# JSON values, parsed and encoded as RFC 8259. Source: https://github.com/paymog/bend-kit/tree/main/json import Base import 0x49814d83de8f70993a43e1002be29ecd/bytes.bend as Bytes # JSON (RFC 8259). parse and encode work on text; parse.bytes and encode.bytes # work on UTF-8 in Bytes, with no String of the whole text. Nesting has no # depth limit other than input size: the parser keeps its own stack, since a # Bend def cannot call itself through another def. # import ./json/json.bend as Json # Num keeps the number's exact text, so no precision is lost. type Val is Data: Null{} Flag{on: Bool} Num{s: String} Str{s: String} Arr{xs: List<&2, Val>} Obj{m: Map<&2, Val>} type Next is Data: Next{v: Val, rest: String} # An open container: array items so far (reversed), or object fields and the # key whose value comes next. type Open is Data: OArr{xs: List<&2, Val>} OObj{m: Map<&2, Val>, key: String} type St is Data: SValue{s: String, stack: List<&2, Open>} SKey{s: String, stack: List<&2, Open>} SColon{key: String, s: String, stack: List<&2, Open>} SStr{s: String, acc: String, key: Bool, stack: List<&2, Open>} SAfter{v: Val, s: String, stack: List<&2, Open>} SDone{v: Val, s: String} SFail{} # Whitespace # ---------- def is_ws(+c: U32) -> Bool: Bool.or(U32.is_eq(c, 32), Bool.or(U32.is_eq(c, 9), Bool.or(U32.is_eq(c, 10), U32.is_eq(c, 13)))) # Bool.pick runs both arms, so loops peek: the next char's class is a Bool # param and only the matching arm recurses. def ws.go(t: String, +c: U32, ws: Bool) -> String: match t: case SNil{}: match ws: case True{}: SNil{} case False{}: SCon{Chr{c}, SNil{}} case SCon{Chr{+d}, u}: match ws: case True{}: ws.go(u, d, is_ws(d)) case False{}: SCon{Chr{c}, SCon{Chr{d}, u}} def skip_ws(s: String) -> String: match s: case SNil{}: SNil{} case SCon{Chr{+c}, t}: ws.go(t, c, is_ws(c)) # Numbers (§6) # ------------ # States: 0 start, 1 after "-", 2 after a leading "0", 3 int digits, 4 after # ".", 5 fraction digits, 6 after "e", 7 after the exponent sign, 8 exponent # digits. 99 means the char does not continue the number. def digit(+c: U32) -> Bool: Bool.and(U32.is_le(48, c), U32.is_le(c, 57)) def num.trans(+st: U32, +c: U32) -> U32: +d = digit(c) +z = U32.is_eq(c, 48) +dot = U32.is_eq(c, 46) +e = Bool.or(U32.is_eq(c, 101), U32.is_eq(c, 69)) +sign = Bool.or(U32.is_eq(c, 43), U32.is_eq(c, 45)) Bool.pick(U32, U32.is_eq(st, 0), Bool.pick(U32, U32.is_eq(c, 45), 1, Bool.pick(U32, z, 2, Bool.pick(U32, d, 3, 99))), Bool.pick(U32, U32.is_eq(st, 1), Bool.pick(U32, z, 2, Bool.pick(U32, d, 3, 99)), Bool.pick(U32, Bool.or(U32.is_eq(st, 2), U32.is_eq(st, 3)), Bool.pick(U32, Bool.and(d, U32.is_eq(st, 3)), 3, Bool.pick(U32, dot, 4, Bool.pick(U32, e, 6, 99))), Bool.pick(U32, Bool.or(U32.is_eq(st, 4), U32.is_eq(st, 5)), Bool.pick(U32, d, 5, Bool.pick(U32, Bool.and(e, U32.is_eq(st, 5)), 6, 99)), Bool.pick(U32, U32.is_eq(st, 6), Bool.pick(U32, sign, 7, Bool.pick(U32, d, 8, 99)), Bool.pick(U32, Bool.or(U32.is_eq(st, 7), U32.is_eq(st, 8)), Bool.pick(U32, d, 8, 99), 99)))))) def num.accepts(+st: U32) -> Bool: Bool.or(Bool.or(U32.is_eq(st, 2), U32.is_eq(st, 3)), Bool.or(U32.is_eq(st, 5), U32.is_eq(st, 8))) def num.end(+st: U32, acc: String, rest: String) -> Maybe<&2, Next>: Bool.pick(Maybe<&2, Next>, num.accepts(st), Some{Next{Num{String.reverse(acc)}, rest}}, None{}) def num.go(t: String, +c: U32, +st: U32, acc: String, stop: Bool) -> Maybe<&2, Next>: match t: case SNil{}: match stop: case True{}: num.end(st, acc, SCon{Chr{c}, SNil{}}) case False{}: num.end(num.trans(st, c), SCon{Chr{c}, acc}, SNil{}) case SCon{Chr{+d}, u}: match stop: case True{}: num.end(st, acc, SCon{Chr{c}, SCon{Chr{d}, u}}) case False{}: +n = num.trans(st, c) num.go(u, d, n, SCon{Chr{c}, acc}, U32.is_eq(num.trans(n, d), 99)) def num(s: String) -> Maybe<&2, Next>: match s: case SNil{}: None{} case SCon{Chr{+c}, t}: num.go(t, c, 0, SNil{}, U32.is_eq(num.trans(0, c), 99)) # Literals # -------- def lit.if(+s: String, v: Val, n: Nat, ok: Bool) -> Maybe<&2, Next>: match ok: case False{}: None{} case True{}: Some{Next{v, String.drop(s, n)}} def lit(+s: String, word: String, v: Val, +n: Nat) -> Maybe<&2, Next>: lit.if(s, v, n, String.starts_with(s, word)) # Strings (§7) # ------------ type Esc is Data: Esc{cp: U32, rest: String} def hexd(+c: U32) -> U32: Bool.pick(U32, digit(c), (c - 48 : U32), Bool.pick(U32, Bool.and(U32.is_le(97, c), U32.is_le(c, 102)), (c - 87 : U32), Bool.pick(U32, Bool.and(U32.is_le(65, c), U32.is_le(c, 70)), (c - 55 : U32), 99))) def hex4.go(s: String, +acc: U32, n: Nat) -> Maybe<&2, Esc>: match s: case SNil{}: match n: case 0n: Some{Esc{acc, SNil{}}} case 1n+k: None{} case SCon{Chr{+c}, t}: match n: case 0n: Some{Esc{acc, SCon{Chr{c}, t}}} case 1n+k: +h = hexd(c) Bool.pick(Maybe<&2, Esc>, U32.is_eq(h, 99), None{}, hex4.go(t, (acc * 16 + h : U32), k)) def hex4(s: String) -> Maybe<&2, Esc>: hex4.go(s, 0, 4n) def is_hi(+v: U32) -> Bool: Bool.and(U32.is_le(55296, v), U32.is_le(v, 56319)) def is_lo(+v: U32) -> Bool: Bool.and(U32.is_le(56320, v), U32.is_le(v, 57343)) def pair.lo(+hi: U32, +rest: String, lo: Maybe<&2, Esc>) -> Esc: match lo: case None{}: Esc{65533, rest} case Some{Esc{+v, r}}: Bool.pick(Esc, is_lo(v), Esc{(65536 + (hi - 55296 : U32) * 1024 + (v - 56320 : U32) : U32), r}, Esc{65533, rest}) # §8.2: a high surrogate pairs with a following \uDC00-\uDFFF; a lone one is U+FFFD. def pair(+hi: U32, +rest: String) -> Esc: Bool.pick(Esc, String.starts_with(rest, "\\u"), pair.lo(hi, rest, hex4(String.drop(rest, 2n))), Esc{65533, rest}) def esc.u(e: Maybe<&2, Esc>) -> Maybe<&2, Esc>: match e: case None{}: None{} case Some{Esc{+v, +r}}: Some{Bool.pick(Esc, is_hi(v), pair(v, r), Bool.pick(Esc, is_lo(v), Esc{65533, r}, Esc{v, r}))} def esc.simple(+e: U32) -> U32: Bool.pick(U32, U32.is_eq(e, 34), 34, Bool.pick(U32, U32.is_eq(e, 92), 92, Bool.pick(U32, U32.is_eq(e, 47), 47, Bool.pick(U32, U32.is_eq(e, 98), 8, Bool.pick(U32, U32.is_eq(e, 102), 12, Bool.pick(U32, U32.is_eq(e, 110), 10, Bool.pick(U32, U32.is_eq(e, 114), 13, Bool.pick(U32, U32.is_eq(e, 116), 9, 99)))))))) # One escape after the backslash. def esc.one(t: String) -> Maybe<&2, Esc>: match t: case SNil{}: None{} case SCon{Chr{+e}, +u}: +v = esc.simple(e) Bool.pick(Maybe<&2, Esc>, U32.is_eq(e, 117), esc.u(hex4(u)), Bool.pick(Maybe<&2, Esc>, U32.is_eq(v, 99), None{}, Some{Esc{v, u}})) def str.esc(e: Maybe<&2, Esc>, acc: String, key: Bool, stack: List<&2, Open>) -> St: match e: case None{}: SFail{} case Some{Esc{cp, r}}: SStr{r, SCon{Chr{cp}, acc}, key, stack} def str.close(+acc: String, t: String, key: Bool, stack: List<&2, Open>) -> St: match key: case True{}: SColon{String.reverse(acc), t, stack} case False{}: SAfter{Str{String.reverse(acc)}, t, stack} def str.c(quote: Bool, ctl: Bool, bs: Bool, c: U32, t: String, acc: String, key: Bool, stack: List<&2, Open>) -> St: match quote: case True{}: str.close(acc, t, key, stack) case False{}: match ctl: case True{}: SFail{} case False{}: match bs: case True{}: str.esc(esc.one(t), acc, key, stack) case False{}: SStr{t, SCon{Chr{c}, acc}, key, stack} # §7: unescaped chars below U+0020 are not allowed in a string. def str.step(s: String, acc: String, key: Bool, stack: List<&2, Open>) -> St: match s: case SNil{}: SFail{} case SCon{Chr{+c}, t}: str.c(U32.is_eq(c, 34), U32.is_lt(c, 32), U32.is_eq(c, 92), c, t, acc, key, stack) # Machine # ------- def got(m: Maybe<&2, Next>, stack: List<&2, Open>) -> St: match m: case None{}: SFail{} case Some{Next{v, r}}: SAfter{v, r, stack} def open.close(r: String, close: Val, stack: List<&2, Open>, empty: Bool, open: Open) -> St: match empty: case True{}: SAfter{close, String.drop(r, 1n), stack} case False{}: match open: case OArr{xs}: SValue{r, Con{OArr{xs}, stack}} case OObj{m, k}: SKey{r, Con{OObj{m, k}, stack}} def value.step(+s: String, +c: U32, +t: String, +stack: List<&2, Open>) -> St: +r = skip_ws(t) Bool.pick(St, U32.is_eq(c, 91), open.close(r, Arr{Nil{}}, stack, String.starts_with(r, "]"), OArr{Nil{}}), Bool.pick(St, U32.is_eq(c, 