rhs) not in [1, 2]:
+                raise ParseError("CYK doesn't support empty rules")
+            if len(r.rhs) == 1 and isinstance(r.rhs[0], T):
+                self.terminal_rules[r.rhs[0]].append(r)
+            elif len(r.rhs) == 2 and all(isinstance(x, NT) for x in r.rhs):
+                self.nonterminal_rules[tuple(r.rhs)].append(r)
+            else:
+                assert False, r
+
+    def __eq__(self, other):
+        return self.grammar == other.grammar
+
+    def __repr__(self):
+        return repr(self.grammar)
+
+
+class UnitSkipRule(Rule):
+    """A rule that records NTs that were skipped during transformation."""
+
+    def __init__(self, lhs, rhs, skipped_rules, weight, alias):
+        super(UnitSkipRule, self).__init__(lhs, rhs, weight, alias)
+        self.skipped_rules = skipped_rules
+
+    def __eq__(self, other):
+        return isinstance(other, type(self)) and self.skipped_rules == other.skipped_rules
+
+    __hash__ = Rule.__hash__
+
+
+def build_unit_skiprule(unit_rule, target_rule):
+    skipped_rules = []
+    if isinstance(unit_rule, UnitSkipRule):
+        skipped_rules += unit_rule.skipped_rules
+    skipped_rules.append(target_rule)
+    if isinstance(target_rule, UnitSkipRule):
+        skipped_rules += target_rule.skipped_rules
+    return UnitSkipRule(unit_rule.lhs, target_rule.rhs, skipped_rules,
+                      weight=unit_rule.weight + target_rule.weight, alias=unit_rule.alias)
+
+
+def get_any_nt_unit_rule(g):
+    """Returns a non-terminal unit rule from 'g', or None if there is none."""
+    for rule in g.rules:
+        if len(rule.rhs) == 1 and isinstance(rule.rhs[0], NT):
+            return rule
+    return None
+
+
+def _remove_unit_rule(g, rule):
+    """Removes 'rule' from 'g' without changing the language produced by 'g'."""
+    new_rules = [x for x in g.rules if x != rule]
+    refs = [x for x in g.rules if x.lhs == rule.rhs[0]]
+    new_rules += [build_unit_skiprule(rule, ref) for ref in refs]
+    return Grammar(new_rules)
+
+
+def _split(rule):
+    """Splits a rule whose len(rhs) > 2 into shorter rules."""
+    rule_str = str(rule.lhs) + '__' + '_'.join(str(x) for x in rule.rhs)
+    rule_name = '__SP_%s' % (rule_str) + '_%d'
+    yield Rule(rule.lhs, [rule.rhs[0], NT(rule_name % 1)], weight=rule.weight, alias=rule.alias)
+    for i in range(1, len(rule.rhs) - 2):
+        yield Rule(NT(rule_name % i), [rule.rhs[i], NT(rule_name % (i + 1))], weight=0, alias='Split')
+    yield Rule(NT(rule_name % (len(rule.rhs) - 2)), rule.rhs[-2:], weight=0, alias='Split')
+
+
+def _term(g):
+    """Applies the TERM rule on 'g' (see top comment)."""
+    all_t = {x for rule in g.rules for x in rule.rhs if isinstance(x, T)}
+    t_rules = {t: Rule(NT('__T_%s' % str(t)), [t], weight=0, alias='Term') for t in all_t}
+    new_rules = []
+    for rule in g.rules:
+        if len(rule.rhs) > 1 and any(isinstance(x, T) for x in rule.rhs):
+            new_rhs = [t_rules[x].lhs if isinstance(x, T) else x for x in rule.rhs]
+            new_rules.append(Rule(rule.lhs, new_rhs, weight=rule.weight, alias=rule.alias))
+            new_rules.extend(v for k, v in t_rules.items() if k in rule.rhs)
+        else:
+            new_rules.append(rule)
+    return Grammar(new_rules)
+
+
+def _bin(g):
+    """Applies the BIN rule to 'g' (see top comment)."""
+    new_rules = []
+    for rule in g.rules:
+        if len(rule.rhs) > 2:
+            new_rules += _split(rule)
+        else:
+            new_rules.append(rule)
+    return Grammar(new_rules)
+
+
+def _unit(g):
+    """Applies the UNIT rule to 'g' (see top comment)."""
