Пример #1
0
def ast(pfas,
        check=True,
        name=None,
        randseed=None,
        doc=None,
        version=None,
        metadata={},
        options={},
        tryYaml=False,
        verbose=False):
    """Create a single PFA from a chained workflow, returning the result as an abstract syntax tree.

    :type pfas: list of titus.pfaast.EngineConfig, Pythonized JSON, or JSON strings
    :param pfas: PFA documents for which the output of document *i* is the input to document *i + 1*
    :type check: bool
    :param check: test the chained PFA for validity
    :type name: string or ``None``
    :param name: optional name for the chained PFA
    :type randseed: integer or ``None``
    :param randseed: optional random number seed for the chained PFA
    :type doc: string or ``None``
    :param doc: optional documentation string for the chained PFA
    :type version: integer or ``None``
    :param version: optional version number for the chained PFA
    :type metadata: dict of strings
    :param metadata: metadata for the chained PFA (default is ``{}``)
    :type options: dict of Pythonized JSON
    :param options: implementation options for the chained PFA (default is ``{}``)
    :type tryYaml: bool
    :param tryYaml: if ``True``, attempt to interpret ``pfas`` as YAML (assuming they fail as JSON)
    :type verbose: bool
    :param verbose: if ``True``, write status messages to standard output
    :rtype: titus.pfaast.EngineConfig
    :return: a PFA document representing the chained workflow
    """

    # normalize all input forms to ASTs
    if verbose:
        sys.stderr.write(time.asctime() + " Converting all inputs to ASTs\n")
    asts = []
    for i, src in enumerate(pfas):
        if verbose:
            sys.stderr.write(time.asctime() + "     step {0}\n".format(i + 1))
        if isinstance(src, EngineConfig):
            ast = src
        elif isinstance(src, dict):
            ast = titus.reader.jsonToAst(src)
        else:
            try:
                ast = titus.reader.jsonToAst(src)
            except ValueError:
                if tryYaml:
                    ast = titus.reader.yamlToAst(src)
                else:
                    raise
        asts.append(ast)
    pfas = asts

    # helper functions for transforming names
    def split(t):
        if "." in t:
            return t[:t.rindex(".")], t[t.rindex(".") + 1:]
        else:
            return None, t

    def join(ns, n):
        if ns is None or ns == "":
            return n
        else:
            return ns + "." + n

    def prefixType(i, pfa, t):
        ns, n = split(t)
        return join(ns, "Step{0:d}_{1}_{2}".format(i + 1, pfa.name, n))

    def prefixAction(i, pfa):
        return "step{0:d}_{1}_action".format(i + 1, pfa.name)

    def prefixFcnRef(i, pfa, x):
        if x.startswith("u."):
            return "u.step{0:d}_{1}_fcn_{2}".format(i + 1, pfa.name, x[2:])
        else:
            return x

    def prefixFcnDef(i, pfa, x):
        return "step{0:d}_{1}_fcn_{2}".format(i + 1, pfa.name, x)

    def prefixCell(i, pfa, x):
        return "step{0:d}_{1}_{2}".format(i + 1, pfa.name, x)

    def prefixPool(i, pfa, x):
        return "step{0:d}_{1}_{2}".format(i + 1, pfa.name, x)

    # define new names for all types to avoid type name collisions
    if verbose:
        sys.stderr.write(time.asctime() +
                         " Changing type names to avoid collisions\n")
    originalNameToNewName = {}
    for i, pfa in enumerate(pfas):
        originalNameToNewName[i] = {}
        for typeName in list(pfa.inputPlaceholder.parser.names.names.keys()):
            keyTypeName = typeName
            if (typeName[0] == "."):
                keyTypeName = keyTypeName[1:]
            originalNameToNewName[i][keyTypeName] = prefixType(
                i, pfa, typeName)

    # but any names in the input to the first and the output from the last should not be changed
    def trivialName(i, avroType, memo):
        if isinstance(avroType, AvroArray):
            trivialName(i, avroType.items, memo)
        elif isinstance(avroType, AvroMap):
            trivialName(i, avroType.values, memo)
        elif isinstance(avroType, AvroUnion):
            for t in avroType.types:
                trivialName(i, t, memo)
        elif isinstance(avroType, (AvroFixed, AvroEnum)):
            t = avroType.fullName
            originalNameToNewName[i][t] = t
        elif isinstance(avroType, AvroRecord):
            t = avroType.fullName
            if t not in memo:
                memo.add(t)
                for f in avroType.fields:
                    trivialName(i, f.avroType, memo)
                originalNameToNewName[i][t] = t

    trivialName(0, pfas[0].input, set())
    trivialName(len(pfas) - 1, pfas[-1].output, set())

