Example #1
0
def wrap_function(func):
    """
    RETURN A THREE-PARAMETER WINDOW FUNCTION TO MATCH
    """
    if isinstance(func, basestring):
        return compile_expression(func)

    numarg = func.__code__.co_argcount
    if numarg == 0:

        def temp(row, rownum, rows):
            return func()

        return temp
    elif numarg == 1:

        def temp(row, rownum, rows):
            return func(row)

        return temp
    elif numarg == 2:

        def temp(row, rownum, rows):
            return func(row, rownum)

        return temp
    elif numarg == 3:
        return func
Example #2
0
def wrap_function(func):
    """
    RETURN A THREE-PARAMETER WINDOW FUNCTION TO MATCH
    """
    if isinstance(func, basestring):
        return compile_expression(func)

    numarg = func.__code__.co_argcount
    if numarg == 0:

        def temp(row, rownum, rows):
            return func()

        return temp
    elif numarg == 1:

        def temp(row, rownum, rows):
            return func(row)

        return temp
    elif numarg == 2:

        def temp(row, rownum, rows):
            return func(row, rownum)

        return temp
    elif numarg == 3:
        return func
Example #3
0
    def where(self, where):
        temp = None
        if isinstance(where, Mapping):
            exec("def temp(row):\n    return "+jx_expression(where).to_python())
        elif isinstance(where, Expression):
            temp = compile_expression(where.to_python())
        else:
            temp = where

        return ListContainer("from "+self.name, filter(temp, self.data), self.schema)
    def where(self, where):
        temp = None
        if isinstance(where, Mapping):
            exec("def temp(row):\n    return " +
                 jx_expression(where).to_python())
        elif isinstance(where, Expression):
            temp = compile_expression(where.to_python())
        else:
            temp = where

        return ListContainer("from " + self.name, filter(temp, self.data),
                             self.schema)
Example #5
0
def list_aggs(frum, query):
    frum = wrap(frum)
    select = listwrap(query.select)

    for e in query.edges:
        if isinstance(e.domain, DefaultDomain):
            accessor = jx_expression_to_function(e.value)
            unique_values = set(map(accessor, frum))
            if None in unique_values:
                e.allowNulls = coalesce(e.allowNulls, True)
                unique_values -= {None}
            e.domain = SimpleSetDomain(partitions=list(sorted(unique_values)))
        else:
            pass

    s_accessors = [(ss.name, compile_expression(ss.value.to_python())) for ss in select]

    result = {
        s.name: Matrix(
            dims=[len(e.domain.partitions) + (1 if e.allowNulls else 0) for e in query.edges],
            zeros=lambda: windows.name2accumulator.get(s.aggregate)(**s)
        )
        for s in select
    }
    where = jx_expression_to_function(query.where)
    coord = [None]*len(query.edges)
    edge_accessor = [(i, make_accessor(e)) for i, e in enumerate(query.edges)]

    net_new_edge_names = set(wrap(query.edges).name) - UNION(e.value.vars() for e in query.edges)
    if net_new_edge_names & UNION(ss.value.vars() for ss in select):
        # s_accessor NEEDS THESE EDGES, SO WE PASS THEM ANYWAY
        for d in filter(where, frum):
            d = d.copy()
            for c, get_matches in edge_accessor:
                coord[c] = get_matches(d)

            for s_name, s_accessor in s_accessors:
                mat = result[s_name]
                for c in itertools.product(*coord):
                    acc = mat[c]
                    for e, cc in zip(query.edges, c):
                        d[e.name] = e.domain.partitions[cc]
                    val = s_accessor(d, c, frum)
                    acc.add(val)
    else:
        # FASTER
        for d in filter(where, frum):
            for c, get_matches in edge_accessor:
                coord[c] = get_matches(d)

            for s_name, s_accessor in s_accessors:
                mat = result[s_name]
                for c in itertools.product(*coord):
                    acc = mat[c]
                    val = s_accessor(d, c, frum)
                    acc.add(val)

    for s in select:
        # if s.aggregate == "count":
        #     continue
        m = result[s.name]
        for c, var in m.items():
            if var != None:
                m[c] = var.end()

    from pyLibrary.queries.containers.cube import Cube

    output = Cube(select, query.edges, result)
    return output
Example #6
0
def list_aggs(frum, query):
    frum = wrap(frum)
    select = listwrap(query.select)

    for e in query.edges:
        if isinstance(e.domain, DefaultDomain):
            accessor = jx_expression_to_function(e.value)
            unique_values = set(map(accessor, frum))
            if None in unique_values:
                e.allowNulls = coalesce(e.allowNulls, True)
                unique_values -= {None}
            e.domain = SimpleSetDomain(partitions=list(sorted(unique_values)))
        else:
            pass

    s_accessors = [(ss.name, compile_expression(ss.value.to_python())) for ss in select]

    result = {
        s.name: Matrix(
            dims=[len(e.domain.partitions) + (1 if e.allowNulls else 0) for e in query.edges],
            zeros=lambda: windows.name2accumulator.get(s.aggregate)(**s)
        )
        for s in select
    }
    where = jx_expression_to_function(query.where)
    coord = [None]*len(query.edges)
    edge_accessor = [(i, make_accessor(e)) for i, e in enumerate(query.edges)]

    net_new_edge_names = set(wrap(query.edges).name) - UNION(e.value.vars() for e in query.edges)
    if net_new_edge_names & UNION(ss.value.vars() for ss in select):
        # s_accessor NEEDS THESE EDGES, SO WE PASS THEM ANYWAY
        for d in filter(where, frum):
            d = d.copy()
            for c, get_matches in edge_accessor:
                coord[c] = get_matches(d)

            for s_name, s_accessor in s_accessors:
                mat = result[s_name]
                for c in itertools.product(*coord):
                    acc = mat[c]
                    for e, cc in zip(query.edges, c):
                        d[e.name] = e.domain.partitions[cc]
                    val = s_accessor(d, c, frum)
                    acc.add(val)
    else:
        # FASTER
        for d in filter(where, frum):
            for c, get_matches in edge_accessor:
                coord[c] = get_matches(d)

            for s_name, s_accessor in s_accessors:
                mat = result[s_name]
                for c in itertools.product(*coord):
                    acc = mat[c]
                    val = s_accessor(d, c, frum)
                    acc.add(val)

    for s in select:
        # if s.aggregate == "count":
        #     continue
        m = result[s.name]
        for c, var in m.items():
            if var != None:
                m[c] = var.end()

    from pyLibrary.queries.containers.cube import Cube

    output = Cube(select, query.edges, result)
    return output