JosephFerano/flan
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
flan/lib/parse.ml
Joseph Ferano 5a62770e52 Classes and generic functions, milestone 2's last item 
A defclass is a named dyn map with a shape tag, and a generic function
dispatches on it two ways: CLOS's, where the dispatch value is the class
of the first argument, and Clojure's, where a body computes it. They are
one mechanism and not two — a class dispatcher is (class-of arg0) as the
dispatch function, which is what lets a method written for the class
point and one written for the value :point be the same branch.

    (defclass point [x y])
    (point 3 4)                 ; the constructor, positional
    (class-of p)                ; :point, or nil for anything else
    (defgeneric area [self] dyn)
    (defmethod area point [p] (* (get p :x) (get p :y)))
    (defmulti describe [x] dyn (get x :kind))
    (defmethod describe :square [s] ...)
    (defmethod describe :else [s] ...)

A slot is a key in the instance's own map, so get, put and has-key? are
how one is read and written and no operation was added for any of it.
What the class adds is the tag, and the tag lives in the object's header
rather than in a reserved entry — the queue's note said a reserved key
and this departs from it, because a key would be counted by len, walked
by the renderer and compared by equality, so every instance would answer
a length one larger than its slot count and print a key nobody wrote. A
header field cannot be reached by get or put at all, so no user key can
collide with it. It costs nothing: the map arm of flan_obj's union grows
to the size the view arm already had, and sizeof(flan_obj) is unchanged.
It needs no tracing either — the tag is an interned keyword entry, which
is immortal and is not a collector object.

The tag shows up in exactly three places: class-of answers it, equality
compares it (two instances of one class compare by their slots; an
instance and a plain map with the same entries do not, which is
Clojure's answer for a record beside a map), and both renderers print it
— #point{ :x 1 :y 2}, Clojure's own spelling.

None of the four forms reaches the checker. lib/classes.ml turns the
whole declaration list into ordinary defns at the top of build_program,
the way Shim.expand already turns a declare-c into a declare plus a
defn: a class becomes its constructor, a generic becomes one function
whose body binds the dispatch value and compares it down a chain, and a
method becomes a branch of that chain. It is a pass and not a macro
because a macro sees one form and the generic's body is not decidable
until every method is in hand — a method may be written above its
generic, below it, or arrive at a reload an hour later.

That last case is why the method bodies are inlined rather than lifted.
A generic is exactly one top-level name, so adding a method to a running
program is the ordinary redefinition of one function, through the cell
every call site already goes through. session.ml names the generic
alongside the method's own declaration name for that reason. The cost,
recorded rather than hidden: a method is not separately callable and is
not a frame of its own.

A dispatch that finds no method signals NoMethod, a prelude struct
carrying the generic's name and the dispatch value that missed. A
condition and not a trap, because a miss is something a program can be
written to answer, and handler-case around the call is the shape. Its
value field is dyn, the first condition here with one; the per-type
descriptor an item-2 struct carries is what the collector reaches it by.
No restart is established at the miss, which is BoundsError's decision
taken for BoundsError's reason.

Both backends, identically: the two new runtime entry points are
declared in emit.ml and the x86 backend needs nothing, since a dyn call
is a dyn call there. Deferred and written down in FIX.org: inheritance,
multi-argument dispatch, :before/:after/:around, named-slot
construction, unknown-slot checking, and computed dispatch values.
2026-09-20 15:36:13 +07:00

1742 lines
86 KiB
OCaml
Raw Blame History

(** Forms → AST. Recognises special forms, desugars sugar, reports malformed
syntax with the location of the offending form.
Everything not recognised here is a call, which is how a Lisp should work:
[at], [len], [push], [abort] and the rest are ordinary functions resolved
by the checker, not syntax. *)
open Form
let fail (f : Form.t) fmt = Loc.fail f.loc fmt
let sym (f : Form.t) =
match f.v with
| Sym s -> s
| _ -> fail f "expected a name, found %s" (Form.to_string f)
(* Names for the temporaries this file mints — the value is bound once and
everything that needs it reads *that*, so a destructuring pattern over a
call calls it once and a short-circuit operand is evaluated once. [~] is a
delimiter in the reader, so no symbol anyone can write contains one: these
cannot collide with a source name and a source name cannot shadow one.
Reset per program so the names, and therefore the slot numbering
downstream, are the same every run.
The purpose is part of the name because these names are shown: the
inspector lists a frame's locals by name, and a short-circuit temp called
[destructure~3] is a plain lie about where it came from. One counter across
all purposes, so a name is still unique whatever minted it. *)
let temps = ref 0
let fresh_temp what = incr temps; Printf.sprintf "%s~%d" what !temps
(* Destructuring binds in [let] and nowhere else. Every other binding position —
a [defn] parameter, a [defstruct] field, an [fn] parameter, a [dotimes]
counter, a [match] arm's binds — takes a plain name, and a pattern written
there is refused here rather than falling out of [sym] as "expected a name".
A parameter is the one worth saying why about: it is a name/type pair, and a
pattern has no name to pair the type with, so supporting it means a pattern
inside [Ast.field] — a record [Load] and [Shim] both build and read, and
neither is this file's to change. *)
let no_pattern (f : Form.t) =
match f.v with
| Map _ | Vec _ ->
fail f
"%s is a destructuring pattern, and a pattern binds only in let — this \
position takes a plain name. Take the value under a name and \
destructure it in the body"
(Form.to_string f)
| _ -> ()
(* ── Type expressions ──────────────────────────────────────────────── *)
let rec texpr (f : Form.t) : Ast.texpr =
let mk t = { Ast.t; tloc = f.loc } in
match f.v with
(* Unit is spelled [()], ML's spelling. It is the honest name, and it cannot
collide with anything: an empty call is not a valid expression, so [()] has
no reading in value position to be confused with. Internally it stays
[Tname "Unit"] -- the resolver, the shim and the emitter all speak that
name, and diagnostics still print it. *)
| List [] -> mk (Ast.Tname "Unit")
(* One spelling. Two accepted spellings is how two spellings become
permanent, and the refusal names the new one -- the same rule the
colon-to-dot change followed. [Tname "Unit"] still exists below this
point: it is what [()] parses to, and what the resolver, the shim and the
emitter go on speaking. *)
| Sym "Unit" -> fail f "unit is written (), not Unit"
| Sym s -> mk (Ast.Tname s)
| Vec [ elem ] -> mk (Ast.Tslice (texpr elem))
| Vec [ n; elem ] -> mk (Ast.Tarray (len n, texpr elem))
| Vec _ ->
fail f "a type in brackets is [T] for a slice or [n T] for a fixed array"
(* Braces are not a type. [{K V}] used to spell [(Map K V)] and the two
resolved to the same thing; the brace spelling is withdrawn, and the
refusal names the surviving one rather than letting the form fall through
to "expected a type".
Two reasons, and the second is the one that decided it. The brace's value
meaning and its type meaning do not correspond the way the bracket's do:
[[1 2 3]] is a value whose type is [[3 i32]], but [{.x 1 .y 0}] is a
value whose type is a *name*, and a map value is built by [map-new] with
no braces anywhere. And dropping it reserves [{}] in type position for
anonymous struct types, [{.x f32 .y f32}], which is a likelier thing to
want than a second spelling of a type that already has one.
It also settles the one syntax question generics had: a defn's constraint
map, [{:where (ordered? $t)}], sits immediately after the return type,
and with braces gone from type position there is nothing for it to be
confused with. *)
| Map _ ->
fail f "a map type is written (Map K V), not in braces — braces in type \
position are not a type"
| List ({ v = Sym "Fn"; _ } :: rest) ->
(match rest with
| [ { v = Vec params; _ }; ret ] ->
mk (Ast.Tfn (List.map texpr params, texpr ret))
| _ -> fail f "a function type is (Fn [T ...] R)")
| List ({ v = Sym name; _ } :: args) when args <> [] ->
mk (Ast.Tapp (name, List.map texpr args))
| _ -> fail f "expected a type, found %s" (Form.to_string f)
and len (f : Form.t) : Ast.len =
match f.v with
| Int n -> Ast.Lint n
| Sym s -> Ast.Lname s
| _ -> fail f "an array length is an integer or a constant's name"
(* Inline name/type pairs: [x i32 y f32] — as in let, defstruct and defn. *)
let rec fields (f : Form.t) (items : Form.t list) : Ast.field list =
match items with
| [] -> []
| name :: ty :: rest ->
no_pattern name;
{ Ast.fname = sym name; fty = texpr ty; floc = name.loc } :: fields f rest
| [ odd ] ->
Loc.fail odd.loc "field %s has no type — these come in name/type pairs"
(Form.to_string odd)
(* A [defn]'s parameter vector, left undecided — the long argument is beside
the [defn] case. A bare symbol could be either half of a pair and is carried
as one; everything else is a type by its shape alone, and is resolved now so
that a malformed type is still reported at the character that is wrong.