    # ensure that chained types match and will be given the same names
    if verbose:
        sys.stderr.write(
            time.asctime() +
            " Verifying that input/output schemas match along the chain\n")

    def chainPair(i, first, second, memo):
        if isinstance(first, AvroNull) and isinstance(second, AvroNull):
            return True
        elif isinstance(first, AvroBoolean) and isinstance(
                second, AvroBoolean):
            return True
        elif isinstance(first, AvroInt) and isinstance(second, AvroInt):
            return True
        elif isinstance(first, AvroLong) and isinstance(second, AvroLong):
            return True
        elif isinstance(first, AvroFloat) and isinstance(second, AvroFloat):
            return True
        elif isinstance(first, AvroDouble) and isinstance(second, AvroDouble):
            return True
        elif isinstance(first, AvroBytes) and isinstance(second, AvroBytes):
            return True
        elif isinstance(first, AvroFixed) and isinstance(second, AvroFixed):
            if first.size == second.size:
                originalNameToNewName[i + 1][
                    second.fullName] = originalNameToNewName[i][first.fullName]
                return True
            else:
                return False
        elif isinstance(first, AvroString) and isinstance(second, AvroString):
            return True
        elif isinstance(first, AvroEnum) and isinstance(second, AvroEnum):
            if first.symbols == second.symbols:
                originalNameToNewName[i + 1][
                    second.fullName] = originalNameToNewName[i][first.fullName]
                return True
            else:
                return False
        elif isinstance(first, AvroArray) and isinstance(second, AvroArray):
            return chainPair(i, first.items, second.items, memo)
        elif isinstance(first, AvroMap) and isinstance(second, AvroMap):
            return chainPair(i, first.values, second.values, memo)
        elif isinstance(first, AvroRecord) and isinstance(second, AvroRecord):
            if first.fullName not in memo:
                memo.add(first.fullName)
                if len(first.fields) != len(second.fields):
                    return False
                for f1, f2 in zip(first.fields, second.fields):
                    if f1.name != f2.name:
                        return False
                    elif not chainPair(i, f1.avroType, f2.avroType, memo):
                        return False
                originalNameToNewName[i + 1][
                    second.fullName] = originalNameToNewName[i][first.fullName]
                return True
        elif isinstance(first, AvroUnion) and isinstance(second, AvroUnion):
            for yt in second.types:
                if not any(chainPair(i, xt, yt, memo) for xt in first.types):
                    return False
            return True
        else:
            return False

    for i in range(len(pfas) - 1):
        first = pfas[i].output
        second = pfas[i + 1].input
        if not chainPair(i, first, second, set()):
            raise PFAChainError(
                "output of engine {0}: {1} not compatible with input of engine {2}: {3}"
                .format(i + 1, ts(first), i + 2, ts(second)))

    def rename(i, avroType, memo):
        if isinstance(avroType, AvroArray):
            return {"type": "array", "items": rename(i, avroType.items, memo)}
        elif isinstance(avroType, AvroMap):
            return {"type": "map", "values": rename(i, avroType.values, memo)}
        elif isinstance(avroType, AvroUnion):
            return [rename(i, t, memo) for t in avroType.types]
        elif isinstance(avroType, AvroFixed):
            ns, n = split(originalNameToNewName[i][avroType.fullName])
            out = {"type": "fixed", "name": n, "size": avroType.size}
            if ns is not None:
                out["namespace"] = ns
            return out
        elif isinstance(avroType, AvroEnum):
            ns, n = split(originalNameToNewName[i][avroType.fullName])
            out = {"type": "enum", "name": n, "symbols": avroType.symbols}
            if ns is not None:
                out["namespace"] = ns
            return out
        elif isinstance(avroType, AvroRecord):
            newName = originalNameToNewName[i][avroType.fullName]
            if newName in memo:
                return memo[newName]
            else:
                ns, n = split(newName)
                out = {"type": "record", "name": n, "fields": []}
                if ns is not None:
                    out["namespace"] = ns
                memo[newName] = join(ns, n)
                for f in avroType.fields:
                    newf = {
                        "name": f.name,
                        "type": rename(i, f.avroType, memo)
                    }
                    if f.default is not None:
                        newf["default"] = f.default
                    if f.order is not None:
                        newf["order"] = f.order
                    out["fields"].append(newf)
                return out
        else:
            return jsonlib.loads(repr(avroType))

    avroTypeBuilder = AvroTypeBuilder()
    memo = {}

    def newPlaceholder(i, oldAvroType):
        newAvroType = rename(i, oldAvroType, {})
        return avroTypeBuilder.makePlaceholder(jsonlib.dumps(newAvroType),
                                               memo)

    # combined name, if not explicitly set
    if name is None:
        name = "Chain_" + "_".join(pfa.name for pfa in pfas)

    # combined method (fold not supported yet, but could be)
    method = Method.MAP
    for pfa in pfas:
        if pfa.method == Method.EMIT:
            method = Method.EMIT
        elif pfa.method == Method.FOLD:
            raise NotImplementedError(
                "chaining of fold-type scoring engines has not been implemented yet"
            )