123), open.close(r, Obj{Map.new(&2, Val)}, stack, String.starts_with(r, "}"), OObj{Map.new(&2, Val), ""}), Bool.pick(St, U32.is_eq(c, 34), SStr{t, SNil{}, False{}, stack}, Bool.pick(St, U32.is_eq(c, 116), got(lit(s, "true", Flag{True{}}, 4n), stack), Bool.pick(St, U32.is_eq(c, 102), got(lit(s, "false", Flag{False{}}, 5n), stack), Bool.pick(St, U32.is_eq(c, 110), got(lit(s, "null", Null{}, 4n), stack), got(num(s), stack))))))) def value(s: String, stack: List<&2, Open>) -> St: match s: case SNil{}: SFail{} case SCon{Chr{+c}, +t}: value.step(SCon{Chr{c}, t}, c, t, stack) def key(s: String, stack: List<&2, Open>) -> St: match s: case SNil{}: SFail{} case SCon{Chr{+c}, t}: Bool.pick(St, U32.is_eq(c, 34), SStr{t, SNil{}, True{}, stack}, SFail{}) def colon.top(k: String, r: String, stack: List<&2, Open>) -> St: match stack: case Nil{}: SFail{} case Con{OArr{xs}, up}: SFail{} case Con{OObj{m, old}, up}: SValue{r, Con{OObj{m, k}, up}} def colon(k: String, s: String, stack: List<&2, Open>) -> St: match s: case SNil{}: SFail{} case SCon{Chr{+c}, t}: Bool.pick(St, U32.is_eq(c, 58), colon.top(k, t, stack), SFail{}) # A finished value joins its container; then "," or the closing bracket. def after.arr.c(comma: Bool, close: Bool, t: String, xs: List<&2, Val>, up: List<&2, Open>) -> St: match comma: case True{}: SValue{t, Con{OArr{xs}, up}} case False{}: match close: case True{}: SAfter{Arr{List.reverse(&2, Val, xs)}, t, up} case False{}: SFail{} def after.arr(xs: List<&2, Val>, r: String, up: List<&2, Open>) -> St: match r: case SNil{}: SFail{} case SCon{Chr{+c}, t}: after.arr.c(U32.is_eq(c, 44), U32.is_eq(c, 93), t, xs, up) def after.obj.c(comma: Bool, close: Bool, t: String, m: Map<&2, Val>, up: List<&2, Open>) -> St: match comma: case True{}: SKey{t, Con{OObj{m, ""}, up}} case False{}: match close: case True{}: SAfter{Obj{m}, t, up} case False{}: SFail{} def after.obj(m: Map<&2, Val>, r: String, up: List<&2, Open>) -> St: match r: case SNil{}: SFail{} case SCon{Chr{+c}, t}: after.obj.c(U32.is_eq(c, 44), U32.is_eq(c, 125), t, m, up) def after(v: Val, s: String, stack: List<&2, Open>) -> St: match stack: case Nil{}: SDone{v, s} case Con{OArr{xs}, up}: after.arr(Con{v, xs}, skip_ws(s), up) case Con{OObj{m, k}, up}: after.obj(Map.set(&2, Val, m, k, v), skip_ws(s), up) def step(st: St) -> St: match st: case SValue{s, stack}: value(skip_ws(s), stack) case SKey{s, stack}: key(skip_ws(s), stack) case SColon{k, s, stack}: colon(k, skip_ws(s), stack) case SStr{s, acc, k, stack}: str.step(s, acc, k, stack) case SAfter{v, s, stack}: after(v, s, stack) case SDone{v, s}: SDone{v, s} case SFail{}: SFail{} def done(v: Val, rest: String) -> Maybe<&2, Val>: match rest: case SNil{}: Some{v} case SCon{h, t}: None{} # fuel: every step but the last eats at least one char. def run(fuel: Nat, st: St) -> Maybe<&2, Val>: match fuel: case 0n: None{} case 1n+f: match st: case SDone{v, s}: done(v, skip_ws(s)) case SFail{}: None{} case SValue{s, stack}: run(f, step(SValue{s, stack})) case SKey{s, stack}: run(f, step(SKey{s, stack})) case SColon{k, s, stack}: run(f, step(SColon{k, s, stack})) case SStr{s, acc, k, stack}: run(f, step(SStr{s, acc, k, stack})) case SAfter{v, s, stack}: run(f, step(SAfter{v, s, stack})) # One JSON text: a value with optional whitespace around it. def parse(+s: String) -> Maybe<&2, Val>: run(Nat.add(String.length(s), 2n), SValue{s, Nil{}}) # Access # ------ def get.hit(r: Map<&2, Val> & Bool) -> Bool: (m, b) = r b def get.some(r: Map<&2, Val> & Val) -> Maybe<&2, Val>: (m, x) = r Some{x} def get.has(+m: Map<&2, Val>, +k: String, hit: Bool) -> Maybe<&2, Val>: match hit: case False{}: None{} case True{}: get.some(Map.get(Val, Null{}, m, k)) def get.of(+m: Map<&2, Val>, +k: String) -> Maybe<&2, Val>: get.has(m, k, get.hit(Map.has(&2, Val, m, k))) # The value under key k of an object. def get(v: Val, k: String) -> Maybe<&2, Val>: match v: case Obj{m}: get.of(m, k) case Null{}: None{} case Flag{on}: None{} case Num{n}: None{} case