1	import Base
2	import bend-kit-bytes@0.3.2.0/bytes.bend as Bytes
3	
4	# State for the strict single-pass UTF-8 decoder.
5	type Utf8State is Data:
6	  Utf8State{need: U32, lo: U32, hi: U32, cp: U32, out: String, valid: Bool}
7	
8	def utf8.lead(+byte: U32, +out: String) -> Utf8State:
9	  Utf8State{
10	    Bool.pick(U32,
11	      U32.is_lt(byte, 128), 0,
12	      Bool.pick(U32, U32.is_le(194, byte) && U32.is_le(byte, 223), 1,
13	        Bool.pick(U32, U32.is_eq(byte, 224) || U32.is_le(225, byte) && U32.is_le(byte, 236) || U32.is_le(238, byte) && U32.is_le(byte, 239) || U32.is_eq(byte, 237), 2,
14	          Bool.pick(U32, U32.is_le(240, byte) && U32.is_le(byte, 244), 3, 0)))),
15	    Bool.pick(U32, U32.is_eq(byte, 224), 160,
16	      Bool.pick(U32, U32.is_eq(byte, 240), 144, 128)),
17	    Bool.pick(U32, U32.is_eq(byte, 237), 159,
18	      Bool.pick(U32, U32.is_eq(byte, 244), 143, 191)),
19	    Bool.pick(U32,
20	      U32.is_lt(byte, 128), byte,
21	      Bool.pick(U32, U32.is_le(194, byte) && U32.is_le(byte, 223), U32.and(byte, 31),
22	        Bool.pick(U32, U32.is_le(224, byte) && U32.is_le(byte, 239), U32.and(byte, 15),
23	          Bool.pick(U32, U32.is_le(240, byte) && U32.is_le(byte, 244), U32.and(byte, 7), 0)))),
24	    Bool.pick(String, U32.is_lt(byte, 128), SCon{Chr{byte}, out}, out),
25	    U32.is_lt(byte, 128) || U32.is_le(194, byte) && U32.is_le(byte, 244)
26	  }
27	
28	def utf8.cont(+need: U32, cp: U32, +out: String, ok: Bool, in_range: Bool, byte: U32) -> Utf8State:
29	  match in_range:
30	    case True{}:
31	      +next_cp = (cp * 64 + U32.and(byte, 63) : U32)
32	      Utf8State{(need - 1 : U32), 128, 191, next_cp,
33	        Bool.pick(String, U32.is_eq(need, 1), SCon{Chr{next_cp}, out}, out), ok}
34	    case False{}:
35	      Utf8State{0, 128, 191, 0, out, False{}}
36	
37	def utf8.step(state: Utf8State, +byte: U32) -> Utf8State:
38	  match state:
39	    case Utf8State{0, _, _, _, out, True{}}:
40	      utf8.lead(byte, out)
41	    case Utf8State{0, _, _, _, out, False{}}:
42	      Utf8State{0, 128, 191, 0, out, False{}}
43	    case Utf8State{need, lo, hi, cp, out, ok}:
44	      utf8.cont(need, cp, out, ok, U32.is_le(lo, byte) && U32.is_le(byte, hi), byte)
45	
46	def utf8.scan.fin.done(out: String, complete: Bool) -> Result<&1, &1, U32 & String, String>:
47	  match complete:
48	    case True{}:
49	      Done{String.reverse(out)}
50	    case False{}:
51	      Fail{(1, "invalid UTF-8")}
52	
53	def utf8.scan.fin(state: Utf8State) -> Result<&1, &1, U32 & String, String>:
54	  match state:
55	    case Utf8State{need, _, _, _, out, True{}}:
56	      utf8.scan.fin.done(out, U32.is_eq(need, 0))
57	    case Utf8State{_, _, _, _, _, False{}}:
58	      Fail{(1, "invalid UTF-8")}
59	
60	def utf8.scan(
61	  remaining: Nat,
62	  +offset: U32,
63	  pair: Array<U32> & U32,
64	  state: Utf8State
65	) -> Result<&1, &1, U32 & String, String>:
66	  match remaining:
67	    case 0n:
68	      match pair:
69	        case (_, _):
70	          utf8.scan.fin(state)
71	    case 1n+p:
72	      match pair:
73	        case (buf, b):
74	          utf8.scan(p, (offset + 1 : U32), Bytes.peek(buf, (offset + 1 : U32)), utf8.step(state, b))
75	
76	# Return None rather than wrapping when U32 addition overflows.
