qnet.algebra.pattern_matching package¶
QNET’s pattern matching engine.
Patterns may be constructed by either instantiating a Pattern instance
directly, or (preferred) by calling the pattern(), pattern_head(),
or wc() helper routines.
The pattern may then be matched against an expression using
match_pattern(). The result of a match is a MatchDict object,
which evaluates to True or False in a boolean context to indicate the success
or failure of the match (or alternatively, through the success attribute).
The MatchDict object also maps any wildcard names to the expression
that the corresponding wildcard Pattern matches.
Summary¶
__all__ Classes:
MatchDict |
Result of a Pattern.match() |
Pattern |
Pattern for matching an expression |
Private Classes:
ProtoExpr |
Object representing an un-instantiated Expression |
__all__ Functions:
match_pattern |
Recursively match expr with the given expr_or_pattern |
pattern |
‘Flat’ constructor for the Pattern class |
pattern_head |
Constructor for a Pattern matching a ProtoExpr |
wc |
Constructor for a wildcard-Pattern |
Reference¶
-
class
qnet.algebra.pattern_matching.MatchDict(*args)[source]¶ Bases:
collections.OrderedDictResult of a
Pattern.match()Dictionary of wildcard names to expressions. Once the value for a key is set, attempting to set it again with a different value raises a
KeyError. The attribute merge_lists may be set to modify this behavior for values that are lists: If it is set to a value different from zero, two lists that are set via the same key are merged. If merge_lists is negative, the new values are appended to the existing values; if it is positive, the new values are prepended.In a boolean context, a
MatchDictalways evaluates as True (even if empty, unlike a normal dictionary), unless the success attribute is explicitly set to False (which a failedPattern.match()should do)Attributes: - success (bool) – Value of the
MatchDictobject in a boolean context:bool(match) == match.success - reason (str) – If success is False, string explaining why the match failed
- merge_lists (int) – Code that indicates how to combine multiple values that are lists
- success (bool) – Value of the
-
class
qnet.algebra.pattern_matching.Pattern(head=None, args=None, kwargs=None, *, mode=1, wc_name=None, conditions=None)[source]¶ Bases:
objectPattern for matching an expression
Parameters: - head (type or None) – The type (or tuple of types) of the expression that can be matched. If None, any type of Expression matches
- args (list or None) – List or tuple of positional arguments of the matched Expression (cf. Expression.args). Each element is an expression (to be matched exactly) or another Pattern instance (matched recursively). If None, no arguments are checked
- kwargs (dict or None) – Dictionary of keyword arguments of the expression (cf. Expression.kwargs). As for args, each value is an expression or Pattern instance.
- mode (int) – If the pattern is used to match the arguments of an expression, code to indicate how many arguments the Pattern can consume: Pattern.single, Pattern.one_or_more, Pattern.zero_or_more
- wc_name (str or None) – If pattern matches an expression, key in the
resulting
MatchDictfor the expression. If None, the match will not be recorded in the result - conditions (list of callables, or None) – If not None, a list of callables that take expr and return a boolean value. If the return value is False, the pattern is determined not to match expr.
Note
For (sub-)patterns that occur nested in the args attribute of another pattern, only the first or last sub-pattern may have a mode other than Pattern.single. This also implies that only one of the args may have a mode other than Pattern.single. This restrictions ensures that patterns can be matched without backtracking, thus guaranteeing numerical efficiency.
