diff options
Diffstat (limited to 'bsfs/query')
| -rw-r--r-- | bsfs/query/ast/filter_.py | 59 | ||||
| -rw-r--r-- | bsfs/query/validator.py | 90 |
2 files changed, 93 insertions, 56 deletions
diff --git a/bsfs/query/ast/filter_.py b/bsfs/query/ast/filter_.py index b129ded..2f0270c 100644 --- a/bsfs/query/ast/filter_.py +++ b/bsfs/query/ast/filter_.py @@ -252,8 +252,7 @@ class Has(FilterExpression): class _Value(FilterExpression): - """ - """ + """Matches some value.""" # target value. value: typing.Any @@ -277,13 +276,13 @@ class Is(_Value): class Equals(_Value): """Value matches exactly. - NOTE: Value format must correspond to literal type; can be a string, a number, or a Node + NOTE: Value must correspond to literal type. """ class Substring(_Value): """Value matches a substring - NOTE: value format must be a string + NOTE: value must be a string. """ @@ -295,9 +294,49 @@ class EndsWith(_Value): """Value ends with a given string.""" +class Distance(FilterExpression): + """Distance to a reference is (strictly) below a threshold. Assumes a Feature literal.""" + + # FIXME: + # (a) pass a node/predicate as anchor instead of a value. + # Then we don't need to materialize the reference. + # (b) pass a FilterExpression (_Bounded) instead of a threshold. + # Then, we could also query values greater than a threshold. + + # reference value. + reference: typing.Any + + # distance threshold. + threshold: float + + # closed (True) or open (False) bound. + strict: bool + + def __init__( + self, + reference: typing.Any, + threshold: float, + strict: bool = False, + ): + self.reference = reference + self.threshold = float(threshold) + self.strict = bool(strict) + + def __repr__(self) -> str: + return f'{typename(self)}({self.reference}, {self.threshold}, {self.strict})' + + def __hash__(self) -> int: + return hash((super().__hash__(), tuple(self.reference), self.threshold, self.strict)) + + def __eq__(self, other) -> bool: + return super().__eq__(other) \ + and self.reference == other.reference \ + and self.threshold == other.threshold \ + and self.strict == other.strict + + class _Bounded(FilterExpression): - """ - """ + """Value is bounded by a threshold. Assumes a Number literal.""" # bound. threshold: float @@ -327,15 +366,11 @@ class _Bounded(FilterExpression): class LessThan(_Bounded): - """Value is (strictly) smaller than threshold. - NOTE: only on numerical literals - """ + """Value is (strictly) smaller than threshold. Assumes a Number literal.""" class GreaterThan(_Bounded): - """Value is (strictly) larger than threshold - NOTE: only on numerical literals - """ + """Value is (strictly) larger than threshold. Assumes a Number literal.""" class Predicate(PredicateExpression): diff --git a/bsfs/query/validator.py b/bsfs/query/validator.py index 352203a..904ac14 100644 --- a/bsfs/query/validator.py +++ b/bsfs/query/validator.py @@ -34,9 +34,6 @@ class Filter(): """ - # vertex types - T_VERTEX = typing.Union[bsc.Node, bsc.Literal] # FIXME: Shouldn't this be in the schema? - # schema to validate against. schema: bsc.Schema @@ -64,7 +61,7 @@ class Filter(): ## routing methods - def _parse_filter_expression(self, type_: T_VERTEX, node: ast.filter.FilterExpression): + def _parse_filter_expression(self, type_: bsc.Vertex, node: ast.filter.FilterExpression): """Route *node* to the handler of the respective FilterExpression subclass.""" if isinstance(node, ast.filter.Is): return self._is(type_, node) @@ -72,6 +69,8 @@ class Filter(): return self._not(type_, node) if isinstance(node, ast.filter.Has): return self._has(type_, node) + if isinstance(node, ast.filter.Distance): + return self._distance(type_, node) if isinstance(node, (ast.filter.Any, ast.filter.All)): return