Str{s}: None{} case Arr{xs}: None{} # Encode # ------ def hexc(+n: U32) -> Char: Chr{Bool.pick(U32, U32.is_lt(n, 10), (48 + n : U32), (87 + n : U32))} # The escaped form of one char, reversed (it is pushed onto a reversed acc). def esc.put(+c: U32, +acc: String) -> String: Bool.pick(String, U32.is_eq(c, 34), SCon{Chr{34}, SCon{Chr{92}, acc}}, Bool.pick(String, U32.is_eq(c, 92), SCon{Chr{92}, SCon{Chr{92}, acc}}, Bool.pick(String, U32.is_eq(c, 10), SCon{Chr{110}, SCon{Chr{92}, acc}}, Bool.pick(String, U32.is_eq(c, 13), SCon{Chr{114}, SCon{Chr{92}, acc}}, Bool.pick(String, U32.is_eq(c, 9), SCon{Chr{116}, SCon{Chr{92}, acc}}, Bool.pick(String, U32.is_lt(c, 32), SCon{hexc(U32.and(c, 15)), SCon{hexc(U32.shrn(c, 4n)), SCon{Chr{48}, SCon{Chr{48}, SCon{Chr{117}, SCon{Chr{92}, acc}}}}}}, SCon{Chr{c}, acc})))))) def esc.go(s: String, acc: String) -> String: match s: case SNil{}: String.reverse(acc) case SCon{Chr{+c}, t}: esc.go(t, esc.put(c, acc)) def esc(s: String) -> String: esc.go(s, SNil{}) type Item is Data: ILit{s: String} IKey{k: String} IVal{v: Val} def items.sep(first: Bool, tail: List<&2, Item>) -> List<&2, Item>: match first: case True{}: tail case False{}: Con{ILit{","}, tail} def items.arr(xs: List<&2, Val>, first: Bool, rest: List<&2, Item>) -> List<&2, Item>: match xs: case Nil{}: Con{ILit{"]"}, rest} case Con{x, t}: items.sep(first, Con{IVal{x}, items.arr(t, False{}, rest)}) def items.obj(kvs: List<&2, Sigma<&2, &2, String, _ => Val>>, first: Bool, rest: List<&2, Item>) -> List<&2, Item>: match kvs: case Nil{}: Con{ILit{"}"}, rest} case (k, v) <> t: items.sep(first, Con{IKey{k}, Con{IVal{v}, items.obj(t, False{}, rest)}}) def put(s: String, racc: String) -> String: String.reverse(s) ++ racc # ponytail: @unsafe because it walks a work list, not the Val; each step # emits text or swaps a container for its strictly smaller parts. @unsafe def enc.go(items: List<&2, Item>, racc: String) -> String: match items: case Nil{}: String.reverse(racc) case Con{ILit{s}, t}: enc.go(t, put(s, racc)) case Con{IKey{k}, t}: enc.go(t, put("\"" ++ esc(k) ++ "\":", racc)) case Con{IVal{Null{}}, t}: enc.go(t, put("null", racc)) case Con{IVal{Flag{True{}}}, t}: enc.go(t, put("true", racc)) case Con{IVal{Flag{False{}}}, t}: enc.go(t, put("false", racc)) case Con{IVal{Num{n}}, t}: enc.go(t, put(n, racc)) case Con{IVal{Str{s}}, t}: enc.go(t, put("\"" ++ esc(s) ++ "\"", racc)) case Con{IVal{Arr{xs}}, t}: enc.go(Con{ILit{"["}, items.arr(xs, True{}, t)}, racc) case Con{IVal{Obj{m}}, t}: enc.go(Con{ILit{"{"}, items.obj(Map.to_list(&2, Val, m), True{}, t)}, racc) # Compact JSON; object keys come out sorted. def encode(v: Val) -> String: enc.go(Con{IVal{v}, Nil{}}, "") def at.go(xs: List<&2, Val>, n: Nat) -> Maybe<&2, Val>: match xs: case Nil{}: None{} case Con{v, t}: match n: case 0n: Some{v} case 1n+p: at.go(t, p) def at(v: Val, n: Nat) -> Maybe<&2, Val>: match v: case Arr{xs}: at.go(xs, n) case Null{}: None{} case Flag{on}: None{} case Num{s}: None{} case Str{s}: None{} case Obj{m}: None{} def u32.over(+acc: U32, +c: U32) -> Bool: Bool.or(U32.is_lt(429496729, acc), Bool.and(U32.is_eq(acc, 429496729), U32.is_lt(53, c))) def u32.dig(s: String, +acc: U32, bad: Bool) -> Maybe<&2, U32>: match s: case SNil{}: match bad: case True{}: None{} case False{}: Some{acc} case SCon{Chr{+c}, t}: u32.dig(t, (acc * 10 + (c - 48 : U32) : U32), Bool.or(bad, Bool.or(Bool.not(digit(c)), u32.over(acc, c)))) def u32.zero(t: String) -> Maybe<&2, U32>: match t: case SNil{}: Some{0} case SCon{d, u}: None{} def u32.lead(+c: U32, t: String, zero: Bool) -> Maybe<&2, U32>: match zero: case True{}: u32.zero(t) case False{}: u32.dig(t, (c - 48 : U32), Bool.or(Bool.not(digit(c)), u32.over(0, c))) def u32.num(s: String) -> Maybe<&2, U32>: match s: case SNil{}: None{} case SCon{Chr{+c}, t}: u32.lead(c, t, U32.is_eq(c, 48)) def u32(v: Val) -> Maybe<&2, U32>: match v: case Num{s}: u32.num(s) case Null{}: None{} case Flag{on}: None{} case Str{s}: None{} case Arr{xs}: None{} case Obj{m}: None{} # Bytes: parse # ------------ # The same machine over UTF-8 in a Bytes buffer. A state holds the offset of # its next byte; the buffer goes beside it. type Sb is Data: BValue{i: U32, stack: List<&2, Open>} BFirst{i: U32, open: Open, stack: List<&2, Open>} BKey{i: U32, stack: List<&2, Open>} BColon{key: String, i: U32, stack: List<&2, Open>} BStr{i: U32, acc: String, key: Bool, stack: List<&2, Open>} BNum{i: U32, st: U32, acc: String, stack: List<&2, Open>} BAfter{v: Val, i: U32, stack: List<&2, Open>} BSep{i: U32, open: Open, stack: List<&2, Open>} BEnd{v: Val, i: U32} BOk{v: Val} BFail{} type Cp is Data: Cp{c: U32, w: U32} def pk.if(ok: Bool, a: Array, +i: U32) -> Array & U32: match ok: case True{}: Bytes.peek(a, i) case False{}: (a, 256) # Byte i, or 256 at the end, which no rule accepts. def pk(a: Array, +len: U32, +i: U32) -> Array & U32: pk.if(U32.is_lt(i, len), a, i) def word.go(w: String, ok: Bool, r: Array & U32, +len: U32, +j: U32) -> Array & Bool: match w: case SNil{}: (a, b) = r (a, ok) case SCon{Chr{+c}, t}: (a, +b) = r word.go(t, ok && U32.is_eq(b, c), pk(a, len, (j + 1 : U32)), len, (j + 1 : U32)) # Do the bytes from j spell w? def word(a: Array, +len: U32, w: String, +j: U32) -> Array & Bool: word.go(w, True{}, pk(a, len, j), len, j) def hex4.b.go(n: Nat, r: Array & U32, +len: U32, +j: U32, +acc: U32) -> Array & U32: match n: case 0n: (a, b) = r (a, acc) case 1n+k: (a, +b) = r +h = hexd(b) hex4.b.go(k, pk(a, len, (j + 1 : U32)), len, (j + 1 : U32), Bool.pick(U32, U32.is_eq(h, 99) || U32.is_eq(acc, 65536), 65536, (acc * 16 + h : U32))) # Four hex digits from byte j, or 65536 if one is not a hex digit. def hex4.b(a: Array, +len: U32, +j: U32) -> Array & U32: hex4.b.go(4n, pk(a, len, j), len, j, 0) def cont(+b: U32) -> Bool: U32.is_eq((b .&. 192 : U32), 128) # RFC 3629 §4: overlong forms, surrogates, and code points past U+10FFFF are invalid, width 0. def utf8.n(+b0: U32, +b1: U32, +b2: U32, +b3: U32) -> Cp: +x1 = (b1 .&. 63 : U32) +x2 = (b2 .&. 63 : U32) +x3 = (b3 .&. 63 : U32) +c2 = (((b0 .&. 31) << 6n) .|. x1 : U32) +c3 = ((((b0 .&. 15) << 12n) .|. (x1 << 6n)) .|. x2 : U32) +c4 = (((((b0 .&. 7) << 18n) .|. (x1 << 12n)) .|. (x2 << 6n)) .|. x3 : U32) +ok2 = U32.is_le(192, b0) && U32.is_lt(b0, 224) && cont(b1) && U32.is_le(128, c2) +ok3 = U32.is_le(224, b0) && U32.is_lt(b0, 240) && cont(b1) && cont(b2) && U32.is_le(2048, c3) && (U32.is_lt(c3, 55296) || U32.is_lt(57343, c3)) +ok4 = U32.is_le(240, b0) && U32.is_lt(b0, 248) && cont(b1) && cont(b2) && cont(b3) && U32.is_le(65536, c4) && U32.is_le(c4, 1114111) Bool.pick(Cp, ok2, Cp{c2, 2}, Bool.pick(Cp, ok3, Cp{c3, 3}, Bool.pick(Cp, ok4, Cp{c4, 4}, Cp{0, 0}))) def utf8.b3(+b0: U32, +b1: U32, +b2: U32, r: Array & U32) -> Array & Cp: (a, +b3) = r (a, utf8.n(b0, b1, b2, b3)) def utf8.b2(+len: U32, +i: U32, +b0: U32, +b1: U32, r: Array & U32) -> Array & Cp: (a, +b2) = r utf8.b3(b0, b1, b2, pk(a, len, (i + 3 : U32))) def utf8.b1(+len: U32, +i: U32, +b0: U32, r: Array & U32) -> Array & Cp: (a, +b1) = r utf8.b2(len, i, b0, b1, pk(a, len, (i + 2 : U32))) # The code point whose lead byte b0 is at i. def utf8.at(a: Array, +len: U32, +i: U32, +b0: U32) -> Array & Cp: utf8.b1(len, i, b0, pk(a, len, (i + 1 : U32))) def str.b.cp(acc: String, key: Bool, stack: List<&2, Open>, +i: U32, r: Array & Cp) -> Array & Sb: (a, Cp{+c, +w}) = r (a, Bool.pick(Sb, U32.is_eq(w, 0), BFail{}, BStr{(i + w : U32), SCon{Chr{c}, acc}, key, stack})) def str.b.lo(+i: U32, +hi: U32, +acc: String, +key: Bool, +stack: List<&2, Open>, r: Array & U32) -> Array & Sb: (a, +lo) = r (a, Bool.pick(Sb, is_lo(lo), BStr{(i + 12 : U32), SCon{Chr{(65536 + (hi - 55296 : U32) * 1024 + (lo - 56320 : U32) : U32)}, acc}, key, stack}, BStr{(i + 6 : U32), SCon{Chr{65533}, acc}, key, stack})) # §8.2: a high surrogate pairs with a following \uDC00-\uDFFF; a lone one is U+FFFD. def str.b.pair(+len: U32, +i: U32, +hi: U32, acc: String, key: Bool, stack: List<&2, Open>, r: Array & Bool) -> Array & Sb: (a, ok) = r match