    # no zero or merge until we support fold method
    zero = None
    merge = None

    # input/output types from first and last
    inputPlaceholder = newPlaceholder(0, pfas[0].input)
    outputPlaceholder = newPlaceholder(len(pfas) - 1, pfas[-1].output)

    if verbose:
        sys.stderr.write(
            time.asctime() +
            " Adding [name, instance, metadata, actionsStarted, actionsFinished, version] as model parameters\n"
        )

    cells = {
        "name":
        Cell(newPlaceholder(0, AvroString()), jsonlib.dumps(""), False, False,
             CellPoolSource.EMBEDDED),
        "instance":
        Cell(newPlaceholder(0, AvroInt()), jsonlib.dumps(0), False, False,
             CellPoolSource.EMBEDDED),
        "metadata":
        Cell(newPlaceholder(0, AvroMap(AvroString())), jsonlib.dumps({}),
             False, False, CellPoolSource.EMBEDDED),
        "actionsStarted":
        Cell(newPlaceholder(0, AvroLong()), jsonlib.dumps(0), False, False,
             CellPoolSource.EMBEDDED),
        "actionsFinished":
        Cell(newPlaceholder(0, AvroLong()), jsonlib.dumps(0), False, False,
             CellPoolSource.EMBEDDED)
    }
    if version is not None:
        cells["version"] = Cell(newPlaceholder(0, AvroInt()), 0, False, False,
                                CellPoolSource.EMBEDDED)
    pools = {}

    if verbose:
        sys.stderr.write(time.asctime() +
                         " Converting scoring engine algorithm\n")

    # all code will go into user functions, including begin/action/end
    fcns = {}

    begin = [
        CellTo("name", [], Ref("name")),
        CellTo("instance", [], Ref("instance")),
        CellTo("metadata", [], Ref("metadata"))
    ]
    if version is not None:
        begin.append(CellTo("version", [], Ref("version")))

    action = [
        CellTo("actionsStarted", [], Ref("actionsStarted")),
        CellTo("actionsFinished", [], Ref("actionsFinished"))
    ]

    end = [
        CellTo("actionsStarted", [], Ref("actionsStarted")),
        CellTo("actionsFinished", [], Ref("actionsFinished"))
    ]

    for i, pfa in enumerate(pfas):
        if verbose:
            sys.stderr.write(time.asctime() +
                             "     step {0}: {1}\n".format(i + 1, pfa.name))

        thisActionFcnName = prefixAction(i, pfa)
        if i + 1 < len(pfas):
            nextActionFcnName = prefixAction(i + 1, pfas[i + 1])
        else:
            nextActionFcnName = None

        # this is a closure; it must be defined in the loop to pick up free variables
        lazyFcnReplacer = None

        def genericReplacer(expr, self):
            if isinstance(expr, FcnDef):
                return FcnDef(
                    [{
                        list(t.keys())[0]: newPlaceholder(
                            i,
                            list(t.values())[0])
                    } for t in expr.params],
                    newPlaceholder(i, expr.ret),
                    [
                        x.replace(lazyFcnReplacer) for x in expr.body
                    ],  # this is the one place where we should pass down fcnReplacer rather than self
                    expr.pos)
            elif isinstance(expr, FcnRef):
                return FcnRef(prefixFcnRef(i, pfa, expr.name), expr.pos)
            elif isinstance(expr, FcnRefFill):
                return FcnRefFill(
                    prefixFcnRef(i, pfa, expr.name),
                    dict((k, v.replace(self))
                         for k, v in list(expr.fill.items())), expr.pos)
            elif isinstance(
                    expr, CallUserFcn
            ):  # TODO: need to change the symbols of the corresponding enum
                return CallUserFcn(expr.name.replace(self),
                                   [x.replace(self) for x in expr.args],
                                   expr.pos)
            elif isinstance(expr, Call):
                if pfa.method == Method.EMIT and i + 1 < len(
                        pfas) and expr.name == "emit":
                    return Call("u." + nextActionFcnName,
                                [x.replace(self) for x in expr.args], expr.pos)
                else:
                    return Call(prefixFcnRef(i, pfa, expr.name),
                                [x.replace(self) for x in expr.args], expr.pos)
            elif isinstance(expr, Literal):
                return Literal(newPlaceholder(i, expr.avroType), expr.value,
                               expr.pos)
            elif isinstance(expr, NewObject):
                return NewObject(
                    dict((k, v.replace(self))
                         for k, v in list(expr.fields.items())),
                    newPlaceholder(i, expr.avroType), expr.pos)
            elif isinstance(expr, NewArray):
                return NewArray([x.replace(self) for x in expr.items],
                                newPlaceholder(i, expr.avroType), expr.pos)
            elif isinstance(expr, CellGet):
                return CellGet(prefixCell(i, pfa, expr.cell),
                               [x.replace(self) for x in expr.path], expr.pos)
            elif isinstance(expr, CellTo):
                return CellTo(prefixCell(i, pfa, expr.cell),
                              [x.replace(self) for x in expr.path],
                              expr.to.replace(self), expr.pos)
            elif isinstance(expr, PoolGet):
                return PoolGet(prefixPool(i, pfa, expr.pool),
                               [x.replace(self) for x in expr.path], expr.pos)
            elif isinstance(expr, PoolTo):
                return PoolTo(prefixPool(i, pfa, expr.pool),
                              [x.replace(self) for x in expr.path],
                              expr.to.replace(self), expr.init.replace(self),
                              expr.pos)
            elif isinstance(expr, CastCase):
                return CastCase(newPlaceholder(i, expr.avroType), expr.named,
                                [x.replace(self) for x in expr.body], expr.pos)
            elif isinstance(expr, Upcast):
                return Upcast(expr.expr.replace(self),
                              newPlaceholder(i, expr.avroType), expr.pos)