+    nt_unit_rule = get_any_nt_unit_rule(g)
+    while nt_unit_rule:
+        g = _remove_unit_rule(g, nt_unit_rule)
+        nt_unit_rule = get_any_nt_unit_rule(g)
+    return g
+
+
+def to_cnf(g):
+    """Creates a CNF grammar from a general context-free grammar 'g'."""
+    g = _unit(_bin(_term(g)))
+    return CnfWrapper(g)
+
+
+def unroll_unit_skiprule(lhs, orig_rhs, skipped_rules, children, weight, alias):
+    if not skipped_rules:
+        return RuleNode(Rule(lhs, orig_rhs, weight=weight, alias=alias), children, weight=weight)
+    else:
+        weight = weight - skipped_rules[0].weight
+        return RuleNode(
+            Rule(lhs, [skipped_rules[0].lhs], weight=weight, alias=alias), [
+                unroll_unit_skiprule(skipped_rules[0].lhs, orig_rhs,
+                                skipped_rules[1:], children,
+                                skipped_rules[0].weight, skipped_rules[0].alias)
+            ], weight=weight)
+
+
+def revert_cnf(node):
+    """Reverts a parse tree (RuleNode) to its original non-CNF form (Node)."""
+    if isinstance(node, T):
+        return node
+    # Reverts TERM rule.
+    if node.rule.lhs.name.startswith('__T_'):
+        return node.children[0]
+    else:
+        children = []
+        for child in map(revert_cnf, node.children):
+            # Reverts BIN rule.
+            if isinstance(child, RuleNode) and child.rule.lhs.name.startswith('__SP_'):
+                children += child.children
+            else:
+                children.append(child)
+        # Reverts UNIT rule.
+        if isinstance(node.rule, UnitSkipRule):
+            return unroll_unit_skiprule(node.rule.lhs, node.rule.rhs,
+                                    node.rule.skipped_rules, children,
+                                    node.rule.weight, node.rule.alias)
+        else:
+            return RuleNode(node.rule, children)

ccxt/static_dependencies/lark/parsers/earley.py

@@ -0,0 +1,314 @@
+"""This module implements an Earley parser.
+
+The core Earley algorithm used here is based on Elizabeth Scott's implementation, here:
+    https://www.sciencedirect.com/science/article/pii/S1571066108001497
+
+That is probably the best reference for understanding the algorithm here.
+
+The Earley parser outputs an SPPF-tree as per that document. The SPPF tree format
+is explained here: https://lark-parser.readthedocs.io/en/latest/_static/sppf/sppf.html
+"""
+
+from typing import TYPE_CHECKING, Callable, Optional, List, Any
+from collections import deque
+
+from ..lexer import Token
+from ..tree import Tree
+from ..exceptions import UnexpectedEOF, UnexpectedToken
+from ..utils import logger, OrderedSet, dedup_list
+from .grammar_analysis import GrammarAnalyzer
+from ..grammar import NonTerminal
+from .earley_common import Item
+from .earley_forest import ForestSumVisitor, SymbolNode, StableSymbolNode, TokenNode, ForestToParseTree
+
+if TYPE_CHECKING:
+    from ..common import LexerConf, ParserConf
+
+class Parser:
+    lexer_conf: 'LexerConf'
+    parser_conf: 'ParserConf'
+    debug: bool
+
+    def __init__(self, lexer_conf: 'LexerConf', parser_conf: 'ParserConf', term_matcher: Callable,
+                 resolve_ambiguity: bool=True, debug: bool=False,
+                 tree_class: Optional[Callable[[str, List], Any]]=Tree, ordered_sets: bool=True):
+        analysis = GrammarAnalyzer(parser_conf)
+        self.lexer_conf = lexer_conf
+        self.parser_conf = parser_conf
+        self.resolve_ambiguity = resolve_ambiguity
+        self.debug = debug
+        self.Tree = tree_class
+        self.Set = OrderedSet if ordered_sets else set
+        self.SymbolNode = StableSymbolNode if ordered_sets else SymbolNode
+
+        self.FIRST = analysis.FIRST
+        self.NULLABLE = analysis.NULLABLE
+        self.callbacks = parser_conf.callbacks
+        # TODO add typing info
+        self.predictions = {}   # type: ignore[var-annotated]
+
+        ## These could be moved to the grammar analyzer. Pre-computing these is *much* faster than
+        #  the slow 'isupper' in is_terminal.