[fields] applies [no_pattern] to the name half of each pair, and this cannot:
which half a slot is has not been decided. A map is the one shape that can be
settled here anyway — braces are not a type in any position ([texpr] refuses
them), so a map in this vector is a destructuring pattern and nothing else,
and it gets the sentence that says so rather than a complaint about map type
syntax. A bracket cannot be settled the same way, because [[a b]] is a
pattern in a name slot and a slice type in a type slot; one written in a name
slot comes back from [Check] as "a parameter's name was expected here", which
is true and is as close as this can get. *)
and pitems (items : Form.t list) : Ast.pitem list =
List.map
(fun (it : Form.t) ->
match it.v with
| Sym s -> Ast.Pname (s, it.loc)
| Map _ -> no_pattern it; assert false
| _ -> Ast.Ptype (texpr it))
items
(* A generic's or a method's parameter vector. Every slot is a bare name and
every parameter is [dyn], so the types are written out here rather than
left for [Check.pair_params] to decide: the pairing exists because a
[defn]'s vector is ambiguous until every type name is known, and this one
never is. A parameter named after a type is therefore fine here, where in a
[defn] it would be refused. *)
and dyn_params which (items : Form.t list) : Ast.field list =
List.map
(fun (it : Form.t) ->
match it.v with
| Sym s ->
{ Ast.fname = s;
fty = { Ast.t = Ast.Tname "dyn"; tloc = it.loc };
floc = it.loc }
| _ ->
fail it
"a %s's parameter is a bare name, and found %s. Every parameter of \
a generic function is dyn — there is no type to write, and a \
method that wanted one could not be reached by a dispatch that \
does not know types either"
which (Form.to_string it))
items
(* A [defmethod]'s dispatch value. Literals only: the value is compared
against what the dispatch answered at run time, and the method is declared
under a name built from it at compile time, so it has to be something both
passes can read off the source. A computed one — Clojure allows any value a
method is registered under, because registration there is a run-time call —
is not available and the message says so. *)
and dispatch (f : Form.t) : Ast.dispatch =
match f.v with
| Kw "else" -> Ast.Delse
| Sym "true" -> Ast.Dbool true
| Sym "false" -> Ast.Dbool false
| Sym s -> Ast.Dclass s
| Kw k -> Ast.Dkw k
| Str s -> Ast.Dstr s
| Int i -> Ast.Dint i
| _ ->
fail f
"a method's dispatch value is a class's name, a keyword, a string, an \
integer, true, false, or :else for the one that answers when no other \
does — and found %s. It is matched at compile time as well as at run \
time, so it is written out rather than computed"
(Form.to_string f)
(* ── The constraint map at the head of a defn body ──────────────────────
[(defn sort [s [$t]] () {:where (ordered? $t)} body ...)]. Clojure's
[{:pre [...] :post [...]}] is the precedent and the reason it is a map
rather than a bare keyword: it leaves room for further keys without new
syntax.
**The one syntax question it had, and how it stopped being one.** [{K V}]
used to be a legal *return type* spelling for [(Map K V)], which put two
braces in a row meaning different things — [(defn f [xs [$t]] {string i32}
{:where ...} body)]. The brace spelling has since been withdrawn from type
position entirely ([texpr] above), so the slot after the return type can be
nothing but this.
The leading keyword is checked, but is no longer what tells a constraint
map from anything else braces can mean in the position a body starts in —
dyn maps changed what else is possible there. [{.x 1}] is still read as a
struct literal's field list and never as a constraint map, but [expr]
below says so, with its own message; this function no longer sees that
shape at all. What DOES reach here besides [:where]: a keyword-keyed map
with a different key, an empty map, or a map keyed on something that is
neither a keyword nor a [.field] symbol, and only when the body has more
after it — with nothing after, that map is the whole single-form body, a
real dyn value like [(defn f [] dyn {:a 1})] or [(defn f [] dyn {})], and
this function leaves it alone. *)
let constraints (body : Form.t list) : Ast.pred list * Form.t list =
match body with
(* A map literal opening on [:where] is always a constraint map, even with
nothing after it: a where-clause with no body past it is what a moved
closing paren produces, and the whole point of naming the key here is
to catch that as a constraint-map error ("a where predicate is ...", or
whatever [keys] finds wrong with it) rather than let a stray [(ordered?
$t)] surface later as "unknown function ordered?" from inside what was
meant as a predicate. Any OTHER map literal — keyed on a different
keyword, keyed on something that is not a keyword at all, or holding no
keys at all — is read as a constraint map only when something follows
it in the body: unlike [:where], nothing about the map alone says it is
a mistake and not an ordinary map literal until [rest] says whether it
is one body form among several (discarded, so worth flagging as the
typo it almost always is) or the function's entire single-form body
([(defn f [] dyn {:a 1})], where the map is not discarded, it IS the
answer, evaluated for both its side effects and its value like any
other body form). Note that "discarded" here is about the map's
*result*, not about whether evaluating it can do anything: {:a (println
"hi")} still prints, same as any expression statement whose value
nothing uses — this arm's business is a value going unused, not
silence. An empty map, [{}], has no key to check and gets its own
message rather than running [keys] on nothing and saying nothing.
One shape is excluded on purpose: a map opening on a [.field] symbol,
[{.x 1}], is a struct field list with no struct name in front of it, and
[expr] below already gives that its own message, "a bare map is not an
expression; write (Type {.field v})" — the one this file had before any
of the above existed. Catching it here first would bury that dedicated
diagnostic under "a constraint map is keyword/value pairs", which is
true but not what is wrong with it. *)
| ({ Form.v = Form.Map kvs; loc } as m) :: rest
when (match kvs with
| { Form.v = Form.Kw "where"; _ } :: _ -> true
| { Form.v = Form.Sym s; _ } :: _
when String.length s > 1 && s.[0] = '.' -> false
| _ -> rest <> []) ->
if kvs = [] then
Loc.fail loc
"an empty map literal here is discarded — the body has more after \
it, and its value going unused is almost always a typo for \
{:where ...}; write (do {} ...) if the empty map is deliberate"
else
let pred (p : Form.t) =
match p.Form.v with
(* [$t] at a predicate, not bare [t]: the clause talks about the
variable the signature *bound*, and writing it the way the signature
wrote it is the one spelling that cannot be read as a concrete type
that happens to share the name. *)
| Form.List [ { Form.v = Form.Sym name; _ };
{ Form.v = Form.Sym v; loc = vloc } ]
when String.length v > 1 && v.[0] = '$' ->
ignore vloc;
{ Ast.pname = name; pvar = String.sub v 1 (String.length v - 1);
ploc = p.Form.loc }
| _ ->
Loc.fail p.Form.loc
"a where predicate is (name? $t), one predicate about one type \
variable — found %s" (Form.to_string p)
in
let rec keys = function
| [] -> []
| { Form.v = Form.Kw "where"; _ } :: v :: rest ->
(match v.Form.v with
(* A vector, because two predicates on one variable is the ordinary
case — [{:where [(ordered? $t) (copyable? $t)]}] is what a
comparing generic that also reads its parameter twice needs. One
predicate on its own is accepted unwrapped, which is the same
sugar [:pre] does not have and is worth the line it costs. *)
| Form.Vec ps -> List.map pred ps
| _ -> [ pred v ])
@ keys rest
| { Form.v = Form.Kw k; loc } :: _ :: rest ->
Loc.fail loc
"%s is not a key a defn's constraint map takes; :where is the only \
one" (":" ^ k)
|> fun () -> keys rest
| odd :: _ ->
Loc.fail odd.Form.loc
"a constraint map is keyword/value pairs — found %s"
(Form.to_string odd)
in
if List.length kvs mod 2 <> 0 then
Loc.fail m.Form.loc "a constraint map is keyword/value pairs, and this \
one has an odd number of forms";
(keys kvs, rest)
| _ -> ([], body)
(* ── Expressions ───────────────────────────────────────────────────── *)
let rec expr (f : Form.t) : Ast.expr =
let mk e = { Ast.e; loc = f.loc } in
match f.v with
| Int i -> mk (Ast.Int i)
| Float x -> mk (Ast.Float x)
| Byte b -> mk (Ast.Byte b)
| Str s -> mk (Ast.Str s)
| Kw k -> mk (Ast.Kw k)
| Sym s -> mk (Ast.Var s)
(* In value position brackets are a fixed-array literal; in type position
they are a slice or array type. Position disambiguates, as with {}. *)
| Vec items -> mk (Ast.Arr (List.map expr items))
(* Braces in value position are two literals told apart by their first form.
A [.field] symbol says struct, and a bare struct field list still needs
its type written — (Type {.field v}) — because the fields alone do not
name one. Anything else, the empty braces included, is a dyn map literal:
{:a 1 :b s}, keys and values alternating, each an ordinary expression. *)
| Map ({ v = Sym s; _ } :: _)
when String.length s > 1 && s.[0] = '.' ->
fail f "a bare map is not an expression; write (Type {.field v})"
| Map items -> mk (Ast.MapLit (None, map_pairs f items))
| List [] -> fail f "() is not an expression"
| List (head :: args) -> form f mk head args
and form f mk (head : Form.t) (args : Form.t list) : Ast.expr =
match head.v with
(* ── quote ─────────────────────────────────────────────────────── *)
| Sym "quote" ->
(match args with
| [ { v = Sym s; _ } ] -> mk (Ast.Quote s)
| _ -> fail f "quote takes one symbol")
(* ── sequencing and binding ────────────────────────────────────── *)
| Sym "do" -> mk (Ast.Do (List.map expr args))
| Sym "let" ->
(match args with
| { v = Vec bs; _ } :: body -> mk (Ast.Let (bindings f bs, body_of body))
| _ -> fail f "let is (let [name value ...] body ...)")
(* ── conditionals ──────────────────────────────────────────────── *)
| Sym "if" ->
(match args with
| [ c; t ] -> mk (Ast.If (expr c, expr t, None))
| [ c; t; e ] -> mk (Ast.If (expr c, expr t, Some (expr e)))
| _ -> fail f "if is (if test then) or (if test then else)")
(* Sugar, desugared here: special forms until macros land at milestone 5. *)
| Sym "when" ->
(match args with
| c :: body when body <> [] ->
mk (Ast.If (expr c, { Ast.e = Ast.Do (body_of body); loc = f.loc }, None))
| _ -> fail f "when is (when test body ...)")
| Sym "cond" -> cond f args
(* Short-circuiting, so they cannot be ordinary calls. *)
| Sym "and" -> shortcircuit f args ~is_and:true
| Sym "or" -> shortcircuit f args ~is_and:false
(* ── loops ─────────────────────────────────────────────────────── *)
(* An optional label comes first: [(while :outer (< i n) ...)]. A keyword in
the head position is unambiguous because a loop condition is never one and
a [dotimes] binding vector is never one either, so [label] peels it off
whatever follows the form's name. *)
| Sym "while" ->
(match label args with
| lbl, c :: body -> mk (Ast.While (lbl, expr c, body_of body))
| _, [] ->
fail f "while is (while test body ...), or (while :label test body ...)")
| Sym "until" ->
(match label args with
| lbl, c :: body ->
let neg = { Ast.e = Ast.Call ({ Ast.e = Ast.Var "not"; loc = head.loc },
[ expr c ]); loc = f.loc } in
mk (Ast.While (lbl, neg, body_of body))
| _, [] ->
fail f "until is (until test body ...), or (until :label test body ...)")
