1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
|
/* NSC -- new Scala compiler
* Copyright 2005-2013 LAMP/EPFL
* @author Paul Phillips
*/
package scala.tools.nsc
package typechecker
import Checkability._
import scala.language.postfixOps
/** On pattern matcher checkability:
*
* Consider a pattern match of this form: (x: X) match { case _: P => }
*
* There are four possibilities to consider:
* [P1] X will always conform to P
* [P2] x will never conform to P
* [P3] X <: P if some runtime test is true
* [P4] X cannot be checked against P
*
* The first two cases correspond to those when there is enough
* static information to say X <: P or that (x ∈ X) ⇒ (x ∉ P).
* The fourth case includes unknown abstract types or structural
* refinements appearing within a pattern.
*
* The third case is the interesting one. We designate another type, XR,
* which is essentially the intersection of X and |P|, where |P| is
* the erasure of P. If XR <: P, then no warning is emitted.
*
* Examples of how this info is put to use:
* sealed trait A[T] ; class B[T] extends A[T]
* def f(x: B[Int]) = x match { case _: A[Int] if true => }
* def g(x: A[Int]) = x match { case _: B[Int] => }
*
* `f` requires no warning because X=B[Int], P=A[Int], and B[Int] <:< A[Int].
* `g` requires no warning because X=A[Int], P=B[Int], XR=B[Int], and B[Int] <:< B[Int].
* XR=B[Int] because a value of type A[Int] which is tested to be a B can
* only be a B[Int], due to the definition of B (B[T] extends A[T].)
*
* This is something like asSeenFrom, only rather than asking what a type looks
* like from the point of view of one of its base classes, we ask what it looks
* like from the point of view of one of its subclasses.
*/
trait Checkable {
self: Analyzer =>
import global._
import definitions._
import CheckabilityChecker.{ isNeverSubType, isNeverSubClass }
/** The applied type of class 'to' after inferring anything
* possible from the knowledge that 'to' must also be of the
* type given in 'from'.
*/
def propagateKnownTypes(from: Type, to: Symbol): Type = {
def tparams = to.typeParams
val tvars = tparams map (p => TypeVar(p))
val tvarType = appliedType(to, tvars: _*)
val bases = from.baseClasses filter (to.baseClasses contains _)
bases foreach { bc =>
val tps1 = (from baseType bc).typeArgs
val tps2 = (tvarType baseType bc).typeArgs
if (tps1.size != tps2.size)
devWarning(s"Unequally sized type arg lists in propagateKnownTypes($from, $to): ($tps1, $tps2)")
(tps1, tps2).zipped foreach (_ =:= _)
// Alternate, variance respecting formulation causes
// neg/unchecked3.scala to fail (abstract types). TODO -
// figure it out. It seems there is more work to do if I
// allow for variance, because the constraints accumulate
// as bounds and "tvar.instValid" is false.
//
// foreach3(tps1, tps2, bc.typeParams)((tp1, tp2, tparam) =>
// if (tparam.initialize.isCovariant) tp1 <:< tp2
// else if (tparam.isContravariant) tp2 <:< tp1
// else tp1 =:= tp2
// )
}
val resArgs = tparams zip tvars map {
case (_, tvar) if tvar.instValid => tvar.constr.inst
case (tparam, _) => tparam.tpeHK
}
appliedType(to, resArgs: _*)
}
private def isUnwarnableTypeArgSymbol(sym: Symbol) = (
sym.isTypeParameter // dummy
|| (sym.name.toTermName == nme.WILDCARD) // _
|| nme.isVariableName(sym.name) // type variable
)
private def isUnwarnableTypeArg(arg: Type) = (
uncheckedOk(arg) // @unchecked T
|| isUnwarnableTypeArgSymbol(arg.typeSymbolDirect) // has to be direct: see pos/t1439
)
private def uncheckedOk(tp: Type) = tp hasAnnotation UncheckedClass
private def typeArgsInTopLevelType(tp: Type): List[Type] = {
val tps = tp match {
case RefinedType(parents, _) => parents flatMap typeArgsInTopLevelType
case TypeRef(_, ArrayClass, arg :: Nil) => typeArgsInTopLevelType(arg)
case TypeRef(pre, sym, args) => typeArgsInTopLevelType(pre) ++ args
case ExistentialType(tparams, underlying) => tparams.map(_.tpe) ++ typeArgsInTopLevelType(underlying)
case _ => Nil
}
tps filterNot isUnwarnableTypeArg
}
