-
Notifications
You must be signed in to change notification settings - Fork 2
/
README
253 lines (176 loc) · 6.97 KB
/
README
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
Hindley-Milner with contracts.
Class assignment for dvh's fall '14 cmsc631.
labichn
E ::= n # some natural number
| true
| false
| x # some variable name
| (add1 E)
| (+ E E)
| (* E E)
| (/ E E)
| (= E E)
| (and E E)
| (not E)
| (or E E)
| (if E E E)
| (lambda x E)
| (let (x E) E)
| (? E)
| (-> E E)
| (! E E)
| (E E)
T ::= B
| N # includes divide by zero errors
| T -> T
| T/c
| X
TE ∈ ℘(T * T)
EQ ∈ ℘(T * T)
XS ∈ ℘(X)
The notation `T = T'` denotes that types T and T' are equal and is
represented as a pair of types T and T'. The notation `T if EQ`
represents some type T if the set of type equality assertions EQ can
be unified and is represented as a pair of types T and EQ. Most work
of the type system is shunted off to unification. Unification is
implemented as per Van Horn's /Program Analysis and Understanding/
with some extensions detailed below.
The ternary typing relation _ |- _ : _ ∈ TE * E * (B * EQ) is defined
as follows:
Base leaves:
-------------------- true
TE |- true : B if {}
--------------------- false
TE |- false : B if {}
----------------- n
TE |- n : N if {}
`unary` and `binary` handle tedious patterns for expressions that
always return base types:
TE |- E : T' if EQ
----------------------------------- base unary
unary T TE E = T if EQ U (T = T')
TE |- E : T' if EQ TE |- E' : T'' if EQ'
------------------------------------------------------- base binary
binary T TE E E' = T if EQ U EQ' U (T = T') U (T = T'')
Base operations:
N if EQ = unary N TE E
------------------------ add1
TE |- (add1 E) : N if EQ
N if EQ = binary N TE E E'
-------------------------- *
TE |- (* E E') : N if EQ
N if EQ = binary N TE E E'
-------------------------- +
TE |- (+ E E') : N if EQ
N if EQ = binary N TE E E'
-------------------------- /
TE |- (/ E E') : N if EQ
B if EQ = unary B TE E
----------------------- not
TE |- (not E) : B if EQ
B if EQ = binary B TE E E'
-------------------------- or
TE |- (or E E') : B if EQ
B if EQ = binary B TE E E'
-------------------------- and
TE |- (and E E') : B if EQ
Equality is defined for everything and always returns a boolean. Bit
of a cop-out, but we'll just assume that function equality is defined
to be false.
TE |- E : T if EQ TE |- E' : T' if EQ
--------------------------------------- =
(= E E') |- B if EQ U EQ'
For if, the predicate must be in B and the branches must have the same
type.
TE |- E' : T' if EQ' TE |- E'' : T'' if EQ'' TE |- E : T if EQ
------------------------------------------------------------------- if
TE |- (if E E' E'') : T' if EQ U EQ' U EQ'' U (T = B) U (T' = T'')
The metafunction `genvar` always returns a fresh type variable. It is
used here to generate a new type variable for lambda's input,
T = genvar TE, X : T |- E : T' if EQ
-------------------------------------- lambda
TE |- (lambda X E) : T -> T' if EQ
`genvar` is used again to generate a type variable for the result of
an application.
TE |- E : T if EQ TE |- E' : T' if EQ' T'' = genvar
------------------------------------------------------- app
TE |- (E E') : T'' if EQ U EQ' U (T = T' -> T'')
The metafunction `fv` is defined using structural recursion over the
given type:
fv(X) = {X}
fv(T -> T') = fv(T) U fv(T')
fv(T/c) = fv(T)
fv(∀ X ... . T) = fv(T) - BS ...
fv(_) = {} otherwise
`fv` can be lifted to TE as the union of the free variables of every
element in its codomain.
The metafunction `gen` generalizes some type T in environment TE as
follows:
gen(TE, T) = ∀ (fv T) - (fv TE) . T
and enables let polymorphism in the following rule.
TE |- E : T if EQ TE, X : gen(TE, T) |- E' : T' if EQ'
-------------------------------------------------------- let
TE |- (let (X E) E') : T' if EQ U EQ'
Typing contracts is easy. Predicates generate a new assertion that E's
type is a function from some type T to B.
T = genvar TE |- E : T' if EQ
--------------------------------------- pred contract
TE |- (? E) : T/c if EQ U (T' = T -> B)
Contract arrows generate new type variables T'' and T''' for the type
parameters of the contract types T and T', respectively.
TE |- E : T if EQ TE |- E' : T' if EQ'
T'' = genvar T''' = genvar
--------------------------------------------- arr contract
TE |- (-> E E') : (T'' -> T''')/c
if EQ U EQ' U (T'' = T/c) U (T''' = T''/c)
Monitors generate a new type variable T'' for the type parameter of
the contract type T.
TE |- E : T if EQ TE |- E' : T' if EQ'
T'' = genvar
--------------------------------------------- monitor
TE |- (! E E') : T' if EQ U EQ' U (T = T''/c)
Beta reduction is defined for types as follows:
β(T -> T', X, U) = β(T, X, U) -> β(T', X, U)
β(T/c, X, U) = β(T, X, U)/c
β(Y, X, U) = U when X = Y
β(Y, X, U) = X when X != Y
β(∀ YS . T, X, U) = ∀ YS . T when X ∈ YS
β(∀ YS . T, X, U) = ∀ YS . β(T, X, U) when X ∉ YS
β(T, X, U) = T otherwise
Variables mapping to polytypes are instantiated with the metafunction
`inst`, iteratively generating fresh variables for all bound variables
using `β`. Monotypes are just returned:
inst(∀ XS . T) = inst'(XS, T)
inst(T) = T otherwise
inst'({}, T) = T
inst'(X U XS, T) = inst'(XS, β(T, X, genvar))
Finally the typing rule for variables:
X : T ∈ TE T' = inst(T)
------------------------- var
TE |- X : T' if {}
After all this we have some type T if EQ can be unified. Unification
is extended componentwise through arrows and contracts, and handles
recursive types by checking if a variable occurs in its own type
expansion:
occurs(TE, X, T -> T') = occurs(TE, X, T) or occurs(TE, X, T')
occurs(TE, X, T/c) = occurs(TE, X, T)
occurs(TE, X, Y) = X = Y
occurs(TE, X, _) = false otherwise
unify(T) = unify'(EQ, T)
unify'( {}, TE) = TE
unify'( (T = T') U EQ, TE) = unify'(EQ, TE)
when T = T'
unify'( (X = T) U EQ, TE) = unify'(EQ U (T = TE(X)), TE)
when not occurs(TE, X, T)
unify'( (T = X) U EQ, TE) = unify'(EQ U (X = T), TE)
unify'( (T/c = T'/c) U EQ, TE) = unify'(EQ U (T = T'), TE)
unify'((T -> U = T' -> U') U EQ, TE) = unify'(EQ U (T = T') U (U = U'), TE)
References:
Clement, (1987). The Natural Dynamic Semantics of Mini-Standard
ML. TAPSOFT'87, Vol 2. LNCS, Vol. 250, pp 67–81
Damas, Milner (1982), "Principal type-schemes for functional
programs". 9th Symposium on Principles of programming languages
(POPL'82) pp. 207–212, ACM
Findler, Robert Bruce, and Matthias Felleisen. "Contracts for
higher-order functions." ACM SIGPLAN Notices. Vol. 37. No. 9. ACM,
2002. Van Horn, Program Analysis and Understanding