(* Break and continue. Not a goto: the label names one of the loops this form
is lexically inside, and the checker resolves it against exactly those, so
control can only leave a loop it is already in — the same restriction
Odin's labelled break has. Bare, each means the innermost loop. *)
| Sym "break" ->
(match label args with
| lbl, [] -> mk (Ast.Break lbl)
| _ -> fail f "break is (break) or (break :label)")
| Sym "continue" ->
(match label args with
| lbl, [] -> mk (Ast.Continue lbl)
| _ -> fail f "continue is (continue) or (continue :label)")
(* ── control ───────────────────────────────────────────────────── *)
| Sym "return" ->
(match args with
| [] -> mk (Ast.Return None)
| [ v ] -> mk (Ast.Return (Some (expr v)))
| _ -> fail f "return takes at most one value")
| Sym "set" ->
(match args with
| [ target; value ] -> mk (Ast.Set (place target, expr value))
| _ -> fail f "set is (set place value)")
(* ── (array 4 rl/Vector2) ───────────────────────────────────────────
A zeroed fixed array, told its count and its element type. The type
spelling [4 rl/Vector2] is unchanged and still works everywhere a type is
expected; what it cannot do is appear in a [let] binding, which has no
type slot, because there the brackets are an array *literal* of two
elements and the second of them is a name nothing declares. So the count
and the type arrive as plain arguments and Parse assembles the type
itself. [(zeroed)] keeps its own job — the empty value of whatever the
destination wants — and this is the one that is told. *)
| Sym "array" ->
(match args with
| [ n; t ] ->
mk (Ast.ArrayOf { Ast.t = Ast.Tarray (len n, texpr t); tloc = f.loc })
| _ ->
fail f
"array is (array COUNT TYPE), as in (array 4 rl/Vector2) — a zeroed \
fixed array of COUNT of them")
| Sym "match" ->
(match args with
| scrutinee :: rest -> mk (Ast.Match (expr scrutinee, arms f rest))
| [] -> fail f "match is (match value pattern body ...)")
(* ── binding and control: never a call ─────────────────────────── *)
(* A form that binds a name or alters control flow cannot fall through to
Call — it would parse cleanly and mean the wrong thing, silently. *)
| Sym "fn" ->
(match args with
| { v = Vec ps; _ } :: body when body <> [] ->
List.iter no_pattern ps;
mk (Ast.Fn (List.map sym ps, body_of body))
| _ -> fail f "fn is (fn [param ...] body ...)")
| Sym "dotimes" ->
(match label args with
| lbl, ({ v = Vec [ n; count ]; _ } :: body) ->
no_pattern n;
mk (Ast.Dotimes (lbl, sym n, expr count, body_of body))
| _ -> fail f "dotimes is (dotimes [name count] body ...)")
(* [(loop [x 0 acc 1] body ...)]. No label: [break] and [continue] may not
leave a loop — a loop answers with the value of its body, and a jump out
of one has no value to give — so there is nothing here for a label to
name. A leading keyword is caught here rather than left to [bindings],
which would complain that [:outer] has no value. *)
| Sym "loop" ->
(match args with
| { v = Kw k; _ } :: _ ->
fail f
":%s — loop takes no label. break and continue may not leave a loop, \
because a loop answers with the value of its body; there is nothing \
for a label to name" k
| { v = Vec bs; _ } :: body -> mk (Ast.Loop (loop_bindings f bs, body_of body))
| _ -> fail f "loop is (loop [name value ...] body ...)")
(* Rebind and jump to the top. Its arguments are checked against the loop's
names in order, so the count is the binding vector's count. *)
| Sym "recur" -> mk (Ast.Recur (List.map expr args))
| Sym "defer" ->
(match args with
| [] -> fail f "defer is (defer body ...)"
| body -> mk (Ast.Defer (body_of body)))
| Sym "some" ->
(match args with
| [ v ] -> mk (Ast.Unwrap (Ast.Usome, expr v))
| _ -> fail f "some is (some option-value)")
| Sym "try" ->
(match args with
| [ v ] -> mk (Ast.Unwrap (Ast.Utry, expr v))
| _ -> fail f "try is (try result-value)")
(* (signal c) : Unit, always. When every applicable handler returns normally
the signalling function simply carries on, and with no handler at all it is
a no-op — spec-conditions.md §1 and §2. *)
(* And (error c) : Never — §2's diverging variant. Same lookup, but a
handler that returns normally does not answer it: with nothing
transferring the program stops. *)
| Sym (("signal" | "error") as how) ->
let kind = if how = "signal" then Ast.Ssignal else Ast.Serror in
(match args with
| [ c ] -> mk (Ast.Signal (kind, expr c))
| _ -> fail f "%s is (%s condition)" how how)
(* (handler-bind [(Type [c] body ...) ...] body ...)
A clause names a condition type, binds the condition, and runs for effect;
matching is by type, since there is no condition hierarchy. *)
| Sym "handler-bind" ->
let clauses, body =
match args with
| { v = Vec clauses; _ } :: body when body <> [] -> (clauses, body)
| _ ->
fail f "handler-bind is (handler-bind [(Type [name] body ...) ...] body ...)"
in
let clause (c : Form.t) =
match c.Form.v with
| Form.List (ty :: { v = Form.Vec [ { v = Form.Sym n; _ } ]; _ } :: cbody)
when cbody <> [] ->
{ Ast.hty = texpr ty; hname = n; hbody = List.map expr cbody;
hloc = c.Form.loc }
| _ ->
fail c "a handler-bind clause is (Type [name] body ...)"
in
mk (Ast.HandlerBind (List.map clause clauses, body_of body))
(* (handler-case BODY [(Type [c] body ...) ...]) — spec-conditions.md, the
unwinding half of the pair.
The body comes first and the clauses after it, which is the opposite of
handler-bind's order and is deliberate: a handler-bind is read as
something established *around* a body, and a handler-case is read as a
body with answers hung off the end of it. A clause is spelled exactly as
handler-bind spells one, because it names the same thing — a condition
type and a name to bind it to. At least one clause, since a handler-case
with none would be its body and nothing else. *)
| Sym "handler-case" ->
let body, clauses =
match args with
| [ body; { v = Vec clauses; _ } ] when clauses <> [] -> (body, clauses)
| _ ->
fail f
"handler-case is (handler-case body [(Type [name] body ...) ...]) \
with at least one clause"
in
let clause (c : Form.t) =
match c.Form.v with
| Form.List (ty :: { v = Form.Vec [ { v = Form.Sym n; _ } ]; _ } :: cbody)
when cbody <> [] ->
{ Ast.hty = texpr ty; hname = n; hbody = List.map expr cbody;
hloc = c.Form.loc }
| _ -> fail c "a handler-case clause is (Type [name] body ...)"
in
mk (Ast.HandlerCase (expr body, List.map clause clauses))
(* (restart-case BODY (name [p T ...] BODY-1) ...) — spec-conditions.md §3.
The body and every clause have the same type, which is the form's. A
clause's parameters are inline name/type pairs, like any other binding
form; what fills them in is the [invoke-restart] that chose the clause. *)
| Sym "restart-case" ->
let body, clauses =
match args with
| body :: clauses when clauses <> [] -> (body, clauses)
| _ ->
fail f "restart-case is (restart-case body (name [p T] body ...) ...)"
in
let clause (c : Form.t) =
match c.Form.v with
| Form.List ({ v = Form.Sym n; _ } :: { v = Form.Vec ps; _ } :: cbody)
when cbody <> [] ->
{ Ast.rname = n; rparams = fields c ps;
rbody = List.map expr cbody; rloc = c.Form.loc }
| _ -> fail c "a restart-case clause is (name [p T] body ...)"
in
mk (Ast.RestartCase (expr body, List.map clause clauses))
(* (invoke-restart 'name arg ...) : Never. The name is a quoted symbol — that
is what the reader's quote is for — and it is resolved on the restart
stack at run time, since restarts are dynamically scoped. The arguments
fill in the clause's parameters, and how many there are and what they are
is settled at run time too, against the frame the name found (§3). *)
| Sym "invoke-restart" ->
(match args with
| { v = Form.List [ { v = Form.Sym "quote"; _ }; { v = Form.Sym n; _ } ]; _ }
:: rest ->
mk (Ast.InvokeRestart (n, List.map expr rest))
| _ ->
fail f
"invoke-restart takes a quoted restart name and then its arguments, \
as in (invoke-restart 'use-value 42)")
(* ── macros ────────────────────────────────────────────────────── *)
(* The reader now produces these three, so they arrive here as ordinary heads
and would fall through to Call — coming back from the checker as "unknown
name quasiquote", which says nothing about what is actually missing. *)
(* [Expand.quasiquote] runs over every form on the way into [program] and
[decl], so a quasiquote is gone before this file looks at it and this arm
cannot be reached by anything that came through either. It is kept as the
backstop for the path that did not: a form built by hand and handed
straight to [expr]. *)
| Sym "quasiquote" ->
fail f "a quasiquote reached the parser undesugared, which means this form \
did not come through Parse.program or Parse.decl"
(* Not a milestone, a mistake: these two mean nothing anywhere else, and the
reader cannot tell, because it does not track where it is. *)
| Sym "unquote" ->
fail f "~x means nothing outside a quasiquote"
| Sym "unquote-splicing" ->
fail f "~@x means nothing outside a quasiquote, and splices only into a \
list or a vector"
(* A declaration is not an expression, and this arm says so for all of them
rather than for [defmacro] alone. It used to be that one, because it was
the only head anyone typed by mistake; now that [Parse.expr] expands, a
macro can *produce* one, and the head this dispatches on is the only place
that sees it — the walk recurses, so a [defn] nested inside what a macro
answered is caught with the same message as one typed at the top.