        genericReplacer.isDefinedAt = lambda x: isinstance(
            x, (FcnDef, FcnRef, FcnRefFill, CallUserFcn, Call, Literal,
                NewObject, CellGet, CellTo, PoolGet, PoolTo, CastCase, Upcast))

        def fcnReplacer(expr):
            return genericReplacer(expr, fcnReplacer)

        fcnReplacer.isDefinedAt = genericReplacer.isDefinedAt

        lazyFcnReplacer = fcnReplacer

        # add statements to begin
        def beginReplacer(expr):
            if isinstance(expr, Ref):
                if expr.name in ("name", "instance",
                                 "metadata") or (version is not None
                                                 and expr.name == "version"):
                    return CellGet(expr.name, [], expr.pos)
                else:
                    return expr
            else:
                return genericReplacer(expr, beginReplacer)

        beginReplacer.isDefinedAt = lambda x: isinstance(
            x, Ref) or genericReplacer.isDefinedAt(x)
        begin.extend([x.replace(beginReplacer) for x in pfa.begin])

        # add statements to end
        def endReplacer(expr):
            if isinstance(expr, Ref):
                if expr.name in ("name", "instance", "metadata",
                                 "actionsStarted", "actionsFinished") or (
                                     version is not None
                                     and expr.name == "version"):
                    return CellGet(expr.name, [], expr.pos)
                else:
                    return expr
            else:
                return genericReplacer(expr, endReplacer)

        endReplacer.isDefinedAt = lambda x: isinstance(
            x, Ref) or genericReplacer.isDefinedAt(x)
        end.extend([x.replace(endReplacer) for x in pfa.end])

        # convert the action into a user function
        def actionReplacer(expr):
            if isinstance(expr, Ref):
                if expr.name in ("name", "instance", "metadata",
                                 "actionsStarted", "actionsFinished") or (
                                     version is not None
                                     and expr.name == "version"):
                    return CellGet(expr.name, [], expr.pos)
                else:
                    return expr
            else:
                return genericReplacer(expr, actionReplacer)

        actionReplacer.isDefinedAt = lambda x: isinstance(
            x, Ref) or genericReplacer.isDefinedAt(x)

        body = [x.replace(actionReplacer) for x in pfa.action]

        if method == Method.MAP:
            # if the overall method is MAP, then we know that all of the individual engines are MAP
            # the overall action calls a nested chain of engines-as-functions and each engine-as-a-function just does its job and returns (body is unmodified)
            fcns[thisActionFcnName] = FcnDef(
                [{
                    "input": newPlaceholder(i, pfa.input)
                }], newPlaceholder(i, pfa.output), body)
            if i == 0:
                action.append(Call("u." + thisActionFcnName, [Ref("input")]))
            else:
                action[-1] = Call("u." + thisActionFcnName, [action[-1]])

        elif method == Method.EMIT:
            # if the overall method is EMIT, then some individual engines might be MAP or might be EMIT
            # the overall action calls the first engine-as-a-function and the engines-as-functions call each other (body is modified)
            if pfa.method == Method.MAP and i + 1 < len(pfas):
                body = [
                    Call("u." + nextActionFcnName, [Do(body)]),
                    LiteralNull()
                ]
            elif pfa.method == Method.MAP:
                body = [Call("emit", [Do(body)])]
            elif pfa.method == Method.EMIT:
                body.append(LiteralNull())

            fcns[thisActionFcnName] = FcnDef(
                [{
                    "input": newPlaceholder(i, pfa.input)
                }], newPlaceholder(i, AvroNull()), body)
            if i == 0:
                action.append(Call("u." + thisActionFcnName, [Ref("input")]))

        # convert all of the user functions into user functions
        for fcnName, fcnDef in list(pfa.fcns.items()):
            # note: some of these user-defined functions may call emit; if so, they'll call the right emit
            fcns[prefixFcnDef(i, pfa, fcnName)] = FcnDef(
                [{
                    list(t.keys())[0]: newPlaceholder(i,
                                                      list(t.values())[0])
                } for t in fcnDef.paramsPlaceholder],
                newPlaceholder(i, fcnDef.ret),
                [x.replace(fcnReplacer) for x in fcnDef.body], fcnDef.pos)

    if verbose:
        sys.stderr.write(time.asctime() +
                         " Create types for model parameters\n")

    for i, pfa in enumerate(pfas):
        if verbose and len(pfa.cells) > 0:
            sys.stderr.write(time.asctime() + "     step {0}:\n".format(i + 1))
        for cellName, cell in list(pfa.cells.items()):
            if verbose:
                sys.stderr.write(time.asctime() +
                                 "         cell {0}\n".format(cellName))
            newCell = Cell(newPlaceholder(i, cell.avroType), cell.init,
                           cell.shared, cell.rollback, cell.source, cell.pos)
            cells[prefixCell(i, pfa, cellName)] = newCell
            if cell.source == "embedded":