Example
Consider the following nested circuit expression:
>>> C1 = CircuitSymbol('C1', cdim=3) >>> C2 = CircuitSymbol('C2', cdim=3) >>> C3 = CircuitSymbol('C3', cdim=3) >>> C4 = CircuitSymbol('C4', cdim=3) >>> perm1 = CPermutation((2, 1, 0)) >>> perm2 = CPermutation((0, 2, 1)) >>> concat_expr = Concatenation( ... (C1 << C2 << perm1), ... (C3 << C4 << perm2))
We may match this with the following pattern:
>>> conditions = [lambda c: c.cdim == 3, ... lambda c: c.label[0] == 'C'] >>> A__Circuit = wc("A__", head=CircuitSymbol, ... conditions=conditions) >>> C__Circuit = wc("C__", head=CircuitSymbol, ... conditions=conditions) >>> B_CPermutation = wc("B", head=CPermutation) >>> D_CPermutation = wc("D", head=CPermutation) >>> pattern_concat = pattern( ... Concatenation, ... pattern(SeriesProduct, A__Circuit, B_CPermutation), ... pattern(SeriesProduct, C__Circuit, D_CPermutation)) >>> m = pattern_concat.match(concat_expr)
The match returns the following dictionary:
>>> result = {'A': [C1, C2], 'B': perm1, 'C': [C3, C4], 'D': perm2} >>> assert m == result
-
single= 1¶
-
one_or_more= 2¶
-
zero_or_more= 3¶
-
extended_arg_patterns()[source]¶ Iterator over patterns for positional arguments to be matched
This yields the elements of
args, extended by their mode value
-
match(expr)[source]¶ Match the given expression (recursively)
Returns a
MatchDictinstance that maps any wildcard names to the expressions that the corresponding wildcard pattern matches. For (sub-)pattern that have a mode attribute other than Pattern.single, the wildcard name is mapped to a list of all matched expression.If the match is successful, the resulting
MatchDictinstance will evaluate to True in a boolean context. If the match is not successful, it will evaluate as False, and the reason for failure is available in the reason attribute of theMatchDictobject.Return type: MatchDict
-
findall(expr)[source]¶ list of all matching (sub-)expressions in expr
See also
finditer()yields the matches (MatchDictinstances) for the matched expressions.
-
finditer(expr)[source]¶ Return an iterator over all matches in expr
Iterate over all
MatchDictresults of matches for any matching (sub-)expressions in expr. The order of the matches conforms to the equivalent matched expressions returned byfindall().
-
wc_names¶ Set of all wildcard names occurring in the pattern
-
qnet.algebra.pattern_matching.pattern(head, *args, mode=1, wc_name=None, conditions=None, **kwargs)[source]¶ ‘Flat’ constructor for the Pattern class
Positional and keyword arguments are mapped into args and kwargs, respectively. Useful for defining rules that match an instantiated Expression with specific arguments
Return type: Pattern
-
qnet.algebra.pattern_matching.pattern_head(*args, conditions=None, wc_name=None, **kwargs)[source]¶ Constructor for a
Patternmatching aProtoExprThe patterns associated with
_rulesand_binary_rulesof anExpressionsubclass, or those passed toExpression.add_rule(), must be instantiated through this routine. The function does not allow to set a wildcard name (wc_name must not be given / be None)Return type: Pattern
-
qnet.algebra.pattern_matching.wc(name_mode='_', head=None, args=None, kwargs=None, *, conditions=None)[source]¶ Constructor for a wildcard-
PatternHelper function to create a Pattern object with an emphasis on wildcard patterns, if we don’t care about the arguments of the matched expressions (otherwise, use
pattern())Parameters: The name_mode argument uses trailing underscored to indicate the mode:
A->Pattern(wc_name="A", mode=Pattern.single, ...)A_->Pattern(wc_name="A", mode=Pattern.single, ...)B__->Pattern(wc_name="B", mode=Pattern.one_or_more, ...)B___->Pattern(wc_name="C", mode=Pattern.zero_or_more, ...)
Return type: Pattern
-
class
qnet.algebra.pattern_matching.ProtoExpr(args, kwargs, cls=None)[source]¶ Bases:
collections.abc.SequenceObject representing an un-instantiated
ExpressionA
ProtoExprmay be matched by aPatterncreated viapattern_head(). This is used inExpression.create(): before an expression is instantiated, aProtoExpris constructed with the positional and keyword arguments passed tocreate(). Then, thisProtoExpris matched against all the automatic rulescreate()knows about.Parameters: - args (list) – positional arguments that would be used in the instantiation of the Expression
- kwargs (dict) – keyword arguments. Will we converted to an
OrderedDict - cls (class or None) – The class of the Expression that will ultimately be instantiated.
The combined values of args and kwargs are accessible as a (mutable) sequence.