+        self.TERMINALS = { sym for r in parser_conf.rules for sym in r.expansion if sym.is_term }
+        self.NON_TERMINALS = { sym for r in parser_conf.rules for sym in r.expansion if not sym.is_term }
+
+        self.forest_sum_visitor = None
+        for rule in parser_conf.rules:
+            if rule.origin not in self.predictions:
+                self.predictions[rule.origin] = [x.rule for x in analysis.expand_rule(rule.origin)]
+
+            ## Detect if any rules/terminals have priorities set. If the user specified priority = None, then
+            #  the priorities will be stripped from all rules/terminals before they reach us, allowing us to
+            #  skip the extra tree walk. We'll also skip this if the user just didn't specify priorities
+            #  on any rules/terminals.
+            if self.forest_sum_visitor is None and rule.options.priority is not None:
+                self.forest_sum_visitor = ForestSumVisitor
+
+        # Check terminals for priorities
+        # Ignore terminal priorities if the basic lexer is used
+        if self.lexer_conf.lexer_type != 'basic' and self.forest_sum_visitor is None:
+            for term in self.lexer_conf.terminals:
+                if term.priority:
+                    self.forest_sum_visitor = ForestSumVisitor
+                    break
+
+        self.term_matcher = term_matcher
+
+
+    def predict_and_complete(self, i, to_scan, columns, transitives):
+        """The core Earley Predictor and Completer.
+
+        At each stage of the input, we handling any completed items (things
+        that matched on the last cycle) and use those to predict what should
+        come next in the input stream. The completions and any predicted
+        non-terminals are recursively processed until we reach a set of,
+        which can be added to the scan list for the next scanner cycle."""
+        # Held Completions (H in E.Scotts paper).
+        node_cache = {}
+        held_completions = {}
+
+        column = columns[i]
+        # R (items) = Ei (column.items)
+        items = deque(column)
+        while items:
+            item = items.pop()    # remove an element, A say, from R
+
+            ### The Earley completer
+            if item.is_complete:   ### (item.s == string)
+                if item.node is None:
+                    label = (item.s, item.start, i)
+                    item.node = node_cache[label] if label in node_cache else node_cache.setdefault(label, self.SymbolNode(*label))
+                    item.node.add_family(item.s, item.rule, item.start, None, None)
+
+                # create_leo_transitives(item.rule.origin, item.start)
+
+                ###R Joop Leo right recursion Completer
+                if item.rule.origin in transitives[item.start]:
+                    transitive = transitives[item.start][item.s]
+                    if transitive.previous in transitives[transitive.column]:
+                        root_transitive = transitives[transitive.column][transitive.previous]
+                    else:
+                        root_transitive = transitive
+
+                    new_item = Item(transitive.rule, transitive.ptr, transitive.start)
+                    label = (root_transitive.s, root_transitive.start, i)
+                    new_item.node = node_cache[label] if label in node_cache else node_cache.setdefault(label, self.SymbolNode(*label))
+                    new_item.node.add_path(root_transitive, item.node)
+                    if new_item.expect in self.TERMINALS:
+                        # Add (B :: aC.B, h, y) to Q
+                        to_scan.add(new_item)
+                    elif new_item not in column:
+                        # Add (B :: aC.B, h, y) to Ei and R
+                        column.add(new_item)
+                        items.append(new_item)
+                ###R Regular Earley completer
+                else:
+                    # Empty has 0 length. If we complete an empty symbol in a particular
+                    # parse step, we need to be able to use that same empty symbol to complete
+                    # any predictions that result, that themselves require empty. Avoids
+                    # infinite recursion on empty symbols.
+                    # held_completions is 'H' in E.Scott's paper.
+                    is_empty_item = item.start == i
+                    if is_empty_item:
+                        held_completions[item.rule.origin] = item.node
+
+                    originators = [originator for originator in columns[item.start] if originator.expect is not None and originator.expect == item.s]
+                    for originator in originators:
+                        new_item = originator.advance()
+                        label = (new_item.s, originator.start, i)
+                        new_item.node = node_cache[label] if label in node_cache else node_cache.setdefault(label, self.SymbolNode(*label))
+                        new_item.node.add_family(new_item.s, new_item.rule, i, originator.node, item.node)
+                        if new_item.expect in self.TERMINALS:
+                            # Add (B :: aC.B, h, y) to Q
+                            to_scan.add(new_item)
+                        elif new_item not in column:
+                            # Add (B :: aC.B, h, y) to Ei and R
+                            column.add(new_item)
+                            items.append(new_item)
+
+            ### The Earley predictor
+            elif item.expect in self.NON_TERMINALS: ### (item.s == lr0)
+                new_items = []
+                for rule in self.predictions[item.expect]:
+                    new_item = Item(rule, 0, i)
+                    new_items.append(new_item)
+
+                # Process any held completions (H).