                def converter(avroType):
                    original = jsonDecoder(cell.avroType,
                                           jsonlib.loads(cell.init))
                    return jsonlib.dumps(jsonEncoder(avroType, original))

                newCell.converter = converter

    for i, pfa in enumerate(pfas):
        if verbose and len(pfa.pools) > 0:
            sys.stderr.write(time.asctime() + "     step {0}:\n".format(i + 1))
        for poolName, pool in list(pfa.pools.items()):
            if verbose:
                sys.stderr.write(time.asctime() +
                                 "         pool {0}\n".format(poolName))
            newPool = Pool(newPlaceholder(i, pool.avroType), pool.init,
                           pool.shared, pool.rollback, pool.source, pool.pos)
            pools[prefixPool(i, pfa, poolName)] = newPool
            if pool.source == "embedded":

                def converter(avroType):
                    original = jsonDecoder(pool.avroType,
                                           jsonlib.loads(pool.init))
                    return jsonlib.dumps(jsonEncoder(avroType, original))

                newPool.converter = converter

    # make sure all the types work together
    if verbose: sys.stderr.write(time.asctime() + " Resolving all types\n")
    avroTypeBuilder.resolveTypes()

    if verbose:
        sys.stderr.write(time.asctime() +
                         " Converting the model parameters themselves\n")

    for i, pfa in enumerate(pfas):
        if verbose and len(pfa.cells) > 0:
            sys.stderr.write(time.asctime() + "     step {0}:\n".format(i + 1))
        for cellName, cell in list(pfa.cells.items()):
            if verbose:
                sys.stderr.write(time.asctime() +
                                 "         cell {0}\n".format(cellName))
            if cell.source == "embedded":
                newCell = cells[prefixCell(i, pfa, cellName)]
                newCell.init = newCell.converter(newCell.avroType)

    for i, pfa in enumerate(pfas):
        if verbose and len(pfa.pools) > 0:
            sys.stderr.write(time.asctime() + "     step {0}:\n".format(i + 1))
        for poolName, pool in list(pfa.pools.items()):
            if verbose:
                sys.stderr.write(time.asctime() +
                                 "         pool {0}\n".format(poolName))
            if pool.source == "embedded":
                newPool = pools[prefixPool(i, pfa, poolName)]
                newPool.init = newPool.converter(newPool.avroType)

    # randseed, doc, version, metadata, and options need to be explicitly set

    # return a (possibly checked) AST
    out = EngineConfig(name, method, inputPlaceholder, outputPlaceholder,
                       begin, action, end, fcns, zero, merge, cells, pools,
                       randseed, doc, version, metadata, options)
    if check:
        if verbose:
            sys.stderr.write(time.asctime() + " Verifying PFA validity\n")
        PFAEngine.fromAst(out)