private class CheckabilityChecker(val X: Type, val P: Type) {
def Xsym = X.typeSymbol
def Psym = P.typeSymbol
def XR = if (Xsym == AnyClass) classExistentialType(Psym) else propagateKnownTypes(X, Psym)
// sadly the spec says (new java.lang.Boolean(true)).isInstanceOf[scala.Boolean]
def P1 = X matchesPattern P
def P2 = !Psym.isPrimitiveValueClass && isNeverSubType(X, P)
def P3 = isNonRefinementClassType(P) && (XR matchesPattern P)
def P4 = !(P1 || P2 || P3)
def summaryString = f"""
|Checking checkability of (x: $X) against pattern $P
|[P1] $P1%-6s X <: P // $X <: $P
|[P2] $P2%-6s x ∉ P // (x ∈ $X) ⇒ (x ∉ $P)
|[P3] $P3%-6s XR <: P // $XR <: $P
|[P4] $P4%-6s None of the above // !(P1 || P2 || P3)
""".stripMargin.trim
val result = (
if (X.isErroneous || P.isErroneous) CheckabilityError
else if (P1) StaticallyTrue
else if (P2) StaticallyFalse
else if (P3) RuntimeCheckable
else if (uncheckableType == NoType) {
// Avoid warning (except ourselves) if we can't pinpoint the uncheckable type
debuglog("Checkability checker says 'Uncheckable', but uncheckable type cannot be found:\n" + summaryString)
CheckabilityError
}
else Uncheckable
)
lazy val uncheckableType = if (Psym.isAbstractType) P else {
val possibles = typeArgsInTopLevelType(P).toSet
val opt = possibles find { targ =>
// Create a derived type with every possibly uncheckable type replaced
// with a WildcardType, except for 'targ'. If !(XR <: derived) then
// 'targ' is uncheckable.
val derived = P map (tp => if (possibles(tp) && !(tp =:= targ)) WildcardType else tp)
!(XR <:< derived)
}
opt getOrElse NoType
}
def neverSubClass = isNeverSubClass(Xsym, Psym)
def neverMatches = result == StaticallyFalse
def isUncheckable = result == Uncheckable
def isCheckable = !isUncheckable
def uncheckableMessage = uncheckableType match {
case NoType => "something"
case tp @ RefinedType(_, _) => "refinement " + tp
case TypeRef(_, sym, _) if sym.isAbstractType => "abstract type " + sym.name
case tp => "non-variable type argument " + tp
}
}
/** X, P, [P1], etc. are all explained at the top of the file.
*/
private object CheckabilityChecker {
/** Are these symbols classes with no subclass relationship? */
def areUnrelatedClasses(sym1: Symbol, sym2: Symbol) = (
sym1.isClass
&& sym2.isClass
&& !(sym1 isSubClass sym2)
&& !(sym2 isSubClass sym1)
)
/** Are all children of these symbols pairwise irreconcilable? */
def allChildrenAreIrreconcilable(sym1: Symbol, sym2: Symbol) = (
sym1.children.toList forall (c1 =>
sym2.children.toList forall (c2 =>
areIrreconcilableAsParents(c1, c2)
)
)
)
/** Is it impossible for the given symbols to be parents in the same class?
* This means given A and B, can there be an instance of A with B? This is the
* case if neither A nor B is a subclass of the other, and one of the following
* additional conditions holds:
* - either A or B is effectively final
* - neither A nor B is a trait (i.e. both are actual classes, not eligible for mixin)
* - both A and B are sealed, and every possible pairing of their children is irreconcilable
*
* TODO: the last two conditions of the last possibility (that the symbols are not of
* classes being compiled in the current run) are because this currently runs too early,
* and .children returns Nil for sealed classes because their children will not be
* populated until typer. It was too difficult to move things around for the moment,
* so I will consult with moors about the optimal time to be doing this.
*/
def areIrreconcilableAsParents(sym1: Symbol, sym2: Symbol): Boolean = areUnrelatedClasses(sym1, sym2) && (
isEffectivelyFinal(sym1) // initialization important
|| isEffectivelyFinal(sym2)
|| !sym1.isTrait && !sym2.isTrait
|| sym1.isSealed && sym2.isSealed && allChildrenAreIrreconcilable(sym1, sym2) && !currentRun.compiles(sym1) && !currentRun.compiles(sym2)
)
private def isEffectivelyFinal(sym: Symbol): Boolean = (
// initialization important
sym.initialize.isEffectivelyFinal || (
settings.future && isTupleSymbol(sym) // SI-7294 step into the future and treat TupleN as final.