+                if item.expect in held_completions:
+                    new_item = item.advance()
+                    label = (new_item.s, item.start, i)
+                    new_item.node = node_cache[label] if label in node_cache else node_cache.setdefault(label, self.SymbolNode(*label))
+                    new_item.node.add_family(new_item.s, new_item.rule, new_item.start, item.node, held_completions[item.expect])
+                    new_items.append(new_item)
+
+                for new_item in new_items:
+                    if new_item.expect in self.TERMINALS:
+                        to_scan.add(new_item)
+                    elif new_item not in column:
+                        column.add(new_item)
+                        items.append(new_item)
+
+    def _parse(self, lexer, columns, to_scan, start_symbol=None):
+
+        def is_quasi_complete(item):
+            if item.is_complete:
+                return True
+
+            quasi = item.advance()
+            while not quasi.is_complete:
+                if quasi.expect not in self.NULLABLE:
+                    return False
+                if quasi.rule.origin == start_symbol and quasi.expect == start_symbol:
+                    return False
+                quasi = quasi.advance()
+            return True
+
+        # def create_leo_transitives(origin, start):
+        #   ...   # removed at commit 4c1cfb2faf24e8f8bff7112627a00b94d261b420
+
+        def scan(i, token, to_scan):
+            """The core Earley Scanner.
+
+            This is a custom implementation of the scanner that uses the
+            Lark lexer to match tokens. The scan list is built by the
+            Earley predictor, based on the previously completed tokens.
+            This ensures that at each phase of the parse we have a custom
+            lexer context, allowing for more complex ambiguities."""
+            next_to_scan = self.Set()
+            next_set = self.Set()
+            columns.append(next_set)
+            transitives.append({})
+            node_cache = {}
+
+            for item in self.Set(to_scan):
+                if match(item.expect, token):
+                    new_item = item.advance()
+                    label = (new_item.s, new_item.start, i)
+                    # 'terminals' may not contain token.type when using %declare
+                    # Additionally, token is not always a Token
+                    # For example, it can be a Tree when using TreeMatcher
+                    term = terminals.get(token.type) if isinstance(token, Token) else None
+                    # Set the priority of the token node to 0 so that the
+                    # terminal priorities do not affect the Tree chosen by
+                    # ForestSumVisitor after the basic lexer has already
+                    # "used up" the terminal priorities
+                    token_node = TokenNode(token, term, priority=0)
+                    new_item.node = node_cache[label] if label in node_cache else node_cache.setdefault(label, self.SymbolNode(*label))
+                    new_item.node.add_family(new_item.s, item.rule, new_item.start, item.node, token_node)
+
+                    if new_item.expect in self.TERMINALS:
+                        # add (B ::= Aai+1.B, h, y) to Q'
+                        next_to_scan.add(new_item)
+                    else:
+                        # add (B ::= Aa+1.B, h, y) to Ei+1
+                        next_set.add(new_item)
+
+            if not next_set and not next_to_scan:
+                expect = {i.expect.name for i in to_scan}
+                raise UnexpectedToken(token, expect, considered_rules=set(to_scan), state=frozenset(i.s for i in to_scan))
+
+            return next_to_scan
+
+
+        # Define parser functions
+        match = self.term_matcher
+
+        terminals = self.lexer_conf.terminals_by_name
+
+        # Cache for nodes & tokens created in a particular parse step.
+        transitives = [{}]
+
+        ## The main Earley loop.
+        # Run the Prediction/Completion cycle for any Items in the current Earley set.
+        # Completions will be added to the SPPF tree, and predictions will be recursively
+        # processed down to terminals/empty nodes to be added to the scanner for the next
+        # step.
+        expects = {i.expect for i in to_scan}
+        i = 0
+        for token in lexer.lex(expects):
+            self.predict_and_complete(i, to_scan, columns, transitives)
+
+            to_scan = scan(i, token, to_scan)
+            i += 1
+
+            expects.clear()
+            expects |= {i.expect for i in to_scan}
+
+        self.predict_and_complete(i, to_scan, columns, transitives)
+
+        ## Column is now the final column in the parse.