A quasiquoted declaration is deliberately not caught: after desugaring,
the name in ``(defn ...)`` is a string inside a [Form.Sym] argument and not
a head, which is the same property that makes a quasiquoted macro call
output rather than a dependency. Building a declaration as a value is what
a macro is for. *)
| Sym ("defmacro" | "defn" | "defvar" | "defconst" | "defstruct" | "defdata"
| "defunion" | "defclass" | "defgeneric" | "defmulti" | "defmethod"
| "defenum" | "defalias" | "import" as name) ->
fail f
"%s is a top-level declaration, not an expression. A quasiquoted one is \
a value and a macro may answer with it; an evaluated one is not a thing \
anything can do" name
(* Recognised, deliberately unimplemented. Rejected rather than left to fall
through to Call, where they would parse and mean nothing. *)
(* [find-restart] and [compute-restarts] are §4's two ways to look at the
restart stack without committing to one. *)
| Sym ("find-restart" | "compute-restarts"
| "errdefer"
| "await" as name) ->
fail f "%s is not implemented yet (see the build sequence in plan.org)" name
(* ── field access: (.pos c) ────────────────────────────────────── *)
| Sym s when String.length s > 1 && s.[0] = '.' ->
let field = String.sub s 1 (String.length s - 1) in
(match args with
| [ target ] -> mk (Ast.Field (expr target, field))
| _ -> fail f "field access is (.%s value)" field)
(* ── struct literal: (Cursor {.src s .pos 0}) ───────────────────── *)
(* Only braces whose first form is a [.field] symbol — or the empty braces,
which have always meant the zero-initialised struct here. A map literal
as an argument, (f {:a 1}), keeps its head as an ordinary call. *)
| Sym name when args <> [] && is_struct_map (List.hd args) ->
(match args with
| [ { v = Map kvs; _ } ] -> mk (Ast.Struct (name, struct_fields f kvs))
| _ -> fail f "a struct literal is (%s {.field value ...})" name)
(* ── anything else is a call ────────────────────────────────────── *)
| _ -> mk (Ast.Call (expr head, List.map expr args))
and is_struct_map (f : Form.t) =
match f.v with
| Map [] -> true
| Map ({ v = Sym s; _ } :: _) -> String.length s > 1 && s.[0] = '.'
| _ -> false
(* A map literal's braces hold key/value pairs, each an expression. *)
and map_pairs f (items : Form.t list) : (Ast.expr * Ast.expr) list =
let rec go = function
| [] -> []
| k :: v :: rest -> (expr k, expr v) :: go rest
| [ odd ] ->
Loc.fail odd.Form.loc
"a map literal is key/value pairs, and this one has an odd number of \
forms — found %s with no value" (Form.to_string odd)
in
ignore f;
go items
and body_of (items : Form.t list) : Ast.expr list = List.map expr items
(* A loop label, or a [break]'s target: a leading keyword, peeled off. Nothing
else in any of these positions is a keyword — a loop condition is not, a
[dotimes] binding vector is not, and [break] takes nothing else at all — so
one function serves all four forms and no form has to say which arguments it
has counted. *)
and label (items : Form.t list) : string option * Form.t list =
match items with
| { v = Kw k; _ } :: rest -> (Some k, rest)
| _ -> (None, items)
(* A loop's binding vector. Pairs like [let]'s, but plain names only: a
destructuring pattern expands to several bindings from one form, and then
[recur]'s argument count would no longer match what is written here. *)
and loop_bindings f (items : Form.t list) : (string * Ast.expr) list =
let rec go = function
| [] -> []
| name :: value :: rest ->
no_pattern name;
(sym name, expr value) :: go rest
| [ odd ] ->
Loc.fail odd.loc
"binding %s has no value — loop takes name/value pairs"
(Form.to_string odd)
in
let bs = go items in
List.iter
(fun (n, _) ->
if List.length (List.filter (fun (m, _) -> m = n) bs) > 1 then
fail f "%s is bound twice in this loop" n)
bs;
bs
and bindings f (items : Form.t list) : Ast.binding list =
(* [name value ...] and [name Type value ...] both read; a type is a form
that is not a value position — disambiguated by pair vs triple is
ambiguous, so let requires (let [name value]) and types are inferred.
Annotated locals are not needed by any acceptance program. *)
let rec go = function
| [] -> []
| pat :: value :: rest ->
let bs = destructure pat (expr value) in
no_duplicates pat bs;
bs @ go rest
| [ odd ] ->
Loc.fail odd.loc "binding %s has no value — let takes name/value pairs"
(Form.to_string odd)
in
if items = [] then Loc.fail f.loc "let needs at least one binding" else go items
(* ── Destructuring ─────────────────────────────────────────────────── *)
(* The temporary every pattern binds its value to before anything reads it, so
that the value is evaluated once however many names come out of it. Returning
the reference as well as the binding is what makes the two impossible to
separate by accident. *)
and temp (p : Form.t) (v : Ast.expr) : Ast.expr * Ast.binding =
let t = fresh_temp "destructure" in
({ Ast.e = Ast.Var t; loc = p.loc },
{ Ast.bname = t; bty = None; bval = v; bloc = p.loc })
(* Clojure's destructuring, desugared here into the bindings and field accesses
the language already has. [Ast.binding] carries a name and nothing else, and
deliberately so: nothing downstream — not [Load]'s renaming, not [Check], not
any backend — learns that a pattern exists. The same reason [dotimes] is a
[Let] plus a [While].
The one thing this cannot decide is whether an array pattern's arity matches
the value's, because that is a type and there are none here. [destructure~nth]
carries the question to [Check], which answers it and emits an ordinary [at].
A binding is a pattern only when it is written in brackets or braces; a bare
name is what it always was. *)
and destructure (p : Form.t) (v : Ast.expr) : Ast.binding list =
match p.v with
| Sym name -> [ { Ast.bname = name; bty = None; bval = v; bloc = p.loc } ]
(* The value goes into a temporary first, so it is evaluated once however
many names the pattern binds, and so that [(let [{:keys [p]} p] ...)]
reads the old [p] rather than the one it is in the middle of rebinding. *)
| Map items -> let t, bind = temp p v in bind :: dmap p t items
| Vec items -> let t, bind = temp p v in bind :: dvec p t items
| _ ->
fail p
"expected a name or a destructuring pattern, found %s — a pattern is \
{:keys [x y]} over a struct or [a b] over a fixed array"
(Form.to_string p)
(* {:keys [x y]} and {inner .field}, over a struct. Clojure's map destructuring
with Flan's structs standing in for its maps: [:keys] is the common case and
the pair form is what nests, since a [:keys] entry is a name and never a
pattern. Everything else Clojure puts in this position — [:as], [:or],
[:strs], [:syms] — is refused by name where it is written.
[:keys] keeps its colon while [.field] takes the dot, and the split is the
point rather than an inconsistency: [.field] names a field of the struct,
[:keys] names no field at all — it is an instruction to the compiler that
happens to sit in the same brace. Keeping them apart leaves the dot meaning
exactly one thing, "this names a field", which is the whole reason the
colon was given up here. *)
and dmap (p : Form.t) (t : Ast.expr) (items : Form.t list) : Ast.binding list =
let ex loc e : Ast.expr = { Ast.e; loc } in
let field loc name = ex loc (Ast.Field (t, name)) in
let rec go = function
| [] -> []
| { v = Kw "keys"; _ } :: names :: rest ->
let ns =
match names.v with
| Vec ns -> ns
| _ ->
Loc.fail names.loc
":keys takes a bracketed list of field names, found %s"
(Form.to_string names)
in
let rec each = function
| [] -> []
| (n : Form.t) :: more ->
let name =
match n.v with
| Sym s -> s
| _ ->
Loc.fail n.loc
":keys binds field names, and %s is not one — a nested pattern \
is written {%s .field}"
(Form.to_string n) (Form.to_string n)
in
{ Ast.bname = name; bty = None; bval = field n.loc name; bloc = n.loc }
:: each more
in
each ns @ go rest
| ({ v = Kw k; _ } as bad) :: _ :: rest ->
ignore rest;
Loc.fail bad.loc
":%s is not implemented in a destructuring pattern — a struct pattern \
is {:keys [x y]} or {name .field}, and nothing else" k
| pat :: ({ v = Sym s; _ } as fform) :: rest
when String.length s > 1 && s.[0] = '.' ->
destructure pat (field fform.loc (String.sub s 1 (String.length s - 1)))
@ go rest
| pat :: ({ v = Kw fld; _ } as bad) :: _ ->
ignore pat;
Loc.fail bad.loc
"a field label is written .%s, not :%s — the colon is for keys, and a \
struct pattern binds {name .%s}" fld fld fld
| pat :: other :: _ ->
Loc.fail other.loc
"expected .field after %s, found %s — a struct pattern binds \
{name .field}" (Form.to_string pat) (Form.to_string other)
| [ odd ] ->
Loc.fail odd.loc "%s has no .field — a struct pattern comes in pairs"
(Form.to_string odd)
in
if items = [] then
fail p "an empty struct pattern {} binds nothing — write the names it should bind"
else go items
(* [a b] and [a b & rest], over a fixed array. Not over a slice: see [Check]. *)
and dvec (p : Form.t) (t : Ast.expr) (items : Form.t list) : Ast.binding list =
let ex loc e : Ast.expr = { Ast.e; loc } in
let var loc n = ex loc (Ast.Var n) in
let rec split acc = function
| [] -> (List.rev acc, None)
| ({ v = Sym "&"; _ } as amp) :: rest ->
(match rest with
| [ r ] -> (List.rev acc, Some r)
| [] -> Loc.fail amp.loc "& needs a name after it, as in [a b & rest]"
| _ :: extra :: _ ->
Loc.fail extra.loc
"& takes one name and it is the last thing in the pattern")
| x :: rest -> split (x :: acc) rest
in
let elems, rest = split [] items in
let n = List.length elems in
(match elems, rest with
| [], None ->
fail p "an empty array pattern [] binds nothing — write the names it should bind"
| [], Some r ->
Loc.fail r.loc
"[& %s] binds the whole value — write %s on its own instead of a pattern"
(Form.to_string r) (Form.to_string r)
| _ -> ());
(* With a [& rest] the pattern says "at least this many"; without one it says
"exactly this many". [Check] is where the array's length is known, so the
count and which of the two it means travel there as arguments. *)
let exact = if rest = None then 1L else 0L in