    if verbose: sys.stderr.write(time.asctime() + " Done\n")
    return out
Пример #2
0
 def genericReplacer(expr, self):
     if isinstance(expr, FcnDef):
         return FcnDef(
             [{
                 list(t.keys())[0]: newPlaceholder(
                     i,
                     list(t.values())[0])
             } for t in expr.params],
             newPlaceholder(i, expr.ret),
             [
                 x.replace(lazyFcnReplacer) for x in expr.body
             ],  # this is the one place where we should pass down fcnReplacer rather than self
             expr.pos)
     elif isinstance(expr, FcnRef):
         return FcnRef(prefixFcnRef(i, pfa, expr.name), expr.pos)
     elif isinstance(expr, FcnRefFill):
         return FcnRefFill(
             prefixFcnRef(i, pfa, expr.name),
             dict((k, v.replace(self))
                  for k, v in list(expr.fill.items())), expr.pos)
     elif isinstance(
             expr, CallUserFcn
     ):  # TODO: need to change the symbols of the corresponding enum
         return CallUserFcn(expr.name.replace(self),
                            [x.replace(self) for x in expr.args],
                            expr.pos)
     elif isinstance(expr, Call):
         if pfa.method == Method.EMIT and i + 1 < len(
                 pfas) and expr.name == "emit":
             return Call("u." + nextActionFcnName,
                         [x.replace(self) for x in expr.args], expr.pos)
         else:
             return Call(prefixFcnRef(i, pfa, expr.name),
                         [x.replace(self) for x in expr.args], expr.pos)
     elif isinstance(expr, Literal):
         return Literal(newPlaceholder(i, expr.avroType), expr.value,
                        expr.pos)
     elif isinstance(expr, NewObject):
         return NewObject(
             dict((k, v.replace(self))
                  for k, v in list(expr.fields.items())),
             newPlaceholder(i, expr.avroType), expr.pos)
     elif isinstance(expr, NewArray):
         return NewArray([x.replace(self) for x in expr.items],
                         newPlaceholder(i, expr.avroType), expr.pos)
     elif isinstance(expr, CellGet):
         return CellGet(prefixCell(i, pfa, expr.cell),
                        [x.replace(self) for x in expr.path], expr.pos)
     elif isinstance(expr, CellTo):
         return CellTo(prefixCell(i, pfa, expr.cell),
                       [x.replace(self) for x in expr.path],
                       expr.to.replace(self), expr.pos)
     elif isinstance(expr, PoolGet):
         return PoolGet(prefixPool(i, pfa, expr.pool),
                        [x.replace(self) for x in expr.path], expr.pos)
     elif isinstance(expr, PoolTo):
         return PoolTo(prefixPool(i, pfa, expr.pool),
                       [x.replace(self) for x in expr.path],
                       expr.to.replace(self), expr.init.replace(self),
                       expr.pos)
     elif isinstance(expr, CastCase):
         return CastCase(newPlaceholder(i, expr.avroType), expr.named,
                         [x.replace(self) for x in expr.body], expr.pos)
     elif isinstance(expr, Upcast):
         return Upcast(expr.expr.replace(self),
                       newPlaceholder(i, expr.avroType), expr.pos)
Пример #3
0
def _readArgument(data, dot, avroTypeBuilder):
    if data is None:
        return LiteralNull(dot)
    elif isinstance(data, bool):
        return LiteralBoolean(data, dot)
    elif isinstance(data, int):
        if -2147483648 <= data <= 2147483647:
            return LiteralInt(data, dot)
        elif -9223372036854775808 <= data <= 9223372036854775807:
            return LiteralLong(data, dot)
        else:
            raise PFASyntaxException("integer out of range: " + str(data), dot)
    elif isinstance(data, float):
        return LiteralDouble(data, dot)
    elif isinstance(data, str):
        if "." in data:
            words = data.split(".")
            ref = words[0]
            rest = words[1:]
            if not validSymbolName(ref):
                raise PFASyntaxException("\"{0}\" is not a valid symbol name".format(ref), dot)
            for i in range(len(rest)):
                try:
                    asint = int(rest[i])
                except ValueError:
                    rest[i] = LiteralString(rest[i], dot)
                else:
                    rest[i] = LiteralInt(asint, dot)
            return AttrGet(Ref(ref), rest, dot)
        elif validSymbolName(data):
            return Ref(data, dot)
        else:
            raise PFASyntaxException("\"{0}\" is not a valid symbol name".format(data), dot)

    elif isinstance(data, (list, tuple)):
        if len(data) == 1 and isinstance(data[0], str):
            return LiteralString(data[0], dot)
        else:
            raise PFASyntaxException("expecting expression, which may be [\"string\"], but no other array can be used as an expression", dot)

    elif isinstance(data, dict):
        at = data.get("@")
        keys = set(x for x in data.keys() if x != "@")

        _path = []
        _seq = True
        _partial = False
        _code = 0
        _newObject = None
        _newArray = None
        _filter = None

        for key in keys:
            if key == "int": _int = _readInt(data[key], dot + " -> " + key)
            elif key == "long": _long = _readLong(data[key], dot + " -> " + key)
            elif key == "float": _float = _readFloat(data[key], dot + " -> " + key)
            elif key == "double": _double = _readDouble(data[key], dot + " -> " + key)
            elif key == "string": _string = _readString(data[key], dot + " -> " + key)
            elif key == "base64": _bytes = _readBase64(data[key], dot + " -> " + key)
            elif key == "type": _avroType = _readAvroPlaceholder(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "value": _value = _readJsonToString(data[key], dot + " -> " + key)

            elif key == "let": _let = _readExpressionMap(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "set": _set = _readExpressionMap(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "for": _forlet = _readExpressionMap(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "step": _forstep = _readExpressionMap(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "ifnotnull": _ifnotnull = _readExpressionMap(data[key], dot + " -> " + key, avroTypeBuilder)

            elif key == "do":
                if isinstance(data[key], (list, tuple)):
                    _body = _readExpressionArray(data[key], dot + " -> " + key, avroTypeBuilder)
                else:
                    _body = [_readExpression(data[key], dot + " -> " + key, avroTypeBuilder)]
            elif key == "then":
                if isinstance(data[key], (list, tuple)):
                    _thenClause = _readExpressionArray(data[key], dot + " -> " + key, avroTypeBuilder)
                else:
                    _thenClause = [_readExpression(data[key], dot + " -> " + key, avroTypeBuilder)]
            elif key == "else":
                if isinstance(data[key], (list, tuple)):
                    _elseClause = _readExpressionArray(data[key], dot + " -> " + key, avroTypeBuilder)
                else:
                    _elseClause = [_readExpression(data[key], dot + " -> " + key, avroTypeBuilder)]
            elif key == "log":
                if isinstance(data[key], (list, tuple)):
                    _log = _readExpressionArray(data[key], dot + " -> " + key, avroTypeBuilder)
                else:
                    _log = [_readExpression(data[key], dot + " -> " + key, avroTypeBuilder)]
            elif key == "path":
                _path = _readExpressionArray(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "args":
                _callwithargs = _readExpressionArray(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "try":
                if isinstance(data[key], (list, tuple)):
                    _trycatch = _readExpressionArray(data[key], dot + " -> " + key, avroTypeBuilder)
                else:
                    _trycatch = [_readExpression(data[key], dot + " -> " + key, avroTypeBuilder)]