ok: case True{}: str.b.lo(i, hi, acc, key, stack, hex4.b(a, len, (i + 8 : U32))) case False{}: (a, BStr{(i + 6 : U32), SCon{Chr{65533}, acc}, key, stack}) # 0: not four hex digits; 1: high surrogate; 2: low surrogate; 3: other. def u.class(+v: U32) -> U32: Bool.pick(U32, U32.is_eq(v, 65536), 0, Bool.pick(U32, is_hi(v), 1, Bool.pick(U32, is_lo(v), 2, 3))) def str.b.u.of(+k: U32, a: Array, +len: U32, +i: U32, +v: U32, acc: String, key: Bool, stack: List<&2, Open>) -> Array & Sb: match k: case 0: (a, BFail{}) case 1: str.b.pair(len, i, v, acc, key, stack, word(a, len, "\\u", (i + 6 : U32))) case 2: (a, BStr{(i + 6 : U32), SCon{Chr{65533}, acc}, key, stack}) case _: (a, BStr{(i + 6 : U32), SCon{Chr{v}, acc}, key, stack}) def str.b.u(+len: U32, +i: U32, acc: String, key: Bool, stack: List<&2, Open>, r: Array & U32) -> Array & Sb: (a, +v) = r str.b.u.of(u.class(v), a, len, i, v, acc, key, stack) # The escape whose backslash is at i. def str.b.esc(+len: U32, +i: U32, acc: String, key: Bool, stack: List<&2, Open>, r: Array & U32) -> Array & Sb: (a, +e) = r match e: case 117: str.b.u(len, i, acc, key, stack, hex4.b(a, len, (i + 2 : U32))) case _: +v = esc.simple(e) (a, Bool.pick(Sb, U32.is_eq(v, 99), BFail{}, BStr{(i + 2 : U32), SCon{Chr{v}, acc}, key, stack})) def str.b.close(+i: U32, acc: String, key: Bool, stack: List<&2, Open>) -> Sb: match key: case True{}: BColon{String.reverse(acc), (i + 1 : U32), stack} case False{}: BAfter{Str{String.reverse(acc)}, (i + 1 : U32), stack} # §7: unescaped chars below U+0020 are not allowed in a string. def str.b.other(ascii: Bool, +c: U32, a: Array, +len: U32, +i: U32, acc: String, key: Bool, stack: List<&2, Open>) -> Array & Sb: match ascii: case True{}: (a, Bool.pick(Sb, U32.is_lt(c, 32), BFail{}, BStr{(i + 1 : U32), SCon{Chr{c}, acc}, key, stack})) case False{}: str.b.cp(acc, key, stack, i, utf8.at(a, len, i, c)) def str.b(r: Array & U32, +len: U32, +i: U32, acc: String, key: Bool, stack: List<&2, Open>) -> Array & Sb: (a, +c) = r match c: case 34: (a, str.b.close(i, acc, key, stack)) case 92: str.b.esc(len, i, acc, key, stack, pk(a, len, (i + 1 : U32))) case 256: (a, BFail{}) case _: str.b.other(U32.is_lt(c, 128), c, a, len, i, acc, key, stack) def num.b.end(ok: Bool, +i: U32, acc: String, stack: List<&2, Open>) -> Sb: match ok: case True{}: BAfter{Num{String.reverse(acc)}, i, stack} case False{}: BFail{} def num.b.if(stop: Bool, +c: U32, +i: U32, +st: U32, +n: U32, acc: String, stack: List<&2, Open>) -> Sb: match stop: case True{}: num.b.end(num.accepts(st), i, acc, stack) case False{}: BNum{(i + 1 : U32), n, SCon{Chr{c}, acc}, stack} def num.b(r: Array & U32, +i: U32, +st: U32, acc: String, stack: List<&2, Open>) -> Array & Sb: (a, +c) = r +n = num.trans(st, c) (a, num.b.if(U32.is_eq(n, 99), c, i, st, n, acc, stack)) def lit.b(+i: U32, +n: U32, v: Val, stack: List<&2, Open>, r: Array & Bool) -> Array & Sb: (a, ok) = r match ok: case True{}: (a, BAfter{v, (i + n : U32), stack}) case False{}: (a, BFail{}) def value.b(r: Array & U32, +len: U32, +i: U32, stack: List<&2, Open>) -> Array & Sb: (a, +c) = r match c: case 32: (a, BValue{(i + 1 : U32), stack}) case 9: (a, BValue{(i + 1 : U32), stack}) case 10: (a, BValue{(i + 1 : U32), stack}) case 13: (a, BValue{(i + 1 : U32), stack}) case 91: (a, BFirst{(i + 1 : U32), OArr{Nil{}}, stack}) case 123: (a, BFirst{(i + 1 : U32), OObj{Map.new(&2, Val), ""}, stack}) case 34: (a, BStr{(i + 1 : U32), SNil{}, False{}, stack}) case 116: lit.b(i, 4, Flag{True{}}, stack, word(a, len, "true", i)) case 102: lit.b(i, 5, Flag{False{}}, stack, word(a, len, "false", i)) case 110: lit.b(i, 4, Null{}, stack, word(a, len, "null", i)) case _: (a, BNum{i, 0, SNil{}, stack}) # After "[" or "{": the closing bracket, or the first item. def first.b.end(open: Open, +c: U32, +i: U32, stack: List<&2, Open>) -> Sb: match open: case OArr{xs}: Bool.pick(Sb, U32.is_eq(c, 93), BAfter{Arr{Nil{}}, (i + 1 : U32), stack}, BFail{}) case OObj{m, k}: Bool.pick(Sb, U32.is_eq(c, 