let nth i =
ex p.loc
(Ast.Call (var p.loc "destructure~nth",
[ t;
ex p.loc (Ast.Int (Int64.of_int i));
ex p.loc (Ast.Int (Int64.of_int n));
ex p.loc (Ast.Int exact) ]))
in
let rec each i = function
| [] -> []
| e :: more -> destructure e (nth i) @ each (i + 1) more
in
let rest_binding =
match rest with
| None -> []
| Some r ->
(* An ordinary (slice t n (len t)): the tail of the temporary, which is a
local and outlives the body that reads it. Nothing new. *)
let name =
match r.v with
| Sym s -> s
| _ ->
Loc.fail r.loc
"& binds one name for the tail, and %s is not one — the tail is a \
slice, so it cannot be destructured further" (Form.to_string r)
in
[ { Ast.bname = name; bty = None; bloc = r.loc;
bval =
ex r.loc
(Ast.Call (var r.loc "slice",
[ t;
ex r.loc (Ast.Int (Int64.of_int n));
ex r.loc (Ast.Call (var r.loc "len", [ t ])) ])) } ]
in
each 0 elems @ rest_binding
(* One pattern binding the same name twice is a mistake, not a shadowing: the
second would win and the first would bind nothing. Across a let's bindings it
*is* shadowing and stays legal, so this looks at one pattern at a time. *)
and no_duplicates (p : Form.t) (bs : Ast.binding list) =
let rec go seen = function
| [] -> ()
| (b : Ast.binding) :: rest ->
if String.contains b.Ast.bname '~' then go seen rest
else if List.mem b.Ast.bname seen then
Loc.fail b.Ast.bloc "this pattern binds %s twice" b.Ast.bname
else go (b.Ast.bname :: seen) rest
in
ignore p; go [] bs
(* A field label is a dot, never a colon. The delimiter is what disambiguates:
[(.x v)] is a call and therefore an access, [{.x 1.0}] is a brace form and
therefore a construction. The colon is left for keys — map keys and enum
members — so the two never share a spelling. *)
and struct_fields f (items : Form.t list) : (string * Ast.expr) list =
let rec go = function
| [] -> []
| { v = Sym s; _ } :: value :: rest
when String.length s > 1 && s.[0] = '.' ->
(String.sub s 1 (String.length s - 1), expr value) :: go rest
| ({ v = Kw k; _ } as bad) :: _ :: _ ->
Loc.fail bad.loc
"a field label is written .%s, not :%s — the colon is for keys, and a \
struct value is (Type {.%s value ...})" k k k
| other :: _ :: _ ->
Loc.fail other.loc "expected .field, found %s" (Form.to_string other)
| [ odd ] -> Loc.fail odd.loc "field %s has no value" (Form.to_string odd)
in
ignore f; go items
and cond f (args : Form.t list) : Ast.expr =
let rec go = function
| [] -> { Ast.e = Ast.Do []; loc = f.loc } (* no clause matched: Unit *)
| { v = Kw "else"; _ } :: body :: _ -> expr body
| test :: body :: rest ->
{ Ast.e = Ast.If (expr test, expr body, Some (go rest)); loc = f.loc }
| [ odd ] -> Loc.fail odd.loc "cond clause %s has no body"
(Form.to_string odd)
in
if args = [] then Loc.fail f.loc "cond needs at least one clause" else go args
(* Every test here is an [if]'s condition, so a dyn operand is truthy-tested
(check.ml's check_truthy) exactly the way a bare [if]'s is, for both
[and] and [or]. The *answer* is the operand that decided the form, which
is Clojure's rule and needs the operand a second time: [(or a b)] is
[(let [t a] (if t t b))] and [(and a b)] is [(let [t a] (if t b t))].
The temp is what makes that a single evaluation — writing the operand
itself into the arm, as [(if a a b)] would, evaluates it twice.
Both used to answer a bare bool sentinel on the deciding path instead.
[and]'s "false" sentinel sat in the else arm, so check_if typed the real
branch first and boxed the sentinel to match: an all-truthy [and] did
hand back its last operand, but a falsey one answered [false] where
Clojure answers the falsey operand itself — (and 1 nil) said false, not
nil. [or]'s "true" sentinel sat in the then arm, the one check_if types
first, so the sentinel decided the whole expression's type and a later
non-bool dyn answer hit the strict bool boundary instead of surviving as
itself: (or nil "x") trapped rather than answering "x", exactly the
canonical (or x default) idiom Clojure is reached for.
The locs are the operand's own, not the whole form's, because the temp's
[Var] node is what lands in the [if] condition and check_truthy reports
the condition's loc when a typed operand is not a bool. Pointing that at
[f.loc] would blame the enclosing (and ...) for whichever operand is
actually wrong.
What answering the operand costs, for both forms alike: the two arms are
now both real values, so mixing a dyn operand with a typed bool one makes
check_if unify them, and the then arm decides. A non-bool dyn value on
the losing side then meets the strict bool boundary at run time —
(or false (box "s")) and (and (box nil) some-bool) both trap, verified on
this tree. Each form used to be safe in exactly one of those directions,
because the sentinel it answered was a bool literal that boxed to fit
whatever the real branch was; neither is now, and they are at least
symmetric about it. Making bool and dyn arms join as dyn is a check_if
question, noted in FIX.org under item 7 and not decided here.
One known wart, measured rather than guessed, and left alone deliberately.
In a want-free position — [(println (and true true (vec-new i32)))] — the
caret lands on the *second* [true] and not on the vec: the last operand is
the then arm, check_if types the then arm first, and the mismatch is
therefore reported against the else arm, which is the previous operand's
temp. An operand anywhere but last is a condition instead, so check_truthy
blames it at its own loc and the caret is right; [or] is right everywhere,
because there the chain and not the sentinel sits in the else arm. Giving
the else arm's [Var] node the *last* operand's loc moves the caret onto the
vec and makes the sentence read backwards — "expected (Vec i32), found
bool" under a caret on the thing that is the (Vec i32) — so it is not an
improvement; answering a bool literal again would revert the paragraph
above; and inverting the condition to move the last operand into the else
arm buys a [not] per operand and worse locs than it fixes. What would
actually fix it is check_if preferring the arm that is not a compiler temp
when it reports, which is check.ml's call. Written up in FIX.org. *)
and shortcircuit f (args : Form.t list) ~is_and : Ast.expr =
let mk e = { Ast.e; loc = f.loc } in
let rec go = function
| [] -> mk (Ast.Var (if is_and then "true" else "false"))
| [ last ] -> expr last
| x :: rest ->
let ex = expr x in
let t = fresh_temp (if is_and then "and" else "or") in
let tvar = { Ast.e = Ast.Var t; loc = ex.Ast.loc } in
let bind = { Ast.bname = t; bty = None; bval = ex; bloc = ex.Ast.loc } in
let rest = go rest in
let body =
if is_and then mk (Ast.If (tvar, rest, Some tvar))
else mk (Ast.If (tvar, tvar, Some rest))
in
mk (Ast.Let ([ bind ], [ body ]))
in
go args
and place (f : Form.t) : Ast.place =
match f.v with
| Sym s -> Ast.Pvar s
| List ({ v = Sym s; _ } :: args)
when String.length s > 1 && s.[0] = '.' ->
let field = String.sub s 1 (String.length s - 1) in
(match args with
| [ target ] -> Ast.Pfield (expr target, field)
| _ -> fail f "field place is (.%s value)" field)
| List ({ v = Sym "at"; _ } :: target :: idx) when idx <> [] ->
Ast.Pindex (expr target, List.map expr idx)
(* Not a place. spec-memory.md gives a map an upsert of its own — [put]
either inserts or replaces — so there is no store into a lookup, and an
entry that is absent has no location to store into. Refused here rather
than parsed into a place form the language does not have. *)
| List ({ v = Sym "get"; _ } :: _) ->
fail f "(get m k) is not a place — a map is written with (put m k v)"
| List [ { v = Sym "deref"; _ }; p ] -> Ast.Pderef (expr p)
| _ ->
fail f
"%s is not assignable. set takes a name, (.field x), (at a i ...), \
or (deref p)"
(Form.to_string f)
and arms f (items : Form.t list) : Ast.arm list =
let rec go = function
| [] -> []
| p :: body :: rest ->
{ Ast.pat = pattern p; body = [ expr body ]; aloc = p.loc } :: go rest
| [ odd ] ->
Loc.fail odd.loc "match arm %s has no body" (Form.to_string odd)
in
if items = [] then Loc.fail f.loc "match needs at least one arm" else go items
and pattern (f : Form.t) : Ast.pattern =
match f.v with
| Sym "_" -> Ast.Pwild
| Kw "else" -> Ast.Pwild
| Sym ctor -> Ast.Pctor (ctor, [])
(* An enum member, which is the one other thing [match] could plausibly be
over: an enum is an i32 at run time, so the arms would be a chain of [=]
and the members are all known, which is exhaustiveness [cond] cannot give.
What stops it is not the lowering, it is that a keyword pattern needs a
case in [Ast.pattern] — and [lib/load.ml] matches that type exhaustively,
so the variant cannot be added from here. Refused by name rather than
spelled as a constructor it is not. *)
| Kw member ->
fail f
":%s is not implemented as a pattern — match is over an Option here, \
and an enum member cannot be one until Ast.pattern can hold a keyword. \
Use cond with (= k :%s)" member member
| List ({ v = Sym ctor; _ } :: binds) ->
List.iter no_pattern binds;
Ast.Pctor (ctor, List.map sym binds)
| _ -> fail f "expected a pattern, found %s" (Form.to_string f)
(* ── The third element of a defvar ─────────────────────────────────────
[(defvar x i32)] declares a zeroed static and [(defvar score 0)] declares a
dyn global holding 0, and which one a form is is decided by whether the
third element is a type. The author's rule, 2026-09-20: "if it's 3 atoms
then it's dyn", and "dispatch the if it's a type do the right thing".
Most forms are settled by their shape alone and are settled here: [0], a
string, a map, [[1 2 3]] and [(f "x")] are not types by any reading, so the
global is dyn and its initialiser is the expression; [[4 u32]], [()] and
[(Fn [i32] i32)] are types by any reading and keep exactly the meaning they
have today. Note which side the bracket falls on: [[n T]] stays a fixed
array, so [(defvar rows [4 u32])] is the zeroed grid it always was, and a
*vector literal* of two names is not reachable in this position.