            elif key == "attr": _attr = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "if": _ifPredicate = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "while": _whilePredicate = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "until": _until = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "unpack": _unpack = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "del": _dell = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)

            elif key == "cond":
                _cond = _readExpressionArray(data[key], dot + " -> " + key, avroTypeBuilder)
                if any(x.elseClause is not None for x in _cond):
                    raise PFASyntaxException("cond expression must only contain else-less if expressions", pos(dot, at))

            elif key == "cases": _cases = _readCastCaseArray(data[key], dot + " -> " + key, avroTypeBuilder)

            elif key == "foreach": _foreach = _readString(data[key], dot + " -> " + key)
            elif key == "forkey": _forkey = _readString(data[key], dot + " -> " + key)
            elif key == "forval": _forval = _readString(data[key], dot + " -> " + key)
            elif key == "fcn": _fcnref = _readString(data[key], dot + " -> " + key)
            elif key == "cell": _cell = _readString(data[key], dot + " -> " + key)
            elif key == "pool": _pool = _readString(data[key], dot + " -> " + key)

            elif key == "in": _in = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "cast": _cast = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "upcast": _upcast = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "init": _init = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "call": _callwith = _readExpression(data[key], dot + " -> " + key, avroTypeBuilder)

            elif key == "seq": _seq = _readBoolean(data[key], dot + " -> " + key)
            elif key == "partial": _partial = _readBoolean(data[key], dot + " -> " + key)

            elif key == "doc": _doc = _readString(data[key], dot + " -> " + key)
            elif key == "error": _error = _readString(data[key], dot + " -> " + key)
            elif key == "code": _code = _readInt(data[key], dot + " -> " + key)
            elif key == "namespace": _namespace = _readString(data[key], dot + " -> " + key)

            elif key == "new":
                if isinstance(data[key], dict):
                    _newObject = _readExpressionMap(data[key], dot + " -> " + key, avroTypeBuilder)
                elif isinstance(data[key], (list, tuple)):
                    _newArray = _readExpressionArray(data[key], dot + " -> " + key, avroTypeBuilder)
                else:
                    raise PFASyntaxException("\"new\" must be an object (map, record) or an array", pos(dot, at))

            elif key == "params": _params = _readParams(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "ret": _ret = _readAvroPlaceholder(data[key], dot + " -> " + key, avroTypeBuilder)
            elif key == "as": _as = _readAvroPlaceholder(data[key], dot + " -> " + key, avroTypeBuilder)

            elif key == "to": _to = _readArgument(data[key], dot + " -> " + key, avroTypeBuilder)

            elif key == "fill": _fill = _readArgumentMap(data[key], dot + " -> " + key, avroTypeBuilder)

            elif key == "filter": _filter = _readStringOrIntArray(data[key], dot + " -> " + key)

            elif key == "format": _format = _readStringPairs(data[key], dot + " -> " + key)
            elif key == "pack": _pack = _readStringExpressionPairs(data[key], dot + " -> " + key, avroTypeBuilder)

            else:
                _callName = key
                if isinstance(data[key], (list, tuple)):
                    _callArgs = _readArgumentArray(data[key], dot + " -> " + key, avroTypeBuilder)
                else:
                    _callArgs = [_readArgument(data[key], dot + " -> " + key, avroTypeBuilder)]

        if "foreach" in keys and not validSymbolName(_foreach):
            raise PFASyntaxException("\"{0}\" is not a valid symbol name".format(data[keys]), pos(dot, at))
        if "forkey" in keys and not validSymbolName(_forkey):
            raise PFASyntaxException("\"{0}\" is not a valid symbol name".format(data[keys]), pos(dot, at))
        if "forval" in keys and not validSymbolName(_forval):
            raise PFASyntaxException("\"{0}\" is not a valid symbol name".format(data[keys]), pos(dot, at))
        if "fcn" in keys and not validFunctionName(_fcnref):
            raise PFASyntaxException("\"{0}\" is not a valid function name".format(data[keys]), pos(dot, at))

        if keys == set(["int"]):                             return LiteralInt(_int, pos(dot, at))
        elif keys == set(["long"]):                          return LiteralLong(_long, pos(dot, at))
        elif keys == set(["float"]):                         return LiteralFloat(_float, pos(dot, at))
        elif keys == set(["double"]):                        return LiteralDouble(_double, pos(dot, at))
        elif keys == set(["string"]):                        return LiteralString(_string, pos(dot, at))
        elif keys == set(["base64"]):                        return LiteralBase64(_bytes, pos(dot, at))
        elif keys == set(["type", "value"]):                 return Literal(_avroType, _value, pos(dot, at))

        elif keys == set(["new", "type"]) and _newObject is not None:
                                                             return NewObject(_newObject, _avroType, pos(dot, at))
        elif keys == set(["new", "type"]) and _newArray is not None:
                                                             return NewArray(_newArray, _avroType, pos(dot, at))