+        assert i == len(columns)-1
+        return to_scan
+
+    def parse(self, lexer, start):
+        assert start, start
+        start_symbol = NonTerminal(start)
+
+        columns = [self.Set()]
+        to_scan = self.Set()     # The scan buffer. 'Q' in E.Scott's paper.
+
+        ## Predict for the start_symbol.
+        # Add predicted items to the first Earley set (for the predictor) if they
+        # result in a non-terminal, or the scanner if they result in a terminal.
+        for rule in self.predictions[start_symbol]:
+            item = Item(rule, 0, 0)
+            if item.expect in self.TERMINALS:
+                to_scan.add(item)
+            else:
+                columns[0].add(item)
+
+        to_scan = self._parse(lexer, columns, to_scan, start_symbol)
+
+        # If the parse was successful, the start
+        # symbol should have been completed in the last step of the Earley cycle, and will be in
+        # this column. Find the item for the start_symbol, which is the root of the SPPF tree.
+        solutions = dedup_list(n.node for n in columns[-1] if n.is_complete and n.node is not None and n.s == start_symbol and n.start == 0)
+        if not solutions:
+            expected_terminals = [t.expect.name for t in to_scan]
+            raise UnexpectedEOF(expected_terminals, state=frozenset(i.s for i in to_scan))
+
+        if self.debug:
+            from .earley_forest import ForestToPyDotVisitor
+            try:
+                debug_walker = ForestToPyDotVisitor()
+            except ImportError:
+                logger.warning("Cannot find dependency 'pydot', will not generate sppf debug image")
+            else:
+                for i, s in enumerate(solutions):
+                    debug_walker.visit(s, f"sppf{i}.png")
+
+
+        if self.Tree is not None:
+            # Perform our SPPF -> AST conversion
+            transformer = ForestToParseTree(self.Tree, self.callbacks, self.forest_sum_visitor and self.forest_sum_visitor(), self.resolve_ambiguity)
+            solutions = [transformer.transform(s) for s in solutions]
+
+            if len(solutions) > 1:
+                t: Tree = self.Tree('_ambig', solutions)
+                t.expand_kids_by_data('_ambig')     # solutions may themselves be _ambig nodes
+                return t
+            return solutions[0]
+
+        # return the root of the SPPF
+        # TODO return a list of solutions, or join them together somehow
+        return solutions[0]

ccxt/static_dependencies/lark/parsers/earley_common.py

@@ -0,0 +1,42 @@
+"""This module implements useful building blocks for the Earley parser
+"""
+
+
+class Item:
+    "An Earley Item, the atom of the algorithm."
+
+    __slots__ = ('s', 'rule', 'ptr', 'start', 'is_complete', 'expect', 'previous', 'node', '_hash')
+    def __init__(self, rule, ptr, start):
+        self.is_complete = len(rule.expansion) == ptr
+        self.rule = rule    # rule
+        self.ptr = ptr      # ptr
+        self.start = start  # j
+        self.node = None    # w
+        if self.is_complete:
+            self.s = rule.origin
+            self.expect = None
+            self.previous = rule.expansion[ptr - 1] if ptr > 0 and len(rule.expansion) else None
+        else:
+            self.s = (rule, ptr)
+            self.expect = rule.expansion[ptr]
+            self.previous = rule.expansion[ptr - 1] if ptr > 0 and len(rule.expansion) else None
+        self._hash = hash((self.s, self.start, self.rule))
+
+    def advance(self):
+        return Item(self.rule, self.ptr + 1, self.start)
+
+    def __eq__(self, other):
+        return self is other or (self.s == other.s and self.start == other.start and self.rule == other.rule)
+
+    def __hash__(self):
+        return self._hash
+
+    def __repr__(self):
+        before = ( expansion.name for expansion in self.rule.expansion[:self.ptr] )
+        after = ( expansion.name for expansion in self.rule.expansion[self.ptr:] )
+        symbol = "{} ::= {}* {}".format(self.rule.origin.name, ' '.join(before), ' '.join(after))
+        return '%s (%d)' % (symbol, self.start)
+
+
+# class TransitiveItem(Item):
+#   ...   # removed at commit 4c1cfb2faf24e8f8bff7112627a00b94d261b420