self._branch(type_, node) if isinstance(node, (ast.filter.And, ast.filter.Or)): @@ -83,7 +82,7 @@ class Filter(): # invalid node raise errors.BackendError(f'expected filter expression, found {node}') - def _parse_predicate_expression(self, node: ast.filter.PredicateExpression) -> typing.Tuple[T_VERTEX, T_VERTEX]: + def _parse_predicate_expression(self, node: ast.filter.PredicateExpression) -> typing.Tuple[bsc.Vertex, bsc.Vertex]: """Route *node* to the handler of the respective PredicateExpression subclass.""" if isinstance(node, ast.filter.Predicate): return self._predicate(node) @@ -95,58 +94,47 @@ class Filter(): ## predicate expressions - def _predicate(self, node: ast.filter.Predicate) -> typing.Tuple[T_VERTEX, T_VERTEX]: + def _predicate(self, node: ast.filter.Predicate) -> typing.Tuple[bsc.Vertex, bsc.Vertex]: # predicate exists in the schema if not self.schema.has_predicate(node.predicate): raise errors.ConsistencyError(f'predicate {node.predicate} is not in the schema') # determine domain and range pred = self.schema.predicate(node.predicate) + if not isinstance(pred.range, (bsc.Node, bsc.Literal)): + raise errors.BackendError(f'the range of predicate {pred} is undefined') dom, rng = pred.domain, pred.range - if rng is None: - # FIXME: It is a design error that Predicates can have a None range... - raise errors.BackendError(f'predicate {pred} has no range') if node.reverse: dom, rng = rng, dom # type: ignore [assignment] # variable re-use confuses mypy # return domain and range return dom, rng - def _one_of(self, node: ast.filter.OneOf) -> typing.Tuple[T_VERTEX, T_VERTEX]: + def _one_of(self, node: ast.filter.OneOf) -> typing.Tuple[bsc.Vertex, bsc.Vertex]: # determine domain and range types # NOTE: select the most specific domain and the most generic range dom, rng = None, None for pred in node: # parse child expression subdom, subrng = self._parse_predicate_expression(pred) - try: - # determine overall domain - if dom is None or subdom < dom: # pick most specific domain - dom = subdom - # domains must be related across all child expressions - if not subdom <= dom and not subdom >= dom: - raise errors.ConsistencyError(f'domains {subdom} and {dom} are not related') - except TypeError as err: # compared literal vs. node - raise errors.ConsistencyError(f'domains {subdom} and {dom} are not of the same type') from err - - try: - # determine overall range - if rng is None or subrng > rng: # pick most generic range - rng = subrng - # ranges must be related across all child expressions - if not subrng <= rng and not subrng >= rng: - raise errors.ConsistencyError(f'ranges {subrng} and {rng} are not related') - except TypeError as err: # compared literal vs. node - raise errors.ConsistencyError(f'ranges {subrng} and {rng} are not of the same type') from err - # check domain and range - if dom is None or rng is None: - # OneOf guarantees at least one expression, these two cases cannot happen - raise errors.UnreachableError() - # return domain and range - return dom, rng + # determine overall domain + if dom is None or subdom < dom: # pick most specific domain + dom = subdom + # domains must be related across all child expressions + if not subdom <= dom and not subdom >= dom: + raise errors.ConsistencyError(f'domains {subdom} and {dom} are not related') + # determine overall range + if rng is None or subrng > rng: # pick most generic range + rng = subrng + # ranges must be related across all child expressions + if not subrng <= rng and not subrng >= rng: + raise errors.ConsistencyError(f'ranges {subrng} and {rng} are not related') + # OneOf guarantees at least one expression, dom and rng are always bsc.Vertex. + # mypy