Two shapes are left over, and they are the ones a name decides rather than
a shape: a bare symbol, which is a type name or a value's name, and
[(head arg ...)] with every argument type-shaped, which is [(Vec i32)] or a
call. Both readings are built and carried — the [texpr] in the [Defvar] and
the [Ast.Ambiguous] expression beside it — and [Check.collect] picks the
type reading whenever the form is a type. Nothing here resolves a name,
because at parse time there are none.
[texpr] is called under a handler on purpose: "does this parse as a type"
is the question, and its refusals are how it answers no. It builds an AST
and touches nothing else, so there is nothing to undo when it raises. *)
let defvar3 (f : Form.t) : Ast.texpr * Ast.init =
let as_type () = match texpr f with t -> Some t | exception Loc.Error _ -> None in
let dyn = { Ast.t = Ast.Tname "dyn"; tloc = f.loc } in
match f.v with
| Sym _ ->
(match as_type () with
(* [Sym "Unit"] is the one symbol [texpr] refuses outright — unit is
spelled [()] — and the refusal is about the spelling of a type, so it
stays the error it is rather than becoming a read of a variable
nobody can have declared. *)
| None -> (texpr f, Ast.Zeroed)
| Some t -> (t, Ast.Ambiguous (expr f)))
| List ({ v = Sym _; _ } :: _ :: _) ->
(match as_type () with
| Some ({ Ast.t = Ast.Tapp _; _ } as t) -> (t, Ast.Ambiguous (expr f))
(* [(Fn [i32] i32)] and anything else [texpr] reads as a type without
going through [Tapp] has no call reading to be confused with. *)
| Some t -> (t, Ast.Zeroed)
| None -> (dyn, Ast.Init (expr f)))
| _ ->
(match as_type () with
| Some t -> (t, Ast.Zeroed)
| None -> (dyn, Ast.Init (expr f)))
(* ── Declarations ──────────────────────────────────────────────────── *)
let rec decl (f : Form.t) : Ast.decl =
let mk d = { Ast.d; dloc = f.loc } in
match f.v with
| List ({ v = Sym "package"; _ } :: args) ->
(match args with
| [ n ] -> mk (Ast.Package (sym n))
| _ -> fail f "package is (package name)")
| List ({ v = Sym "import"; _ } :: args) ->
(match args with
| [ alias; { v = Str path; _ } ] -> mk (Ast.Import (sym alias, path))
| _ -> fail f "import is (import alias \"collection:path\")")
| List ({ v = Sym "defalias"; _ } :: args) ->
(match args with
| [ n; t ] -> mk (Ast.Defalias (sym n, texpr t))
| _ -> fail f "defalias is (defalias Name Type)")
| List ({ v = Sym "defstruct"; _ } :: args) ->
(match args with
| [ n; { v = Vec fs; _ } ] -> mk (Ast.Defstruct (sym n, fields f fs))
| _ -> fail f "defstruct is (defstruct Name [field Type ...])")
| List ({ v = Sym "defdata"; _ } :: args) ->
(match args with
| [ n; { v = Vec vs; _ } ] -> mk (Ast.Defdata (sym n, List.map variant vs))
| _ -> fail f "defdata is (defdata Name [(Case [field Type ...]) ...])")
(* C's union: one storage, as many ways of reading it as there are members.
It carries a field list and not a case list, which is the whole surface
difference from [defdata] — there is no tag, so there is nothing to name
a case with.
The tagged sum was spelled [defunion] until this form wanted the name, and
a file written before the rename is the hazard this arm exists for. It is
not an alias and it is not a near-miss: the old text would *parse* under
the new meaning. [(defunion U [A B])] is two bare symbols, which is
exactly the shape of one member [A] of type [B], and it would have gone on
compiling as an untagged union of one member — the silent misparse the
[defn] case above was rewritten to make impossible, with no diagnostic
anywhere and nothing in the source that looks wrong.
So the name slots are read before anything is built. A member name is
lowercase and a case name is capitalised, and a case *with* fields is a
list where a member name would be; either one means the text in hand is a
tagged sum wearing the old spelling, and it is refused by name. A file
that really did mean an untagged union whose first member is capitalised
is refused too, and it is the right trade: that is not a thing anyone has
written, and being told to rename a member is nothing beside being given
the wrong type in silence. *)
| List ({ v = Sym "defunion"; _ } :: args) ->
(match args with
| [ n; { v = Vec ms; _ } ] ->
List.iteri
(fun i (m : Form.t) ->
let looks_tagged =
i mod 2 = 0
&& (match m.v with
| List _ -> true
| Sym s -> s <> "" && s.[0] = Char.uppercase_ascii s.[0]
| _ -> false)
in
if looks_tagged then
Loc.failk "parse/defunion-renamed" f.loc
"the tagged sum is defdata now — (defdata Name [(Case [field \
Type ...]) ...]) — and defunion is C's untagged union, whose \
members overlay one storage: (defunion Name [member Type \
...]). This reads as the tagged one, so it is refused rather \
than quietly given the other meaning")
ms;
mk (Ast.Defunion (sym n, fields f ms))
| _ -> fail f "defunion is (defunion Name [member Type ...])")
(* The slot after the parameters is unconditionally the return type. It used
to be optional, and the parser decided return-type-versus-body by looking
the symbol up in a set of the file's type names -- sound only because one
top-level namespace means a name cannot be both a type and a value, and
brittle because the set had to be complete. It was wrong twice in one day,
the second time parsing [(defn f [] (Rune {.code 65}) (bar))] as a
function *returning* a Rune with a one-form body, silently, in every file
in the language. A silent misparse is the worst failure class available,
and macros now generate definitions, which widens it.
Mandatory removes the guess: nothing is consulted, [()] is what a function
that returns nothing writes, and a mistyped type is a mistyped type --
[(defn f [] f65 0.0)] reaches the resolver's near-miss check and comes back
as *did you mean f64*, where it used to come back as an unknown name.
The return slot stays mandatory now that parameters may be left
unannotated, and it is worth saying why the two do not move together.
Dynamic-by-default means a *parameter* with no type is [dyn]; the return
type could have been given the same rule, and was not, because the
ambiguity there has no syntactic resolution at all. [(defn f [] (Rune
{.code 65}) (bar))] is the case above: a capitalised head in a list is a
type application and also a struct literal -- see [Struct] in [expr] --
and no rule separates them, so an optional return slot is a coin toss
between a type and the first form of a body. A parameter vector has no
such case: every slot in it is a name or a type and never an expression.
So [dyn] is written out in the return position, which costs one token and
keeps a decision this file paid for twice in one day.
── The parameter vector ──────────────────────────────────────────────
[(defn f [x y])] is one parameter [x] of type [y], or two parameters [x]
and [y] of type [dyn], and which one it is depends on whether [y] names a
type. That is the lookup this comment's first half says was removed for
being brittle, and it is being asked for again -- so it is not done here.
The vector is carried undecided, as [Ast.pitem]s, and paired in [Check],
where the set of type names is complete.
The move is not cosmetic. What the old rule got wrong was consulting a set
that was not finished being built: it ran per-file, at parse time, before
macros had generated their definitions, and macros generating definitions
is exactly what widened the failure. By the time [Check] pairs the vector,
every file is loaded, every macro has expanded and every C header has been
imported, so the set is not a guess about what might be a type -- it is
the types. That is strictly more than the parser could ever know, and it
is the whole of the argument for the placement.
What deferring does not buy is immunity. The set is complete at a point in
time and not across time: [(defn f [x y] ...)] is two dyn parameters until
somebody writes [(defstruct y ...)] or imports a header that declares one,
and then it is one parameter of type [y], with no edit to [f]. The
signature changes under it. That residual is real, it is the dictated
rule's and not this file's, and [Session.compatible] is where it is felt --
a redefinition that changes a signature is refused there, and this is a
way for a signature to change with nothing redefined. *)
| List ({ v = Sym "defn"; _ } :: args) ->
(match args with
| n :: { v = Vec ps; _ } :: ret :: body ->
(* The slot's own failure, because the thing found there is almost
always the old spelling: a body whose first form was a call, written
when the slot could be left out. [texpr]'s "expected a type" alone
would be true and unhelpful. *)
let rty =
try texpr ret with
| Loc.Error { Loc.dloc = loc; dmsg = msg; _ } ->
Loc.fail loc
"%s. This is the return type, which every defn states -- a \
function that returns nothing writes ()" msg
in
let fwhere, body = constraints body in
mk (Ast.Defn { Ast.name = sym n; params = []; praw = Some (pitems ps);
ret = Some rty; fwhere; fbody = body_of body;
nloc = n.loc })
| _ ->
fail f
"defn is (defn name [param Type ...] ReturnType body ...). The return \
type is not optional; a function that returns nothing writes ()")
(* ── The dyn side's classes and generic functions ──────────────────
Four forms, all of them shorthand: nothing below [Classes.expand] knows
they exist, and what it writes in their place is ordinary [defn]s. The
parsing here is only the shape check — which slots are present, and what
kind of thing is in each — because everything that needs the other
declarations to answer (is that a class? is there a generic by that
name? has this dispatch value a method already?) is the expansion's.