77	def checked_add(+left: U32, +right: U32) -> Maybe<&2, U32>:
78	  Bool.pick(Maybe<&2, U32>, U32.is_le(left, (4294967295 - right : U32)), Some{(left + right : U32)}, None{})
79	
80	# Normal target size for an encoded storage block or WAL frame.
81	def MAX_BLOCK_BYTES() -> U32:
82	  1048576
83	
84	# Size at which the SST encoder closes a block and starts the next one.
85	def TARGET_BLOCK_BYTES() -> U32:
86	  4096
87	
88	# Hard cap for an individual encoded key or value.
89	def MAX_RECORD_BYTES() -> U32:
90	  1073741824
91	
92	# Hard cap for the complete serialized Manifest.
93	def MAX_MANIFEST_BYTES() -> U32:
94	  1048576
95	
96	# Checks a record against the storage codec byte cap.
97	def record_bytes_allowed(+length: U32) -> Bool:
98	  U32.is_le(length, MAX_RECORD_BYTES())
99	
100	def utf8.width(+scalar: U32) -> U32:
101	  Bool.pick(U32, U32.is_le(scalar, 127), 1,
102	    Bool.pick(U32, U32.is_le(scalar, 2047), 2,
103	      Bool.pick(U32, U32.is_le(scalar, 65535), 3, 4)))
104	
105	# Returns a string length as a storage byte count.
106	def string_size(+text: String) -> Nat:
107	  match text:
108	    case SNil{}:
109	      0n
110	    case SCon{Chr{scalar}, rest}:
111	      Nat.add(U32.to_nat(utf8.width(scalar)), string_size(rest))
112	
113	def from_string.size(+text: String, +size: U32, fits: Bool) -> Maybe<&2, U32>:
114	  match text fits:
115	    case SNil{} _:
116	      Some{size}
117	    case SCon{Chr{_scalar}, SNil{}} False{}:
118	      None{}
119	    case SCon{Chr{scalar}, SNil{}} True{}:
120	      Some{(size + utf8.width(scalar) : U32)}
121	    case SCon{Chr{_scalar}, SCon{Chr{_next_scalar}, _rest}} False{}:
122	      None{}
123	    case SCon{Chr{scalar}, SCon{Chr{next_scalar}, rest}} True{}:
124	      from_string.size(SCon{Chr{next_scalar}, rest}, (size + utf8.width(scalar) : U32),
125	        U32.is_le((size + utf8.width(scalar) : U32),
126	          (MAX_RECORD_BYTES() - utf8.width(next_scalar) : U32)))
127	
128	def from_string.size.begin(+text: String) -> Maybe<&2, U32>:
129	  match text:
130	    case SNil{}:
131	      Some{0}
132	    case SCon{Chr{scalar}, _}:
133	      from_string.size(text, 0,
134	        U32.is_le(0, (MAX_RECORD_BYTES() - utf8.width(scalar) : U32)))
135	
136	def utf8.put.result(pair: Bytes.Cursor & Bool) -> Bytes.Cursor:
137	  match pair:
138	    case (cursor, _):
139	      cursor
140	
141	def utf8.put.byte(cursor: Bytes.Cursor, +byte: U32) -> Bytes.Cursor:
142	  utf8.put.result(Bytes.Cursor.put.u8(cursor, byte))
143	
144	def utf8.put.two(cursor: Bytes.Cursor, +scalar: U32) -> Bytes.Cursor:
145	  utf8.put.byte(utf8.put.byte(cursor, (192 + U32.shrn(scalar, 6n) : U32)),
146	    (128 + U32.and(scalar, 63) : U32))
147	
148	def utf8.put.three(cursor: Bytes.Cursor, +scalar: U32) -> Bytes.Cursor:
149	  utf8.put.byte(utf8.put.byte(
150	    utf8.put.byte(cursor, (224 + U32.shrn(scalar, 12n) : U32)),
151	    (128 + U32.and(U32.shrn(scalar, 6n), 63) : U32)),