does not realize this, hence we ignore the warning. + return dom, rng # type: ignore [return-value] ## intermediates - def _branch(self, type_: T_VERTEX, node: ast.filter._Branch): + def _branch(self, type_: bsc.Vertex, node: ast.filter._Branch): # type is a Node if not isinstance(type_, bsc.Node): raise errors.ConsistencyError(f'expected a Node, found {type_}') @@ -167,16 +155,16 @@ class Filter(): # child expression is valid self._parse_filter_expression(rng, node.expr) - def _agg(self, type_: T_VERTEX, node: ast.filter._Agg): + def _agg(self, type_: bsc.Vertex, node: ast.filter._Agg): for expr in node: # child expression is valid self._parse_filter_expression(type_, expr) - def _not(self, type_: T_VERTEX, node: ast.filter.Not): + def _not(self, type_: bsc.Vertex, node: ast.filter.Not): # child expression is valid self._parse_filter_expression(type_, node.expr) - def _has(self, type_: T_VERTEX, node: ast.filter.Has): + def _has(self, type_: bsc.Vertex, node: ast.filter.Has): # type is a Node if not isinstance(type_, bsc.Node): raise errors.ConsistencyError(f'expected a Node, found {type_}') @@ -189,19 +177,30 @@ class Filter(): if not type_ <= dom: raise errors.ConsistencyError(f'expected type {dom}, found {type_}') # node.count is a numerical expression - # FIXME: We have to ensure that ns.xsd.integer is always known in the schema! - self._parse_filter_expression(self.schema.literal(ns.xsd.integer), node.count) + self._parse_filter_expression(self.schema.literal(ns.bsfs.Number), node.count) + + def _distance(self, type_: bsc.Vertex, node: ast.filter.Distance): + # type is a Literal + if not isinstance(type_, bsc.Feature): + raise errors.ConsistencyError(f'expected a Feature, found {type_}') + # type exists in the schema + if type_ not in self.schema.literals(): + raise errors.ConsistencyError(f'literal {type_} is not in the schema') + # reference matches type_ + if len(node.reference) != type_.dimension: + raise errors.ConsistencyError(f'reference has dimension {len(node.reference)}, expected {type_.dimension}') + # FIXME: test dtype ## conditions - def _is(self, type_: T_VERTEX, node: ast.filter.Is): # pylint: disable=unused-argument # (node) + def _is(self, type_: bsc.Vertex, node: ast.filter.Is): # pylint: disable=unused-argument # (node) if not isinstance(type_, bsc.Node): raise errors.ConsistencyError(f'expected a Node, found {type_}') if type_ not in self.schema.nodes(): raise errors.ConsistencyError(f'node {type_} is not in the schema') - def _value(self, type_: T_VERTEX, node: ast.filter._Value): # pylint: disable=unused-argument # (node) + def _value(self, type_: bsc.Vertex, node: ast.filter._Value): # pylint: disable=unused-argument # (node) # type is a literal if not isinstance(type_, bsc.Literal): raise errors.ConsistencyError(f'expected a Literal, found {type_}') @@ -211,13 +210,16 @@ class Filter(): # FIXME: Check if node.value corresponds to type_ # FIXME: A specific literal might be requested (i.e., a numeric type when used in Has) - def _bounded(self, type_: T_VERTEX, node: ast.filter._Bounded): # pylint: disable=unused-argument # (node) + def _bounded(self, type_: bsc.Vertex, node: ast.filter._Bounded): # pylint: disable=unused-argument # (node) # type is a literal if not isinstance(type_, bsc.Literal): raise errors.ConsistencyError(f'expected a Literal, found {type_}') # type exists in the schema if type_ not in self.schema.literals(): raise errors.ConsistencyError(f'literal {type_} is not in the schema') + # type must be a numerical + if not type_ <= self.schema.literal(ns.bsfs.Number): + raise errors.ConsistencyError(f'expected a number type, found {type_}') # FIXME: Check if node.value corresponds to type_ |