github.com/AndrienkoAleksandr/go@v0.0.19/src/go/types/validtype.go (about)

     1  // Code generated by "go test -run=Generate -write=all"; DO NOT EDIT.
     2  
     3  // Copyright 2022 The Go Authors. All rights reserved.
     4  // Use of this source code is governed by a BSD-style
     5  // license that can be found in the LICENSE file.
     6  
     7  package types
     8  
     9  // validType verifies that the given type does not "expand" indefinitely
    10  // producing a cycle in the type graph.
    11  // (Cycles involving alias types, as in "type A = [10]A" are detected
    12  // earlier, via the objDecl cycle detection mechanism.)
    13  func (check *Checker) validType(typ *Named) {
    14  	check.validType0(typ, nil, nil)
    15  }
    16  
    17  // validType0 checks if the given type is valid. If typ is a type parameter
    18  // its value is looked up in the type argument list of the instantiated
    19  // (enclosing) type, if it exists. Otherwise the type parameter must be from
    20  // an enclosing function and can be ignored.
    21  // The nest list describes the stack (the "nest in memory") of types which
    22  // contain (or embed in the case of interfaces) other types. For instance, a
    23  // struct named S which contains a field of named type F contains (the memory
    24  // of) F in S, leading to the nest S->F. If a type appears in its own nest
    25  // (say S->F->S) we have an invalid recursive type. The path list is the full
    26  // path of named types in a cycle, it is only needed for error reporting.
    27  func (check *Checker) validType0(typ Type, nest, path []*Named) bool {
    28  	switch t := typ.(type) {
    29  	case nil:
    30  		// We should never see a nil type but be conservative and panic
    31  		// only in debug mode.
    32  		if debug {
    33  			panic("validType0(nil)")
    34  		}
    35  
    36  	case *Array:
    37  		return check.validType0(t.elem, nest, path)
    38  
    39  	case *Struct:
    40  		for _, f := range t.fields {
    41  			if !check.validType0(f.typ, nest, path) {
    42  				return false
    43  			}
    44  		}
    45  
    46  	case *Union:
    47  		for _, t := range t.terms {
    48  			if !check.validType0(t.typ, nest, path) {
    49  				return false
    50  			}
    51  		}
    52  
    53  	case *Interface:
    54  		for _, etyp := range t.embeddeds {
    55  			if !check.validType0(etyp, nest, path) {
    56  				return false
    57  			}
    58  		}
    59  
    60  	case *Named:
    61  		// Exit early if we already know t is valid.
    62  		// This is purely an optimization but it prevents excessive computation
    63  		// times in pathological cases such as testdata/fixedbugs/issue6977.go.
    64  		// (Note: The valids map could also be allocated locally, once for each
    65  		// validType call.)
    66  		if check.valids.lookup(t) != nil {
    67  			break
    68  		}
    69  
    70  		// Don't report a 2nd error if we already know the type is invalid
    71  		// (e.g., if a cycle was detected earlier, via under).
    72  		// Note: ensure that t.orig is fully resolved by calling Underlying().
    73  		if t.Underlying() == Typ[Invalid] {
    74  			return false
    75  		}
    76  
    77  		// If the current type t is also found in nest, (the memory of) t is
    78  		// embedded in itself, indicating an invalid recursive type.
    79  		for _, e := range nest {
    80  			if Identical(e, t) {
    81  				// We have a cycle. If t != t.Origin() then t is an instance of
    82  				// the generic type t.Origin(). Because t is in the nest, t must
    83  				// occur within the definition (RHS) of the generic type t.Origin(),
    84  				// directly or indirectly, after expansion of the RHS.
    85  				// Therefore t.Origin() must be invalid, no matter how it is
    86  				// instantiated since the instantiation t of t.Origin() happens
    87  				// inside t.Origin()'s RHS and thus is always the same and always
    88  				// present.
    89  				// Therefore we can mark the underlying of both t and t.Origin()
    90  				// as invalid. If t is not an instance of a generic type, t and
    91  				// t.Origin() are the same.
    92  				// Furthermore, because we check all types in a package for validity
    93  				// before type checking is complete, any exported type that is invalid
    94  				// will have an invalid underlying type and we can't reach here with
    95  				// such a type (invalid types are excluded above).
    96  				// Thus, if we reach here with a type t, both t and t.Origin() (if
    97  				// different in the first place) must be from the current package;
    98  				// they cannot have been imported.
    99  				// Therefore it is safe to change their underlying types; there is
   100  				// no chance for a race condition (the types of the current package
   101  				// are not yet available to other goroutines).
   102  				assert(t.obj.pkg == check.pkg)
   103  				assert(t.Origin().obj.pkg == check.pkg)
   104  				t.underlying = Typ[Invalid]
   105  				t.Origin().underlying = Typ[Invalid]
   106  
   107  				// Find the starting point of the cycle and report it.
   108  				// Because each type in nest must also appear in path (see invariant below),
   109  				// type t must be in path since it was found in nest. But not every type in path
   110  				// is in nest. Specifically t may appear in path with an earlier index than the
   111  				// index of t in nest. Search again.
   112  				for start, p := range path {
   113  					if Identical(p, t) {
   114  						check.cycleError(makeObjList(path[start:]))
   115  						return false
   116  					}
   117  				}
   118  				panic("cycle start not found")
   119  			}
   120  		}
   121  
   122  		// No cycle was found. Check the RHS of t.