)
)
def isNeverSubClass(sym1: Symbol, sym2: Symbol) = areIrreconcilableAsParents(sym1, sym2)
private def isNeverSubArgs(tps1: List[Type], tps2: List[Type], tparams: List[Symbol]): Boolean = /*logResult(s"isNeverSubArgs($tps1, $tps2, $tparams)")*/ {
def isNeverSubArg(t1: Type, t2: Type, variance: Variance) = (
if (variance.isInvariant) isNeverSameType(t1, t2)
else if (variance.isCovariant) isNeverSubType(t2, t1)
else if (variance.isContravariant) isNeverSubType(t1, t2)
else false
)
exists3(tps1, tps2, tparams map (_.variance))(isNeverSubArg)
}
private def isNeverSameType(tp1: Type, tp2: Type): Boolean = (tp1, tp2) match {
case (TypeRef(_, sym1, args1), TypeRef(_, sym2, args2)) =>
isNeverSubClass(sym1, sym2) || ((sym1 == sym2) && isNeverSubArgs(args1, args2, sym1.typeParams))
case _ =>
false
}
// Important to dealias at any entry point (this is the only one at this writing.)
def isNeverSubType(tp1: Type, tp2: Type): Boolean = /*logResult(s"isNeverSubType($tp1, $tp2)")*/((tp1.dealias, tp2.dealias) match {
case (TypeRef(_, sym1, args1), TypeRef(_, sym2, args2)) =>
isNeverSubClass(sym1, sym2) || {
(sym1 isSubClass sym2) && {
val tp1seen = tp1 baseType sym2
isNeverSubArgs(tp1seen.typeArgs, args2, sym2.typeParams)
}
}
case _ => false
})
}
trait InferCheckable {
self: Inferencer =>
def isUncheckable(P0: Type) = !isCheckable(P0)
def isCheckable(P0: Type): Boolean = (
uncheckedOk(P0) || (P0.widen match {
case TypeRef(_, NothingClass | NullClass | AnyValClass, _) => false
case RefinedType(_, decls) if !decls.isEmpty => false
case p =>
new CheckabilityChecker(AnyClass.tpe, p) isCheckable
})
)
/** TODO: much better error positions.
* Kind of stuck right now because they just pass us the one tree.
* TODO: Eliminate inPattern, canRemedy, which have no place here.
*/
def checkCheckable(tree: Tree, P0: Type, X0: Type, inPattern: Boolean, canRemedy: Boolean = false) {
if (uncheckedOk(P0)) return
def where = if (inPattern) "pattern " else ""
// singleton types not considered here
val P = P0.widen
val X = X0.widen
P match {
// Prohibit top-level type tests for these, but they are ok nested (e.g. case Foldable[Nothing] => ... )
case TypeRef(_, NothingClass | NullClass | AnyValClass, _) =>
InferErrorGen.TypePatternOrIsInstanceTestError(tree, P)
// If top-level abstract types can be checked using a classtag extractor, don't warn about them
case TypeRef(_, sym, _) if sym.isAbstractType && canRemedy =>
;
// Matching on types like case _: AnyRef { def bippy: Int } => doesn't work -- yet.
case RefinedType(_, decls) if !decls.isEmpty =>
getContext.unit.warning(tree.pos, s"a pattern match on a refinement type is unchecked")
case _ =>
val checker = new CheckabilityChecker(X, P)
log(checker.summaryString)
if (checker.neverMatches) {
val addendum = if (checker.neverSubClass) "" else " (but still might match its erasure)"
getContext.unit.warning(tree.pos, s"fruitless type test: a value of type $X cannot also be a $P$addendum")
}
else if (checker.isUncheckable) {
val msg = (
if (checker.uncheckableType =:= P) s"abstract type $where$P"
else s"${checker.uncheckableMessage} in type $where$P"
)
getContext.unit.warning(tree.pos, s"$msg is unchecked since it is eliminated by erasure")
}
}
}
}
}
private[typechecker] final class Checkability(val value: Int) extends AnyVal { }
private[typechecker] object Checkability {
val StaticallyTrue = new Checkability(0)
val StaticallyFalse = new Checkability(1)
val RuntimeCheckable = new Checkability(2)
val Uncheckable = new Checkability(3)
val CheckabilityError = new Checkability(4)
}
|