def _infer_type(self, type_env: Env.Types) -> T.Base:
self.expr.infer_type(type_env, self._check_quant)
if isinstance(self.expr.type, T.Array):
if "sep" not in self.options:
raise Error.StaticTypeMismatch(
self, T.Array(None), self.expr.type,
"array command placeholder must have 'sep'")
# if sum(1 for t in [T.Int, T.Float, T.Boolean, T.String, T.File] if isinstance(self.expr.type.item_type, t)) == 0:
# raise Error.StaticTypeMismatch(self, T.Array(None), self.expr.type, "cannot use array of complex types for command placeholder")
elif "sep" in self.options:
raise Error.StaticTypeMismatch(
self,
T.Array(None),
self.expr.type,
"command placeholder has 'sep' option for non-Array expression",
)
if "true" in self.options or "false" in self.options:
if not isinstance(self.expr.type, T.Boolean):
raise Error.StaticTypeMismatch(
self,
T.Boolean(),
self.expr.type,
"command placeholder 'true' and 'false' options used with non-Boolean expression",
)
if not ("true" in self.options and "false" in self.options):
raise Error.StaticTypeMismatch(
self,
T.Boolean(),
self.expr.type,
"command placeholder with only one of 'true' and 'false' options",
)
return T.String()
def _infer_type(self, type_env: Env.Types) -> T.Base:
if not self.items:
return T.Array(None)
for item in self.items:
item.infer_type(type_env, self._check_quant)
# Start by assuming the type of the first item is the item type
item_type: T.Base = self.items[0].type
# Allow a mixture of Int and Float to construct Array[Float]
if isinstance(item_type, T.Int):
for item in self.items:
if isinstance(item.type, T.Float):
item_type = T.Float()
# If any item is String, assume item type is String
# If any item has optional quantifier, assume item type is optional
# If all items have nonempty quantifier, assume item type is nonempty
all_nonempty = len(self.items) > 0
for item in self.items:
if isinstance(item.type, T.String):
item_type = T.String(optional=item_type.optional)
if item.type.optional:
item_type = item_type.copy(optional=True)
if isinstance(item.type, T.Array) and not item.type.nonempty:
all_nonempty = False
if isinstance(item_type, T.Array):
item_type = item_type.copy(nonempty=all_nonempty)
# Check all items are coercible to item_type
for item in self.items:
try:
item.typecheck(item_type)
except Error.StaticTypeMismatch:
self._type = T.Array(item_type, optional=False, nonempty=True)
raise Error.StaticTypeMismatch(
self, item_type, item.type,
"(inconsistent types within array)") from None
return T.Array(item_type, optional=False, nonempty=True)
def infer_type(self, expr: E.Apply) -> T.Base:
if len(expr.arguments) != 2:
raise Error.WrongArity(expr, 2)
expr.arguments[0].typecheck(T.String())
expr.arguments[1].typecheck(T.Array(T.String()))
return T.Array(
T.String(),
nonempty=(
isinstance(expr.arguments[1].type, T.Array) and expr.arguments[1].type.nonempty
),
)
def infer_type(self, expr: E.Apply) -> T.Base:
if len(expr.arguments) != 1:
raise Error.WrongArity(expr, 1)
if not isinstance(expr.arguments[0].type, T.Array):
raise Error.StaticTypeMismatch(expr.arguments[0], T.Array(None),
expr.arguments[0].type)
if expr.arguments[0].type.item_type is None:
# TODO: error for 'indeterminate type'
raise Error.EmptyArray(expr.arguments[0])
ty = expr.arguments[0].type.item_type
assert isinstance(ty, T.Base)
return T.Array(ty.copy(optional=False))
def infer_type(self, expr: E.Apply) -> T.Base:
if len(expr.arguments) != 1:
raise Error.WrongArity(expr, 1)