   123  		// Every type added to nest is also added to path; thus every type that is in nest
   124  		// must also be in path (invariant). But not every type in path is in nest, since
   125  		// nest may be pruned (see below, *TypeParam case).
   126  		if !check.validType0(t.Origin().fromRHS, append(nest, t), append(path, t)) {
   127  			return false
   128  		}
   129  
   130  		check.valids.add(t) // t is valid
   131  
   132  	case *TypeParam:
   133  		// A type parameter stands for the type (argument) it was instantiated with.
   134  		// Check the corresponding type argument for validity if we are in an
   135  		// instantiated type.
   136  		if len(nest) > 0 {
   137  			inst := nest[len(nest)-1] // the type instance
   138  			// Find the corresponding type argument for the type parameter
   139  			// and proceed with checking that type argument.
   140  			for i, tparam := range inst.TypeParams().list() {
   141  				// The type parameter and type argument lists should
   142  				// match in length but be careful in case of errors.
   143  				if t == tparam && i < inst.TypeArgs().Len() {
   144  					targ := inst.TypeArgs().At(i)
   145  					// The type argument must be valid in the enclosing
   146  					// type (where inst was instantiated), hence we must
   147  					// check targ's validity in the type nest excluding
   148  					// the current (instantiated) type (see the example
   149  					// at the end of this file).
   150  					// For error reporting we keep the full path.
   151  					return check.validType0(targ, nest[:len(nest)-1], path)
   152  				}
   153  			}
   154  		}
   155  	}
   156  
   157  	return true
   158  }
   159  
   160  // makeObjList returns the list of type name objects for the given
   161  // list of named types.
   162  func makeObjList(tlist []*Named) []Object {
   163  	olist := make([]Object, len(tlist))
   164  	for i, t := range tlist {
   165  		olist[i] = t.obj
   166  	}
   167  	return olist
   168  }
   169  
   170  // Here is an example illustrating why we need to exclude the
   171  // instantiated type from nest when evaluating the validity of
   172  // a type parameter. Given the declarations
   173  //
   174  //   var _ A[A[string]]
   175  //
   176  //   type A[P any] struct { _ B[P] }
   177  //   type B[P any] struct { _ P }
   178  //
   179  // we want to determine if the type A[A[string]] is valid.
   180  // We start evaluating A[A[string]] outside any type nest:
   181  //
   182  //   A[A[string]]
   183  //         nest =
   184  //         path =
   185  //
   186  // The RHS of A is now evaluated in the A[A[string]] nest:
   187  //
   188  //   struct{_ B[P₁]}
   189  //         nest = A[A[string]]
   190  //         path = A[A[string]]
   191  //
   192  // The struct has a single field of type B[P₁] with which
   193  // we continue:
   194  //
   195  //   B[P₁]
   196  //         nest = A[A[string]]
   197  //         path = A[A[string]]
   198  //
   199  //   struct{_ P₂}
   200  //         nest = A[A[string]]->B[P]
   201  //         path = A[A[string]]->B[P]
   202  //
   203  // Eventually we reach the type parameter P of type B (P₂):
   204  //
   205  //   P₂
   206  //         nest = A[A[string]]->B[P]
   207  //         path = A[A[string]]->B[P]
   208  //
   209  // The type argument for P of B is the type parameter P of A (P₁).
   210  // It must be evaluated in the type nest that existed when B was
   211  // instantiated:
   212  //
   213  //   P₁
   214  //         nest = A[A[string]]        <== type nest at B's instantiation time
   215  //         path = A[A[string]]->B[P]
   216  //
   217  // If we'd use the current nest it would correspond to the path
   218  // which will be wrong as we will see shortly. P's type argument
   219  // is A[string], which again must be evaluated in the type nest
   220  // that existed when A was instantiated with A[string]. That type
   221  // nest is empty:
   222  //
   223  //   A[string]
   224  //         nest =                     <== type nest at A's instantiation time
   225  //         path = A[A[string]]->B[P]
   226  //
   227  // Evaluation then proceeds as before for A[string]:
   228  //
   229  //   struct{_ B[P₁]}
   230  //         nest = A[string]
   231  //         path = A[A[string]]->B[P]->A[string]
   232  //
   233  // Now we reach B[P] again. If we had not adjusted nest, it would
   234  // correspond to path, and we would find B[P] in nest, indicating
   235  // a cycle, which would clearly be wrong since there's no cycle in
   236  // A[string]:
   237  //
   238  //   B[P₁]
   239  //         nest = A[string]
   240  //         path = A[A[string]]->B[P]->A[string]  <== path contains B[P]!
   241  //
   242  // But because we use the correct type nest, evaluation proceeds without
   243  // errors and we get the evaluation sequence:
   244  //
   245  //   struct{_ P₂}
   246  //         nest = A[string]->B[P]
   247  //         path = A[A[string]]->B[P]->A[string]->B[P]
   248  //   P₂
   249  //         nest = A[string]->B[P]
   250  //         path = A[A[string]]->B[P]->A[string]->B[P]
   251  //   P₁
   252  //         nest = A[string]
   253  //         path = A[A[string]]->B[P]->A[string]->B[P]
   254  //   string
   255  //         nest =
   256  //         path = A[A[string]]->B[P]->A[string]->B[P]
   257  //
   258  // At this point we're done and A[A[string]] and is valid.