Every parameter of a generic and of a method is [dyn], written or not,
so a parameter vector here takes bare names and nothing else. That is
what keeps these off [defn]'s undecided-pairing path: a [defn]'s vector
cannot be read until every type name is known, and one that may hold
only names can be read here. *)
| List ({ v = Sym "defclass"; _ } :: args) ->
(match args with
| [ n; { v = Vec slots; _ } ] ->
mk (Ast.Defclass
(sym n,
List.map
(fun (s : Form.t) ->
match s.v with
| Sym name -> (name, s.loc)
| _ ->
fail s
"a class slot is a name. Its value is dyn and there is \
no type to write: an instance is a dyn map with a \
shape tag on it, and (get p :%s) is how a slot is read"
(Form.to_string s))
slots))
| _ -> fail f "defclass is (defclass Name [slot ...])")
| List ({ v = Sym ("defgeneric" | "defmulti" as which); _ } :: args) ->
let generic = String.equal which "defgeneric" in
let usage =
if generic then
"defgeneric is (defgeneric name [param ...] ReturnType). It has no \
body: its dispatch value is the class of its first argument, which \
is what makes it the class-dispatching half of the pair. Write \
defmulti for a dispatch value of your own"
else
"defmulti is (defmulti name [param ...] ReturnType body ...), and the \
body is the dispatch: it answers the value the methods are keyed by"
in
(match args with
| n :: { v = Vec ps; _ } :: ret :: body
when if generic then body = [] else body <> [] ->
mk ((if generic then (fun fn -> Ast.Defgeneric fn)
else fun fn -> Ast.Defmulti fn)
{ Ast.name = sym n; params = dyn_params which ps; praw = None;
ret = Some (texpr ret); fwhere = []; fbody = body_of body;
nloc = n.loc })
| _ -> fail f "%s" usage)
| List ({ v = Sym "defmethod"; _ } :: args) ->
(match args with
| n :: key :: { v = Vec ps; _ } :: body when body <> [] ->
let gen = sym n and k = dispatch key in
mk (Ast.Defmethod
{ Ast.mgen = gen; mkey = k; mkloc = key.loc;
(* The name is the declaration's, not a symbol anything emits:
no function is ever written under it. *)
mfn = { Ast.name = gen ^ "@" ^ Ast.dispatch_text k;
params = dyn_params "defmethod" ps; praw = None;
ret = None; fwhere = []; fbody = body_of body;
nloc = n.loc } })
| _ ->
fail f
"defmethod is (defmethod generic dispatch [param ...] body ...). \
There is no return type: the generic states it once, for every \
method written for it")
| List ({ v = Sym ("declare" | "declare-c" as which); _ } :: args) ->
(* (declare name [param Type ...] ReturnType? "c_symbol"). The C symbol is
last and is always written: a foreign name is not derivable from a Flan
one, and guessing it would fail at link time rather than here.
[declare-c] is the same shape and a different claim about the symbol.
[declare]'s signature IS the C signature, already flattened by whoever
wrote the C; [declare-c]'s is the *library's* — structs by value — and
[Shim] generates the flattening. The two cannot be one form, because
(declare f [p string] ...) already means the symbol takes ptr+len and
(declare-c f [p string] ...) means it takes a NUL-terminated char *. *)
let mkd fn csym =
if String.equal which "declare-c" then Ast.DeclareC (fn, csym)
else Ast.Declare (fn, csym)
in
let usage =
Printf.sprintf
"%s is (%s name [param Type ...] ReturnType? \"c_symbol\")" which which
in
(match List.rev args with
| { v = Str csym; _ } :: rest ->
(match List.rev rest with
| [ n; { v = Form.Vec ps; _ } ] ->
mk (mkd { Ast.name = sym n; params = fields f ps; praw = None;
ret = None; fwhere = []; fbody = []; nloc = n.loc } csym)
| [ n; { v = Form.Vec ps; _ }; r ] ->
mk (mkd { Ast.name = sym n; params = fields f ps; praw = None;
ret = Some (texpr r); fwhere = []; fbody = [];
nloc = n.loc } csym)
| _ -> fail f "%s" usage)
| _ -> fail f "%s" usage)
(* A member's value is optional and autoincrements, which is C's rule and is
here for C's reason: the enums this language writes are as often a
transcription of a header as they are original, and every value written by
hand is a value that can be wrong. [0 1 2 3 4] typed out is fine until a
member is inserted in the middle, and then the renumbering is a manual
edit of every line below it.
Which values were *written* does not survive this form. [Ast.Defenum]
holds resolved numbers, so by the time [Check] sees an enum the difference
between a number someone chose and one autoincrement produced is gone --
and there are no per-member locations in the AST to point at either. That
is why the collision rule below is enforced here and not in the checker
beside the duplicate-*name* rule: this one is a question about the source
text, and the parser is the last pass that can still answer it. *)
| List ({ v = Sym "defenum"; _ } :: args) ->
(match args with
| [ n; { v = Form.Vec ms; _ } ] ->
let ename = sym n in
(* An enum member is an [i32] at run time. [Shim] lowers the type to
int32_t for C's benefit and [Check] builds every member as a
[Tast.Int (v, I32)] -- but the reader hands this pass an [int64], so
a value the type cannot hold arrives here looking perfectly ordinary.
Left alone it is truncated by the x86 backend and malformed in the
LLVM IR, and worse than either, it defeats the duplicate-value rule
below: that rule compares [int64]s, so in
[(defenum E [A 0 B 4294967296])] the two values differ and the scan
passes, while at run time both members are 0 and a [match] on one is
unreachable through the other. The one rule written to catch two
names for one number waves through the case it exists for.
So every value is checked against its run-time type here -- refused
if it does not fit, never truncated to fit -- the moment it is
resolved and before any of that reasoning runs, and the scan below
therefore compares the numbers the program will actually have.
The bounds are the signed 32-bit ones and they are spelled out rather
than borrowed. [Check.in_range] is the function that does exactly
this for every ordinary literal, and it is the right one -- but
[Check] is built above this module and reading its output, so there
is no call this file could make. If one of the two changes, change
the other. *)
let fits m loc ~explicit (v : int64) =
if Int64.compare v (-2147483648L) >= 0
&& Int64.compare v 2147483647L <= 0
then v
else if explicit then
Loc.failk "parse/enum-value-out-of-range" loc
"the member %s of %s is %Ld, which does not fit i32 — an enum's \
discriminant is an i32, so its members run from -2147483648 to \
2147483647. Give %s a value in that range, or a defconst of a \
wider type if the number itself is what matters"
m ename v m
else
Loc.failk "parse/enum-value-out-of-range" loc
"the member %s of %s has no value of its own, so it \
autoincrements to %Ld, which does not fit i32 — an enum's \
discriminant is an i32, so its members run from -2147483648 to \
2147483647. Write %s's value out, or lower the member above it"
m ename v m
in
(* Each member becomes its name, its value, whether that value was
written, and where the name is. The last two exist only so the
refusals here and below can be made; neither reaches the AST. *)
let rec members next = function
| [] -> []
| { v = Form.Sym m; loc } :: { v = Form.Int k; _ } :: rest ->
(* The [let] is load-bearing rather than tidiness. OCaml leaves the
evaluation order of [::]'s two operands unspecified and in
practice takes the tail first, so an inlined [fits ... k] would
run *after* the recursive call -- and for
[(defenum E [A 9223372036854775807 B])] that recursive call is
[Int64.add max_int 1L], which wraps quietly to min_int and would
have the refusal name B and a number written nowhere in the
source. Binding first refuses A, whose value is the one actually
wrong, and in doing so makes the wrap unreachable: [k] is inside
i32 by the time it is incremented, so the sum cannot overflow. *)
let k = fits m loc ~explicit:true k in
(m, k, true, loc) :: members (Int64.add k 1L) rest
| { v = Form.Sym m; loc } :: rest ->
let next = fits m loc ~explicit:false next in
(m, next, false, loc) :: members (Int64.add next 1L) rest
| bad :: _ ->
fail bad
"an enum member is a name, optionally followed by an integer, \
found %s" (Form.to_string bad)
in
let ms = members 0L ms in
(* A duplicate value that was written is an alias and is meant: a [Count]
or a [Last] pointing at a value another member already holds is how C
spells the end of a range, and refusing it would refuse a real idiom.
A duplicate that autoincrement walked into is nobody's decision. It
happens when a member above is renumbered or one is inserted, and the
result is two names for one number with nothing in the source saying
so -- silently, and the program still compiles, and one of the two is
now unreachable through a [match] on the other. That silence is the
same failure class as a silent misparse, so the implicit member is
refused; writing its value out is both the fix and the way to say the
alias was intended.
Every member is resolved before any of this runs, because the member
an autoincrement collides with is as often below it as above: in
[(defenum E [A B 0])] it is [A], the implicit one, that has to be
refused, and a left-to-right check would never see [B] coming. *)
let indexed =
List.mapi (fun i (m, v, explicit, loc) -> (i, m, v, explicit, loc)) ms
in
List.iter
(fun (i, m, v, explicit, loc) ->
if not explicit then
match
List.find_opt
(fun (j, _, ov, _, _) -> j <> i && Int64.equal ov v)
indexed
with
| None -> ()
| Some (_, other, _, _, oloc) ->
Loc.failk "parse/enum-autoincrement-collision" loc
~notes:
[ Loc.note oloc
(Printf.sprintf "%s has the value %Ld" other v) ]
"%s has no value of its own, so it autoincrements to %Ld, \
which is the value %s already has. Give %s its value \
explicitly if the two are meant to be one number under two \
names, or a value no other member holds"
m v other m)
indexed;
mk (Ast.Defenum (ename, List.map (fun (m, v, _, _) -> (m, v)) ms))
| _ ->
fail f
"defenum is (defenum Name [member value? ...]). A member with no \
value takes the previous member's plus one, and the first takes 0")
| List ({ v = Sym "defvar"; _ } :: args) ->
(match args with
| [ n; t ] ->
let ty, init = defvar3 t in
mk (Ast.Defvar (sym n, Some ty, init))
| [ n; t; { v = Sym "uninit"; _ } ] ->
mk (Ast.Defvar (sym n, Some (texpr t), Ast.Uninit))
| [ n; t; v ] -> mk (Ast.Defvar (sym n, Some (texpr t), Ast.Init (expr v)))
| _ ->
fail f
"defvar is (defvar name Type value?) or (defvar name value) — a \
third element that is not a type is the value of a dyn global")
| List ({ v = Sym "defconst"; _ } :: args) ->
(match args with
| [ n; v ] -> mk (Ast.Defconst (sym n, None, expr v))
| [ n; t; v ] -> mk (Ast.Defconst (sym n, Some (texpr t), expr v))
| _ -> fail f "defconst is (defconst name Type? value)")
(* A macro is an ordinary function, and this is where it becomes one:
[(defmacro m [args] body)] is [(defn m [args [Form]] Form body)]. There is
no [Ast.Defmacro] and there is not going to be one -- a macro has the type
[[Form] -> Form], it is compiled by the same backend as everything else,
and the only thing that makes it a macro is that [Expand] calls it at
compile time instead of the program calling it at run time.