        elif keys == set(["do"]):                            return Do(_body, pos(dot, at))
        elif keys == set(["let"]):                           return Let(_let, pos(dot, at))
        elif keys == set(["set"]):                           return SetVar(_set, pos(dot, at))

        elif keys == set(["attr", "path"]):                  return AttrGet(_attr, _path, pos(dot, at))
        elif keys == set(["attr", "path", "to"]):            return AttrTo(_attr, _path, _to, pos(dot, at))
        elif keys == set(["cell"]) or \
             keys == set(["cell", "path"]):                  return CellGet(_cell, _path, pos(dot, at))
        elif keys == set(["cell", "to"]) or \
             keys == set(["cell", "path", "to"]):            return CellTo(_cell, _path, _to, pos(dot, at))
        elif keys == set(["pool", "path"]):                  return PoolGet(_pool, _path, pos(dot, at))
        elif keys == set(["pool", "path", "to", "init"]):    return PoolTo(_pool, _path, _to, _init, pos(dot, at))
        elif keys == set(["pool", "del"]):                   return PoolDel(_pool, _dell, pos(dot, at))

        elif keys == set(["if", "then"]):                    return If(_ifPredicate, _thenClause, None, pos(dot, at))
        elif keys == set(["if", "then", "else"]):            return If(_ifPredicate, _thenClause, _elseClause, pos(dot, at))
        elif keys == set(["cond"]):                          return Cond(_cond, None, pos(dot, at))
        elif keys == set(["cond", "else"]):                  return Cond(_cond, _elseClause, pos(dot, at))

        elif keys == set(["while", "do"]):                   return While(_whilePredicate, _body, pos(dot, at))
        elif keys == set(["do", "until"]):                   return DoUntil(_body, _until, pos(dot, at))
        elif keys == set(["for", "while", "step", "do"]):    return For(_forlet, _whilePredicate, _forstep, _body, pos(dot, at))

        elif keys == set(["foreach", "in", "do"]) or \
             keys == set(["foreach", "in", "do", "seq"]):    return Foreach(_foreach, _in, _body, _seq, pos(dot, at))
        elif keys == set(["forkey", "forval", "in", "do"]):  return Forkeyval(_forkey, _forval, _in, _body, pos(dot, at))

        elif keys == set(["cast", "cases"]) or \
             keys == set(["cast", "cases", "partial"]):      return CastBlock(_cast, _cases, _partial, pos(dot, at))
        elif keys == set(["upcast", "as"]):                  return Upcast(_upcast, _as, pos(dot, at))

        elif keys == set(["ifnotnull", "then"]):             return IfNotNull(_ifnotnull, _thenClause, None, pos(dot, at))
        elif keys == set(["ifnotnull", "then", "else"]):     return IfNotNull(_ifnotnull, _thenClause, _elseClause, pos(dot, at))

        elif keys == set(["pack"]):                          return Pack(_pack, pos(dot, at))
        elif keys == set(["unpack", "format", "then"]):      return Unpack(_unpack, _format, _thenClause, None, pos(dot, at))
        elif keys == set(["unpack", "format", "then", "else"]):
                                                             return Unpack(_unpack, _format, _thenClause, _elseClause, pos(dot, at))

        elif keys == set(["doc"]):                           return Doc(_doc, pos(dot, at))

        elif keys == set(["error"]):                         return Error(_error, None, pos(dot, at))
        elif keys == set(["error", "code"]):                 return Error(_error, _code, pos(dot, at))
        elif keys == set(["try"]) or \
             keys == set(["try", "filter"]):                 return Try(_trycatch, _filter, pos(dot, at))
        elif keys == set(["log"]):                           return Log(_log, None, pos(dot, at))
        elif keys == set(["log", "namespace"]):              return Log(_log, _namespace, pos(dot, at))

        elif keys == set(["params", "ret", "do"]):           return FcnDef(_params, _ret, _body, pos(dot, at))
        elif keys == set(["fcn"]):                           return FcnRef(_fcnref, pos(dot, at))
        elif keys == set(["fcn", "fill"]):                   return FcnRefFill(_fcnref, _fill, pos(dot, at))
        elif keys == set(["call", "args"]):                  return CallUserFcn(_callwith, _callwithargs, pos(dot, at))

        elif len(keys) == 1 and list(keys)[0] not in \
             set(["args", "as", "attr", "base64", "call", "cases", "cast", "cell", "code", "cond", "do", "doc", "double", "else",
                  "error", "fcn", "fill", "filter", "float", "for", "foreach", "forkey", "format", "forval", "if", "ifnotnull", "in", "init",
                  "int", "let", "log", "long", "namespace", "new", "pack", "params", "partial", "path", "pool", "ret", "seq",
                  "set", "step", "string", "then", "to", "try", "type", "unpack", "until", "upcast", "value", "while"]):
                                                             return Call(_callName, _callArgs, pos(dot, at))

        else: raise PFASyntaxException("unrecognized special form: {0} (not enough arguments? too many?)".format(", ".join(keys)), pos(dot, at))

    else:
        raise PFASyntaxException("expected expression, not " + _trunc(repr(data)), dot)