152	    (128 + U32.and(scalar, 63) : U32))
153	
154	def utf8.put.four(cursor: Bytes.Cursor, +scalar: U32) -> Bytes.Cursor:
155	  utf8.put.byte(utf8.put.byte(
156	    utf8.put.byte(utf8.put.byte(cursor, (240 + U32.shrn(scalar, 18n) : U32)),
157	      (128 + U32.and(U32.shrn(scalar, 12n), 63) : U32)),
158	    (128 + U32.and(U32.shrn(scalar, 6n), 63) : U32)),
159	    (128 + U32.and(scalar, 63) : U32))
160	
161	def utf8.put.scalar(
162	  cursor: Bytes.Cursor,
163	  +scalar: U32,
164	  ascii: Bool,
165	  two_byte: Bool,
166	  three_byte: Bool
167	) -> Bytes.Cursor:
168	  match ascii two_byte three_byte:
169	    case True{} _ _:
170	      utf8.put.byte(cursor, scalar)
171	    case False{} True{} _:
172	      utf8.put.two(cursor, scalar)
173	    case False{} False{} True{}:
174	      utf8.put.three(cursor, scalar)
175	    case False{} False{} False{}:
176	      utf8.put.four(cursor, scalar)
177	
178	def utf8.encode.go(text: String, cursor: Bytes.Cursor) -> Bytes.Cursor:
179	  match text:
180	    case SNil{}:
181	      cursor
182	    case SCon{Chr{+scalar}, tail}:
183	      utf8.encode.go(tail, utf8.put.scalar(cursor, scalar,
184	        U32.is_le(scalar, 127), U32.is_le(scalar, 2047), U32.is_le(scalar, 65535)))
185	
186	def from_string.output.pair(result: Bytes.Bytes & U32) -> Bytes.Bytes:
187	  match result:
188	    case (bytes, _):
189	      bytes
190	
191	def from_string.output(+text: String, +len: U32) -> Bytes.Bytes:
192	  from_string.output.pair(Bytes.Cursor.finish(utf8.encode.go(text,
193	    Bytes.Cursor.new(Bytes.Bytes{len, Bytes.alloc(len)}))))
194	
195	def from_string.preflight(+text: String, size: Maybe<&2, U32>) -> Result<&1, &1, U32 & String, Bytes.Bytes>:
196	  match size:
197	    case None{}:
198	      Fail{(2, "record too large")}
199	    case Some{len}:
200	      Done{from_string.output(text, len)}
201	
202	# Encode public text at the packed storage boundary.
203	def from_string(+text: String) -> Result<&1, &1, U32 & String, Bytes.Bytes>:
204	  from_string.preflight(text, from_string.size.begin(text))
205	
206	# Decode bytes after strict RFC 3629 validation.
207	def decode_checked(bytes: Bytes.Bytes) -> Result<&1, &1, U32 & String, String>:
208	  match bytes:
209	    case Bytes.Bytes{len, buf}:
210	      utf8.scan(U32.to_nat(len), 0, Bytes.peek(buf, 0), Utf8State{0, 128, 191, 0, SNil{}, True{}})
211	
212	# Decode a packed buffer, rejecting malformed UTF-8.
213	def to_string_strict(bytes: Bytes.Bytes) -> Result<&1, &1, U32 & String, String>:
214	  decode_checked(bytes)
215	
216	# Read one bounded big-endian u32 through a package cursor.
217	def read_u32be(cursor: Bytes.Cursor) -> Bytes.Cursor & Maybe<&2, U32>:
218	  Bytes.Cursor.u32be(cursor)
219	
220	# Write one bounded big-endian u32 through a package cursor.
221	def write_u32be(cursor: Bytes.Cursor, value: U32) -> Bytes.Cursor & Bool:
222	  Bytes.Cursor.put.u32be(cursor, value)