One parameter, the slice of the argument forms, rather than one declared
parameter per argument. It needs no reader or parser change and it gives
variadics for free, which is what [unless] and [when] need in a language
with no &rest.
The shape rules stay exactly as they were, because they were enforced
before the feature existed on purpose: getting the shape wrong and getting
the whole feature are different mistakes. *)
| List ({ v = Sym "defmacro"; _ } :: args) ->
(match args with
| n :: { v = Form.Vec [ p ]; _ } :: body when body <> [] ->
let form_t = { Ast.t = Ast.Tname "Form"; tloc = f.loc } in
mk (Ast.Defn
{ Ast.name = sym n;
params = [ { Ast.fname = sym p;
fty = { Ast.t = Ast.Tslice form_t; tloc = p.loc };
floc = p.loc } ];
(* Written out, not deferred: a macro takes [[Form]] and
returns a [Form], and neither half of that is the user's to
leave off. *)
praw = None;
ret = Some form_t; fwhere = []; fbody = body_of body;
nloc = n.loc })
| _ :: { v = Form.Vec ps; _ } :: body when body <> [] ->
List.iter (fun (p : Form.t) -> ignore (sym p)) ps;
fail f
"a macro takes one parameter, the forms at its call site, and this \
one names %d. There is no &rest and no arity: (defmacro m [args] \
...) and (len args) is how many were written"
(List.length ps)
| _ ->
fail f "defmacro is (defmacro name [param ...] body ...)")
(* Only reachable from the single-declaration entry point below: a file's
forms go through [splice] first, and a [do] there is its items. Said by
name because the two paths differ and the difference is not the author's
fault to guess at. *)
| List ({ v = Sym "do"; _ } :: _) ->
fail f
"a top-level (do ...) is several declarations spliced in place, and this \
is a position that takes exactly one — a macro answering several is a \
file's form, not an expression's"
| List ({ v = Sym s; _ } :: _) -> fail f "unknown top-level form (%s ...)" s
| _ -> fail f "expected a top-level declaration, found %s" (Form.to_string f)
and variant (f : Form.t) : Ast.variant =
match f.v with
| Sym n -> { Ast.vname = n; vfields = []; vloc = f.loc }
| List [ { v = Sym n; _ }; { v = Vec fs; _ } ] ->
{ Ast.vname = n; vfields = fields f fs; vloc = f.loc }
| List [ { v = Sym n; _ } ] -> { Ast.vname = n; vfields = []; vloc = f.loc }
| _ -> fail f "a data type case is Name or (Name [field Type ...])"
(* Macro expansion, which runs over [Form] and therefore before anything in
this file. It cannot be called directly: expanding a macro means compiling
it and dlopening it, so the expander sits above [Check] and [Build] and this
module sits below them. [Macro] fills this in, and lib/dune passes -linkall
so that it always has -- an executable that links the library gets the
installation whether or not it names the module.
The default is the identity because [Macro] is what knows which names are
macros; with nothing installed, a call to one arrives at the checker as an
unknown name, which is wrong but not silent. *)
let expander : (Form.t list -> Form.t list) ref = ref (fun fs -> fs)
(* The defmacros an import brought in: already qualified under the alias the
package was imported as, and already quasiquote-desugared, so they are in
exactly the shape [Macro] hands its own round-0 set.
A ref, for the same reason [expander] is one — [Load] sits below [Macro] and
above this file, so there is no call it could make instead — and set rather
than passed because the parse that needs them is not always the parse that
resolved them: [Session.eval] parses one [defn] for C-c C-c, long after the
import that supplied the macro it calls. [Load.program] sets it around the
parse it drives and restores it afterwards; the session sets it around the
whole of an evaluation. Empty is the ordinary case and costs nothing. *)
let imported_macros : Form.t list ref = ref []
(* What those macros are allowed to *call*, and it is the same list the
importing program gets: the package's declarations, qualified under the
alias, as [Load] already built them.
A macro module is compiled from the prelude plus the [defmacro]s, so until
now the header's rule held without anything enforcing it — a macro body
could call prelude functions and other macros and nothing else. A package
macro that called one of its own package's functions was renamed to
[alias/fn] by [Load.rename_form], reached the checker with nothing of that
name declared, and was refused as a call into an imported package.
That refusal was the machinery missing a piece rather than a rule. The
rename says the intent plainly: what a package's macro answers with, and
what its body calls, is spelled the way the importer spells it. So the
declarations travel beside the macros and go into the module with them.
[Macro.compile] prunes them to what the macros actually reach, so a package
whose macros are pure quasiquote — raylib's five [with-*] — pays nothing and
links nothing new.
An [Ast.decl list] and not forms, because [Load] has already done the
qualifying over the Ast and a second renamer over [Form] would be that work
written twice, in the file where the two copies could disagree silently. *)
let imported_decls : Ast.decl list ref = ref []
let with_imported ?(decls = []) (ms : Form.t list) (f : unit -> 'a) : 'a =
let saved = !imported_macros in
let saved_decls = !imported_decls in
imported_macros := ms;
imported_decls := decls;
Fun.protect
~finally:(fun () ->
imported_macros := saved;
imported_decls := saved_decls)
f
(* Two entry points and not one function with a flag, and the reason is the
daemon. [Loc.Errors] is a second exception, and the handlers in the session
and in the daemon name only [Loc.Error] — so a list reaching them would be
an unhandled exception and a dead session, which is the one thing the whole
dev loop exists to prevent. A flag on the function the session already calls
would put that one label away from happening. A separate name cannot: the
session's call site has to be edited by someone for its behaviour to change.
[keep_going] asks for every bad declaration in the file rather than the
first. The resync point is a top-level form, and it is the only honest one
here: the reader already found where each declaration ends, so skipping a
bad one costs nothing and cannot lose its place. Inside a declaration there
is no such landmark, so one bad [defn] is one error. *)
(* ── One call, several declarations ────────────────────────────────
Expansion is form-for-form: [Macro.expand_form] answers one [Form.t] per
input and the loop below turns each into one [Ast.decl]. Every macro written
until now expands to an *expression* — [unless], [into], raylib's [with-*] —
so one-for-one was the whole of what was needed.
A type provider is the first thing that is not. [(defedn Tileset "t.edn")]
has to produce the struct *and* the reader over it, and a nested map in the
data means a struct per nesting level: three declarations and more from one
form. There is no arrangement of one-for-one that reaches that.
So a [do] at the top level is its items, in place. It is the sequencing
spelling the language already has, it is Clojure's answer to exactly this,
and it is only ever reachable by a macro: nobody writes [(do (defn ...))] in
a file, and the message below still says so for anyone who tries and wrote
it wrong. Recursive, because a macro that splices what another macro
answered has a [do] inside a [do] and the nesting is not the author's to
flatten by hand.
It is spliced *after* expansion and before the declaration walk, so what is
spliced is already fully expanded — a [do] holding a call to another macro
settled before it got here. *)
let rec splice (f : Form.t) : Form.t list =
match f.Form.v with
| Form.List ({ Form.v = Form.Sym "do"; _ } :: items) ->
List.concat_map splice items
| _ -> [ f ]
let parse_forms ~keep_going (forms : Form.t list) : Ast.decl list =
(* Quasiquote first and always, because it is pure and needs nothing loaded:
it is what turns a macro body into ordinary code, and the prelude's own
macros have to parse in a process that has not built a macro module yet.
Then expansion, which may need one. *)
let forms = !expander (List.map Expand.quasiquote forms) in
let forms = List.concat_map splice forms in
temps := 0;
let s = Loc.sink ~on:keep_going in
let decls = List.filter_map (fun f -> Loc.caught s (fun () -> decl f)) forms in
Loc.finish s;
decls
(** One file, stopping at the first declaration it cannot parse. Raises
[Loc.Error], never [Loc.Errors]. *)
let program (forms : Form.t list) : Ast.decl list =
parse_forms ~keep_going:false forms
(** One file, reporting every declaration it cannot parse. Raises [Loc.Errors]
when there was more than nothing wrong, so only a caller prepared for a
list should be calling it. *)
let program_all (forms : Form.t list) : Ast.decl list =
parse_forms ~keep_going:true forms
(* Single-declaration entry point, for tests and the REPL. *)
let decl (f : Form.t) : Ast.decl =
temps := 0;
match !expander [ Expand.quasiquote f ] with
| [ f ] -> decl f
| fs ->
(* One declaration in, one out. A macro at the top level would break that,
and there is no top-level macro call: [decl] dispatches on the head and
a macro name is not one of the heads it knows. *)
Loc.fail f.loc "expanding this declaration produced %d of them"
(List.length fs)
(* Single-expression entry point: C-x C-e, and the tests that parse one
expression. It expands, which [Parse.expr] above does not and never did —
so a bare [(unless c a b)] typed at the REPL was an unknown name, the
prelude's macros included. That is the whole of the change here: the wrap is
the one [decl] has, applied to the other entry point.
[Expand.quasiquote] first for [parse_forms]'s reason — it is pure, it needs
nothing loaded, and the arm in [form] that refuses an undesugared quasiquote
is the backstop for the path that skips this, not for this one.
What an expression that expands to a declaration does is decided in [form]:
[defn] and its siblings are refused by name, wherever in the expansion they
appear. Nothing here has to look for them.
[temps] is deliberately not reset. [decl] resets it because a declaration is
a fresh top level; an expression is evaluated into a session that has been
handing out temporaries all along, and restarting the counter would hand out
a name the frame beside it is already using. *)
let expr (f : Form.t) : Ast.expr =
match !expander [ Expand.quasiquote f ] with
| [ f ] -> expr f
| fs ->
(* One expression in, one out. [Macro.program] is a [List.map], so it
cannot answer with anything else — this is here because the invariant is
worth stating where it is relied on, not because it has been seen. *)
Loc.fail f.loc "expanding this expression produced %d forms, and an \
expression is one" (List.length fs)
Reference in New Issue View Git Blame Copy Permalink