expr.arguments[0].typecheck(T.Array(T.Any()))
# TODO: won't handle implicit coercion from T to Array[T]
assert isinstance(expr.arguments[0].type, T.Array)
if expr.arguments[0].type.item_type is None:
return T.Array(T.Any())
if not isinstance(expr.arguments[0].type.item_type, T.Array):
raise Error.StaticTypeMismatch(
expr.arguments[0], T.Array(T.Array(T.Any())), expr.arguments[0].type
)
return expr.arguments[0].type
def infer_type(self, expr: E.Apply) -> T.Base:
if len(expr.arguments) != 1:
raise Error.WrongArity(expr, 1)
if not isinstance(expr.arguments[0].type, T.Array):
raise Error.StaticTypeMismatch(expr, T.Array(None),
expr.arguments[0].type)
return T.Int()
def infer_type(self, expr: E.Apply) -> T.Base:
if len(expr.arguments) != 2:
raise Error.WrongArity(expr, 2)
arg0ty: T.Base = expr.arguments[0].type
if not isinstance(arg0ty, T.Array) or (expr._check_quant and arg0ty.optional):
raise Error.StaticTypeMismatch(expr.arguments[0], T.Array(T.Any()), arg0ty)
if isinstance(arg0ty.item_type, T.Any):
# TODO: error for 'indeterminate type'
raise Error.EmptyArray(expr.arguments[0])
arg1ty: T.Base = expr.arguments[1].type
if not isinstance(arg1ty, T.Array) or (expr._check_quant and arg1ty.optional):
raise Error.StaticTypeMismatch(expr.arguments[1], T.Array(T.Any()), arg1ty)
if isinstance(arg1ty.item_type, T.Any):
# TODO: error for 'indeterminate type'
raise Error.EmptyArray(expr.arguments[1])
return T.Array(
T.Pair(arg0ty.item_type, arg1ty.item_type),
nonempty=(arg0ty.nonempty or arg1ty.nonempty),
)
def array_type(self, items, meta):
assert len(items) >= 1
assert isinstance(items[0], T.Base)
optional = False
nonempty = False
for c in "".join(items[1:]):
if c == "?":
optional = True
if c == "+":
nonempty = True
return T.Array(items[0], optional, nonempty)
def type(self, items, meta):
quantifiers = set()
if len(items) > 1 and isinstance(items[-1], set):
quantifiers = items.pop()
param = items[1] if len(items) > 1 else None
param2 = items[2] if len(items) > 2 else None
if items[0].value == "Array":
if not param or param2:
raise Err.InvalidType(sp(self.filename, meta),
"Array must have one type parameter")
if quantifiers - set(["optional", "nonempty"]):
raise Err.ValidationError(
sp(self.filename, meta),
"invalid type quantifier(s) for Array")
return T.Array(param, "optional" in quantifiers, "nonempty"
in quantifiers)
if "nonempty" in quantifiers:
raise Err.InvalidType(
sp(self.filename, meta),
"invalid type quantifier(s) for " + items[0].value)
atomic_types = {
"Int": T.Int,
"Float": T.Float,
"Boolean": T.Boolean,
"String": T.String,
"File": T.File,
}
if items[0].value in atomic_types:
if param or param2:
raise Err.InvalidType(
sp(self.filename, meta),
items[0] + " type doesn't accept parameters")
return atomic_types[items[0].value]("optional" in quantifiers)
if items[0].value == "Map":
if not (param and param2):
raise Err.InvalidType(sp(self.filename, meta),
"Map must have two type parameters")
return T.Map((param, param2), "optional" in quantifiers)
if items[0].value == "Pair":
if not (param and param2):
raise Err.InvalidType(sp(self.filename, meta),
"Pair must have two type parameters")
return T.Pair(param, param2, "optional" in quantifiers)
if param or param2:
raise Err.InvalidType(sp(self.filename, meta),
"Unexpected type parameter(s)")
return T.StructInstance(items[0].value, "optional" in quantifiers)
def infer_type(self, expr: E.Apply) -> T.Base:
if len(expr.arguments) != 1:
raise Error.WrongArity(expr, 1)
expr.arguments[0].typecheck(T.Int())
nonempty = False
arg0 = expr.arguments[0]
if isinstance(arg0, E.Int) and arg0.value > 0:
nonempty = True
if isinstance(arg0, E.Apply) and arg0.function_name == "length":
arg00ty = arg0.arguments[0].type
if isinstance(arg00ty, T.Array) and arg00ty.nonempty:
nonempty = True
return T.Array(T.Int(), nonempty=nonempty)
def _arrayize_types(type_env: Env.Types, nonempty: bool = False) -> Env.Types:
# Given a type environment, recursively promote each binding of type T to
# Array[T] -- used in Scatter.add_to_type_env
ans = []
for node in type_env:
if isinstance(node, Env.Binding):
ans.append(
Env.Binding(node.name, T.Array(node.rhs, nonempty=nonempty),
node.ctx))
elif isinstance(node, Env.Namespace):
ans.append(
Env.Namespace(node.namespace,
_arrayize_types(node.bindings, nonempty)))
else:
assert False
return ans
def _infer_type(self, type_env: Env.Types) -> T.Base:
kty = None
vty = None
for k, v in self.items:
k.infer_type(type_env, self._check_quant)
if kty is None:
kty = k.type
else:
k.typecheck(kty)
v.infer_type(type_env, self._check_quant)
if vty is None or vty == T.Array(None) or vty == T.Map(None):
vty = v.type
else:
v.typecheck(vty)
if kty is None:
return T.Map(None)
assert vty is not None
return T.Map((kty, vty))
def infer_type(self, expr: E.Apply) -> T.Base:
if not expr.arguments:
raise Error.WrongArity(expr, 1)
if not expr.arguments[0].type.coerces(T.File(optional=True)):
if isinstance(expr.arguments[0].type, T.Array):
if expr.arguments[0].type.optional or not expr.arguments[0].type.item_type.coerces(
T.File(optional=True)
):
raise Error.StaticTypeMismatch(
expr.arguments[0], T.Array(T.File(optional=True)), expr.arguments[0].type
)
else:
raise Error.StaticTypeMismatch(
expr.arguments[0], T.File(optional=True), expr.arguments[0].type
)
if len(expr.arguments) == 2:
if expr.arguments[1].type != T.String():
raise Error.StaticTypeMismatch(
expr.arguments[1], T.String(), expr.arguments[1].type
)
elif len(expr.arguments) > 2:
raise Error.WrongArity(expr, 2)
return T.Float()
def _call_eager(self, expr: E.Apply, arguments: List[V.Base]) -> V.Base:
pfx = arguments[0].coerce(T.String()).value
return V.Array(
T.Array(T.String()),
[V.String(pfx + s.coerce(T.String()).value) for s in arguments[1].value],
)
def __init__(self):
# language built-ins
self._at = _At()
self._land = _And()
self._lor = _Or()
self._negate = StaticFunction(
"_negate", [T.Boolean()], T.Boolean(), lambda x: V.Boolean(not x.value)
)
self._add = _AddOperator()
self._sub = _ArithmeticOperator("-", lambda l, r: l - r)
self._mul = _ArithmeticOperator("*", lambda l, r: l * r)
self._div = _ArithmeticOperator("/", lambda l, r: l // r)
self._rem = StaticFunction(
"_rem", [T.Int(), T.Int()], T.Int(), lambda l, r: V.Int(l.value % r.value)
)
self._eqeq = _ComparisonOperator("==", lambda l, r: l == r)
self._neq = _ComparisonOperator("!=", lambda l, r: l != r)
self._lt = _ComparisonOperator("<", lambda l, r: l < r)
self._lte = _ComparisonOperator("<=", lambda l, r: l <= r)
self._gt = _ComparisonOperator(">", lambda l, r: l > r)
self._gte = _ComparisonOperator(">=", lambda l, r: l >= r)
# static stdlib functions
for (name, argument_types, return_type, F) in [
("floor", [T.Float()], T.Int(), lambda v: V.Int(math.floor(v.value))),
("ceil", [T.Float()], T.Int(), lambda v: V.Int(math.ceil(v.value))),
("round", [T.Float()], T.Int(), lambda v: V.Int(round(v.value))),