125), BAfter{Obj{m}, (i + 1 : U32), stack}, BFail{}) def first.b.open(open: Open, +i: U32, stack: List<&2, Open>) -> Sb: match open: case OArr{xs}: BValue{i, Con{OArr{xs}, stack}} case OObj{m, k}: BKey{i, Con{OObj{m, k}, stack}} def first.b(r: Array & U32, +i: U32, open: Open, stack: List<&2, Open>) -> Array & Sb: (a, +c) = r match c: case 32: (a, BFirst{(i + 1 : U32), open, stack}) case 9: (a, BFirst{(i + 1 : U32), open, stack}) case 10: (a, BFirst{(i + 1 : U32), open, stack}) case 13: (a, BFirst{(i + 1 : U32), open, stack}) case 93: (a, first.b.end(open, c, i, stack)) case 125: (a, first.b.end(open, c, i, stack)) case _: (a, first.b.open(open, i, stack)) def key.b(r: Array & U32, +i: U32, stack: List<&2, Open>) -> Array & Sb: (a, +c) = r match c: case 32: (a, BKey{(i + 1 : U32), stack}) case 9: (a, BKey{(i + 1 : U32), stack}) case 10: (a, BKey{(i + 1 : U32), stack}) case 13: (a, BKey{(i + 1 : U32), stack}) case 34: (a, BStr{(i + 1 : U32), SNil{}, True{}, stack}) case _: (a, BFail{}) def colon.b.top(k: String, +i: U32, stack: List<&2, Open>) -> Sb: match stack: case Nil{}: BFail{} case Con{OArr{xs}, up}: BFail{} case Con{OObj{m, old}, up}: BValue{i, Con{OObj{m, k}, up}} def colon.b(r: Array & U32, k: String, +i: U32, stack: List<&2, Open>) -> Array & Sb: (a, +c) = r match c: case 32: (a, BColon{k, (i + 1 : U32), stack}) case 9: (a, BColon{k, (i + 1 : U32), stack}) case 10: (a, BColon{k, (i + 1 : U32), stack}) case 13: (a, BColon{k, (i + 1 : U32), stack}) case 58: (a, colon.b.top(k, (i + 1 : U32), stack)) case _: (a, BFail{}) # A finished value joins its container. def after.b(v: Val, +i: U32, stack: List<&2, Open>) -> Sb: match stack: case Nil{}: BEnd{v, i} case Con{OArr{xs}, up}: BSep{i, OArr{Con{v, xs}}, up} case Con{OObj{m, k}, up}: BSep{i, OObj{Map.set(&2, Val, m, k, v), ""}, up} def sep.b.comma(open: Open, +i: U32, up: List<&2, Open>) -> Sb: match open: case OArr{xs}: BValue{i, Con{OArr{xs}, up}} case OObj{m, k}: BKey{i, Con{OObj{m, k}, up}} def sep.b.close(open: Open, +c: U32, +i: U32, up: List<&2, Open>) -> Sb: match open: case OArr{xs}: Bool.pick(Sb, U32.is_eq(c, 93), BAfter{Arr{List.reverse(&2, Val, xs)}, i, up}, BFail{}) case OObj{m, k}: Bool.pick(Sb, U32.is_eq(c, 125), BAfter{Obj{m}, i, up}, BFail{}) # After an item: "," or the closing bracket. def sep.b(r: Array & U32, +i: U32, open: Open, up: List<&2, Open>) -> Array & Sb: (a, +c) = r match c: case 32: (a, BSep{(i + 1 : U32), open, up}) case 9: (a, BSep{(i + 1 : U32), open, up}) case 10: (a, BSep{(i + 1 : U32), open, up}) case 13: (a, BSep{(i + 1 : U32), open, up}) case 44: (a, sep.b.comma(open, (i + 1 : U32), up)) case _: (a, sep.b.close(open, c, (i + 1 : U32), up)) def end.b(r: Array & U32, v: Val, +i: U32) -> Array & Sb: (a, +c) = r match c: case 32: (a, BEnd{v, (i + 1 : U32)}) case 9: (a, BEnd{v, (i + 1 : U32)}) case 10: (a, BEnd{v, (i + 1 : U32)}) case 13: (a, BEnd{v, (i + 1 : U32)}) case 256: (a, BOk{v}) case _: (a, BFail{}) # fuel: a step eats a byte, or ends a number, a value, or an opening bracket, so 4 * len + 4 steps are enough. def runb(fuel: Nat, r: Array & Sb, +len: U32) -> Maybe<&2, Val>: match fuel: case 0n: None{} case 1n+f: (a, st) = r match st: case BOk{v}: Some{v} case BFail{}: None{} case BValue{+i, stack}: runb(f, value.b(pk(a, len, i), len, i, stack), len) case BFirst{+i, open, stack}: runb(f, first.b(pk(a, len, i), i, open, stack), len) case BKey{+i, stack}: runb(f, key.b(pk(a, len, i), i, stack), len) case BColon{k, +i, stack}: runb(f, colon.b(pk(a, len, i), k, i, stack), len) case BStr{+i, acc, k, stack}: runb(f, str.b(pk(a, len, i), len, i, acc, k, stack), len) case BNum{+i, +s, acc, stack}: runb(f, num.b(pk(a, len, i), i, s, acc, stack), len) case BAfter{v, +i, stack}: runb(f, (a, after.b(v, i, stack)), len) case BSep{+i, open, stack}: runb(f, sep.b(pk(a, len, i), i, open, stack), len) case BEnd{v, +i}: runb(f, end.b(pk(a, len, i), v, i), len) # One JSON text in UTF-8. None for bad JSON or bad UTF-8 (RFC 8259 §8.1). def parse.bytes(b: Bytes.Bytes) -> Maybe<&2, Val>: Bytes.Bytes{+len, buf} = b runb(Nat.add(Nat.mul(U32.to_nat(len), 4n), 4n), (buf, BValue{0, Nil{}}), len) # Bytes: encode # ------------- # Octets being written: the buffer, its size in words, the count so far, and # the word being filled. Bytes enter at the top of w, as in Bytes.from_string. type W is Type: W{buf: Array, cap: U32, n: U32, w: U32} # The buffer doubles when word k does not fit. def w.room(grow: Bool, buf: Array, +k: U32) -> Array: match grow: case True{}: Bytes.grow((k * 4 : U32), buf, (k * 8 : U32)) case False{}: buf def w.store(full: Bool, buf: Array, +cap: U32, +n: U32, +w: U32) -> W: match full: case True{}: +k = (n >> 2n : U32) +g = U32.is_eq(k, cap) W{Array.set(U32, w.room(g, buf, k), k, w), Bool.pick(U32, g, (cap * 2 : U32), cap), (n + 1 : U32), 0} case False{}: W{buf, cap, (n + 1 : U32), w} def w.put(o: W, +b: U32) -> W: W{buf, +cap, +n, +w} = o w.store(U32.is_eq((n .&. 3 : U32), 3), buf, cap, n, ((w >> 8n) .|. (b << 24n) : U32)) def w.shrink(fit: Bool, +n: U32, buf: Array) -> Array: match fit: case True{}: buf case False{}: Bytes.dst(Bytes.copy(n, buf, Bytes.alloc(n), 0, 0)) # Stores the last partial word, then copies into the fewest words, as Bytes requires. def w.done(o: W) -> Bytes.Bytes: W{buf, +cap, +n, +w} = o +r = (n .&. 3 : U32) +k = (n >> 2n : U32) +g = U32.is_ne(r, 0) && U32.is_eq(k, cap) +cap2 = Bool.pick(U32, g, (cap * 2 : U32), cap) out = Bytes.flush(U32.is_ne(r, 0), w.room(g, buf, k), k, U32.shrn(w, U32.to_nat((((4 - r) .&. 3) * 8 : U32)))) Bytes.Bytes{n, w.shrink(U32.is_eq(U32.shln(1, Bytes.depth(Bytes.words(n))), cap2), n, out)} def w.cont(+c: U32, +k: Nat) -> U32: (128 + (U32.shrn(c, k) .&. 63) : U32) def w.cp.of(+n: U32, +c: U32, o: W) -> W: match n: case 1: w.put(o, c) case 2: w.put(w.put(o, (192 + U32.shrn(c, 6n) : U32)), w.cont(c, 0n)) case 3: w.put(w.put(w.put(o, (224 + U32.shrn(c, 12n) : U32)), w.cont(c, 6n)), w.cont(c, 0n)) case _: w.put(w.put(w.put(w.put(o, (240 + U32.shrn(c, 18n) : U32)), w.cont(c, 12n)), w.cont(c, 6n)), w.cont(c, 0n)) # One code point in UTF-8. def w.cp(+c: U32, o: W) -> W: w.cp.of(Bool.pick(U32, U32.is_lt(c, 128), 1, Bool.pick(U32, U32.is_lt(c, 2048), 2, Bool.pick(U32, U32.is_lt(c, 65536), 3, 4))), c, o) def w.hex(+n: U32) -> U32: Bool.pick(U32, U32.is_lt(n, 10), (48 + n : U32), (87 + n : U32)) def w.ctl(ctl: Bool, +c: U32, o: W) -> W: match ctl: case True{}: w.put(w.put(w.put(w.put(w.put(w.put(o, 92), 117), 48), 48), w.hex(U32.shrn(c, 4n))), w.hex((c .&. 15 : U32))) case False{}: w.cp(c, o) # §7: quote, backslash and control chars are escaped. def w.esc.c(+c: U32, o: W) -> W: match c: case 34: w.put(w.put(o, 92), 34) case 92: w.put(w.put(o, 92), 92) case 10: w.put(w.put(o, 92), 110) case 13: w.put(w.put(o, 92), 114) case 9: w.put(w.put(o, 92), 116) case _: w.ctl(U32.is_lt(c, 32), c, o) def w.text(s: String, o: W) -> W: match s: case SNil{}: o case SCon{Chr{+c}, t}: w.text(t, w.cp(c, o)) # A quoted, escaped string. def w.str.go(s: String, o: W) -> W: match s: case SNil{}: w.put(o, 34) case SCon{Chr{+c}, t}: w.str.go(t, w.esc.c(c, o)) def w.str(s: String, o: W) -> W: w.str.go(s, w.put(o, 34)) # ponytail: @unsafe for the same reason as enc.go. @unsafe def encb.go(items: List<&2, Item>, o: W) -> W: match items: case Nil{}: o case Con{ILit{s}, t}: encb.go(t, w.text(s, o)) case Con{IKey{k}, t}: encb.go(t, w.put(w.str(k, o), 58)) case Con{IVal{Null{}}, t}: encb.go(t, w.text("null", o)) case Con{IVal{Flag{True{}}}, t}: encb.go(t, w.text("true", o)) case Con{IVal{Flag{False{}}}, t}: encb.go(t, w.text("false", o)) case Con{IVal{Num{n}}, t}: encb.go(t, w.text(n, o)) case Con{IVal{Str{s}}, t}: encb.go(t, w.str(s, o)) case Con{IVal{Arr{xs}}, t}: encb.go(Con{ILit{"["}, items.arr(xs, True{}, t)}, o) case Con{IVal{Obj{m}}, t}: encb.go(Con{ILit{"{"}, items.obj(Map.to_list(&2, Val, m), True{}, t)}, o) # encode, as UTF-8 in Bytes. def encode.bytes(v: Val) -> Bytes.Bytes: w.done(encb.go(Con{IVal{v}, Nil{}}, W{Bytes.alloc(64), 16, 0, 0}))