("length", [T.Array(T.Any())], T.Int(), lambda v: V.Int(len(v.value))),
("sub", [T.String(), T.String(), T.String()], T.String(), _sub),
("basename", [T.String(), T.String(optional=True)], T.String(), _basename),
(
"defined",
[T.Any(optional=True)],
T.Boolean(),
lambda v: V.Boolean(not isinstance(v, V.Null)),
),
# context-dependent:
("write_lines", [T.Array(T.String())], T.File(), _notimpl),
("write_tsv", [T.Array(T.Array(T.String()))], T.File(), _notimpl),
("write_map", [T.Map((T.Any(), T.Any()))], T.File(), _notimpl),
("write_json", [T.Any()], T.File(), _notimpl),
("stdout", [], T.File(), _notimpl),
("stderr", [], T.File(), _notimpl),
("glob", [T.String()], T.Array(T.File()), _notimpl),
("read_int", [T.File()], T.Int(), _notimpl),
("read_boolean", [T.File()], T.Boolean(), _notimpl),
("read_string", [T.File()], T.String(), _notimpl),
("read_float", [T.File()], T.Float(), _notimpl),
("read_array", [T.File()], T.Array(T.Any()), _notimpl),
("read_map", [T.File()], T.Map((T.Any(), T.Any())), _notimpl),
("read_lines", [T.File()], T.Array(T.Any()), _notimpl),
("read_tsv", [T.File()], T.Array(T.Array(T.String())), _notimpl),
("read_json", [T.File()], T.Any(), _notimpl),
]:
setattr(self, name, StaticFunction(name, argument_types, return_type, F))
# polymorphically typed stdlib functions which require specialized
# infer_type logic
self.range = _Range()
self.prefix = _Prefix()
self.size = _Size()
self.select_first = _SelectFirst()
self.select_all = _SelectAll()
self.zip = _Zip()
self.cross = _Zip() # FIXME
self.flatten = _Flatten()
self.transpose = _Transpose()
def _call_eager(self, expr: E.Apply, arguments: List[V.Base]) -> V.Base:
arg0 = arguments[0]
assert isinstance(arg0, V.Int)
if arg0.value < 0:
raise Error.EvalError(expr, "range() got negative argument")
return V.Array(T.Array(T.Int()), [V.Int(x) for x in range(arg0.value)])
_static_functions: List[Tuple[str, List[T.Base], T.Base, Any]] = [
("_negate", [T.Boolean()], T.Boolean(),
lambda x: V.Boolean(not x.value)), # pyre-fixme
("_rem", [T.Int(), T.Int()], T.Int(),
lambda l, r: V.Int(l.value % r.value)), # pyre-fixme
("stdout", [], T.File(), _notimpl),
("basename", [T.String(), T.String(optional=True)], T.String(), _notimpl),
# note: size() can take an empty value and probably returns 0 in that case.
# e.g. https://github.com/DataBiosphere/topmed-workflows/blob/31ba8a714b36ada929044f2ba3d130936e6c740e/CRAM-no-header-md5sum/md5sum/CRAM_md5sum.wdl#L39
("size", [T.File(optional=True),
T.String(optional=True)], T.Float(), _notimpl),
("ceil", [T.Float()], T.Int(), _notimpl),
("round", [T.Float()], T.Int(), _notimpl),
("glob", [T.String()], T.Array(T.File()), _notimpl),
("read_int", [T.String()], T.Int(), _notimpl),
("read_boolean", [T.String()], T.Boolean(), _notimpl),
("read_string", [T.String()], T.String(), _notimpl),
("read_float", [T.String()], T.Float(), _notimpl),
("read_array", [T.String()], T.Array(None), _notimpl),
("read_map", [T.String()], T.Map(None), _notimpl),
("read_lines", [T.String()], T.Array(None), _notimpl),
("read_tsv", [T.String()], T.Array(T.Array(T.String())), _notimpl),
("write_lines", [T.Array(T.String())], T.File(), _notimpl),
("write_tsv", [T.Array(T.Array(T.String()))], T.File(), _notimpl),
("write_map", [T.Map(None)], T.File(), _notimpl),
("range", [T.Int()], T.Array(T.Int()), _notimpl),
("sub", [T.String(), T.String(), T.String()], T.String(), _notimpl),
]
for name, argument_types, return_type, F in _static_functions:
