Merge pull request #88 from jolt-lang/error-round-3

errors: unresolved symbols error with Clojure's message (round 3)

jolt-core/clojure/core/10-seq.clj

@@ -8,6 +8,14 @@
 ;; self-hosted compiler (jolt-core/jolt/*.clj). Compiler-facing structural fns go
 ;; in the kernel tier (00-kernel) instead — see its header.
 
+;; Volatiles (moved up from 20-coll: this tier's transducers use them, and the
+;; analyzer now ERRORS on unresolved forward references — jolt-2o7.3). The
+;; constructor (volatile!) stays native; these are pure over ref-put!/get.
+(defn vreset! [vol newval]
+  (jolt.host/ref-put! vol :val newval) newval)
+(defn vswap! [vol f & args]
+  (vreset! vol (apply f (get vol :val) args)))
+
 (defn ffirst [coll] (first (first coll)))
 (defn nfirst [coll] (next (first coll)))
 (defn fnext  [coll] (first (next coll)))

jolt-core/clojure/core/20-coll.clj

@@ -7,6 +7,8 @@
 ;; internal Janet callers, not used by the self-hosted compiler.
 
 ;; Tiny leaves first — fns below in this tier (and 25-sorted) use them.
+(defn some? [x] (not (nil? x)))
+
 (defn identity [x] x)
 
 (defn constantly [x] (fn [& args] x))
@@ -130,6 +132,11 @@
 
 (defn split-with [pred coll] [(take-while pred coll) (drop-while pred coll)])
 
+(defn qualified-keyword? [x] (and (keyword? x) (some? (namespace x))))
+(defn simple-keyword? [x] (and (keyword? x) (nil? (namespace x))))
+(defn qualified-symbol? [x] (and (symbol? x) (some? (namespace x))))
+(defn simple-symbol? [x] (and (symbol? x) (nil? (namespace x))))
+
 (defn ident? [x] (or (keyword? x) (symbol? x)))
 
 (defn qualified-ident? [x] (or (qualified-symbol? x) (qualified-keyword? x)))
@@ -234,10 +241,8 @@
 (defn thread-bound? [& vars]
   (every? (fn [v] (__thread-bound? v)) vars))
 
-;; file-seq: the tree of paths under root (root included), directories walked
-;; via the host dir primitives. Paths (strings), not File objects.
-(defn file-seq [root]
-  (tree-seq __dir? __list-dir root))
+(defn key [e] (if (map-entry? e) (nth e 0) (throw (ex-info "key requires a map entry" {}))))
+(defn val [e] (if (map-entry? e) (nth e 1) (throw (ex-info "val requires a map entry" {}))))
 
 ;; --- Ad-hoc hierarchies (stage 3) — Clojure's canonical pure-map port. -----
 ;; A hierarchy is {:parents {tag #{parents}} :ancestors {tag #{all}} 
@@ -322,6 +327,10 @@
 (defn counted? [x]
   (or (vector? x) (map? x) (set? x) (list? x) (string? x)))
 (defn indexed? [x] (vector? x))
+;; sorted? is defined by the next tier (25-sorted) — declared here so this
+;; tier compiles (forward references are analysis errors now, jolt-2o7.3).
+(declare sorted?)
+
 (defn reversible? [x] (or (vector? x) (sorted? x)))
 (defn seqable? [x]
   (or (nil? x) (coll? x) (string? x)))
@@ -331,10 +340,8 @@
 (defn float? [x] (double? x))
 (defn infinite? [x] (and (number? x) (or (= x ##Inf) (= x ##-Inf))))
 
-(defn qualified-keyword? [x] (and (keyword? x) (some? (namespace x))))
-(defn simple-keyword? [x] (and (keyword? x) (nil? (namespace x))))
-(defn qualified-symbol? [x] (and (symbol? x) (some? (namespace x))))
-(defn simple-symbol? [x] (and (symbol? x) (nil? (namespace x))))
+;; qualified-/simple- keyword?/symbol? moved above qualified-ident? (forward
+;; references are analysis errors now — jolt-2o7.3).
 
 ;; find: the map entry [k v] when k is present (nil values included), nil
 ;; otherwise. contains? gives vectors-by-index for free, matching Clojure.
@@ -342,6 +349,19 @@
   (when (contains? m k) [k (get m k)]))
 
 ;; realized?: defined on the pending types only (delay/lazy-seq/future read
+;; Tagged-value predicates. The constructors (atom/volatile!/...) stay in Janet,
+;; but every tagged value carries its kind under :jolt/type (records under
+;; :jolt/deftype), reachable via get — which is nil on non-tables — so the
+;; predicates are pure over get and move out of the seed.
+(defn atom? [x]               (= (get x :jolt/type) :jolt/atom))
+(defn volatile? [x]           (= (get x :jolt/type) :jolt/volatile))
+(defn reader-conditional? [x] (= (get x :jolt/type) :jolt/reader-conditional))
+(defn tagged-literal? [x]     (= (get x :jolt/type) :jolt/tagged-literal))
+(defn record? [x]             (some? (get x :jolt/deftype)))
+(defn uuid? [x]               (= (get x :jolt/type) :jolt/uuid))
+(defn inst? [x]               (= (get x :jolt/type) :jolt/inst))
+(defn char? [x]               (= (get x :jolt/type) :jolt/char))
+
 ;; their realization slot; promises/atoms always-realized), error otherwise.
 (defn realized? [x]
   (cond
@@ -379,6 +399,14 @@
   ([coll] (dorun coll) coll)
   ([n coll] (dorun n coll) coll))
 
+;; spread: (spread [1 2 [3 4]]) => (1 2 3 4) — list*'s variadic helper
+;; (private in Clojure).
+(defn- spread [arglist]
+  (cond
+    (nil? arglist) nil
+    (nil? (next arglist)) (seq (first arglist))
+    :else (cons (first arglist) (spread (next arglist)))))
+
 ;; list*: cons the leading args onto the final seq argument.
 (defn list*
   ([args] (seq args))
@@ -388,14 +416,6 @@
   ([a b c d & more]
    (cons a (cons b (cons c (cons d (spread more)))))))
 
-;; spread: (spread [1 2 [3 4]]) => (1 2 3 4) — list*'s variadic helper
-;; (private in Clojure; defined after use is fine, vars resolve at call time).
-(defn- spread [arglist]
-  (cond
-    (nil? arglist) nil
-    (nil? (next arglist)) (seq (first arglist))
-    :else (cons (first arglist) (spread (next arglist)))))
-
 ;; print-str family: print/println/prn into a captured *out*.
 (defn print-str [& xs] (__with-out-str (fn* [] (apply print xs))))
 (defn println-str [& xs] (__with-out-str (fn* [] (apply println xs))))
@@ -495,6 +515,12 @@
                          (mapcat walk (children node))))))]
     (walk root)))
 
+;; file-seq: the tree of paths under root (root included), directories walked
+;; via the host dir primitives. Paths (strings), not File objects. (Lives below
+;; tree-seq: forward references are analysis errors now — jolt-2o7.3.)
+(defn file-seq [root]
+  (tree-seq __dir? __list-dir root))
+
 ;; Canonical flatten via tree-seq: the leaves (non-sequential nodes) in order.
 ;; Flattens lists too (sequential?), matching Clojure/CLJS.
 (defn flatten [coll]
@@ -603,19 +629,6 @@
 (defn ex-cause [e]
   (let [e (ex-unwrap e)] (if (ex-info-val? e) (get e :cause) nil)))
 
-;; Tagged-value predicates. The constructors (atom/volatile!/...) stay in Janet,
-;; but every tagged value carries its kind under :jolt/type (records under
-;; :jolt/deftype), reachable via get — which is nil on non-tables — so the
-;; predicates are pure over get and move out of the seed.
-(defn atom? [x]               (= (get x :jolt/type) :jolt/atom))
-(defn volatile? [x]           (= (get x :jolt/type) :jolt/volatile))
-(defn reader-conditional? [x] (= (get x :jolt/type) :jolt/reader-conditional))
-(defn tagged-literal? [x]     (= (get x :jolt/type) :jolt/tagged-literal))
-(defn record? [x]             (some? (get x :jolt/deftype)))
-(defn uuid? [x]               (= (get x :jolt/type) :jolt/uuid))
-(defn inst? [x]               (= (get x :jolt/type) :jolt/inst))
-(defn char? [x]               (= (get x :jolt/type) :jolt/char))
-
 ;; inst-ms: epoch milliseconds of an instant; throws on a non-inst (Clojure
 ;; protocol behavior).
 (defn inst-ms [x]
@@ -635,6 +648,10 @@
   ([n step coll] (map vec (partition n step coll)))
   ([n step pad coll] (map vec (partition n step pad coll))))
 
+;; partition-all is a lazy-tier fn (40-lazy) — declared so partitionv-all
+;; compiles; bound by the time anything calls it.
+(declare partition-all)
+
 (defn partitionv-all
   ([n coll] (map vec (partition-all n coll)))
   ([n step coll] (map vec (partition-all n step coll))))
@@ -684,12 +701,7 @@
 (defn set-validator! [a f]
   (jolt.host/ref-put! a :validator f) nil)
 
-;; Volatiles. The constructor (volatile!) stays native — it builds the mutable box —
-;; but vreset! sets the box's slot through ref-put! and vswap! is pure over it + get.
-(defn vreset! [vol newval]
-  (jolt.host/ref-put! vol :val newval) newval)
-(defn vswap! [vol f & args]
-  (vreset! vol (apply f (get vol :val) args)))
+;; vreset!/vswap! live in the seq tier (10-seq.clj): its transducers use them.
 
 ;; Future status predicates — pure reads of the future's :cached/:cancelled slots.
 ;; future? stays native (deref/future-cancel/realized? call it); future-call and
@@ -761,8 +773,7 @@
 
 ;; Strict, as in Clojure: an entry is what (seq m) yields (a host tuple), NOT
 ;; a plain vector — (key [1 2]) throws.
-(defn key [e] (if (map-entry? e) (nth e 0) (throw (ex-info "key requires a map entry" {}))))
-(defn val [e] (if (map-entry? e) (nth e 1) (throw (ex-info "val requires a map entry" {}))))
+;; key/val moved above the hierarchies section (underive uses them).
 
 ;; find was previously missing from jolt entirely. Presence (contains?), not
 ;; value, decides — so (find {:a nil} :a) is [:a nil]. Works on vectors by
@@ -772,7 +783,7 @@
 (defn find [m k]
   (when (contains? m k) (first {k (get m k)})))
 
-(defn some? [x] (not (nil? x)))
+;; some? lives in the top leaf block now (forward refs are errors).
 (defn true? [x] (= true x))
 (defn false? [x] (= false x))
 

jolt-core/jolt/analyzer.clj

@@ -199,7 +199,17 @@
               (case (:kind r)
                 :var (var-ref (:ns r) (:name r))
                 :host (host-ref (:name r))
-                (var-ref (compile-ns ctx) nm))))))
+                ;; :unresolved — previously emitted a var-ref that auto-interned
+                ;; an UNBOUND var, so a typo'd symbol died later as 'Cannot call
+                ;; nil as a function' with no hint which symbol (jolt-2o7.3).
+                ;; Punt to the interpreter: its resolver raises Clojure's
+                ;; 'Unable to resolve symbol' when the form actually runs (at
+                ;; eval for top-level forms, at call for fn bodies). A punt
+                ;; rather than a hard throw because runtime-interning forms
+                ;; (defmulti's setup call) legitimately reference the var they
+                ;; are about to create when nested in a non-top-level do. Real
+                ;; forward references want (declare ...), as in Clojure.
+                (uncompilable (str "Unable to resolve symbol: " nm " in this context")))))))
 
 (defn- analyze-list [ctx form env]
   (let [items (vec (form-elements form))]

src/jolt/evaluator.janet

@@ -740,7 +740,7 @@
                       # No implicit Janet fallback (Stage 3): an unresolved
                       # Clojure symbol is an error. Host access is the explicit
                       # janet/ prefix above.
-                      (error (string "Unable to resolve symbol: " name))))))))))))))))))
+                      (error (string "Unable to resolve symbol: " name " in this context"))))))))))))))))))
 (defn- parse-arg-names
   "Parse a parameter vector, handling & rest args.
   Returns {:fixed [names...] :rest name-or-nil :all [names...]}"

src/jolt/host_iface.janet

@@ -130,14 +130,23 @@
 #                                      same fallback the interpreter's resolve-sym uses
 #   {:kind :unresolved :name NAME}   — not yet defined (forward reference)
 (defn h-resolve-global [ctx sym]
-  (let [v (resolve-var ctx @{} sym)]
-    (if v
-      {:kind :var :ns (var-ns v) :name (var-name v)}
-      (let [nm (sym :name)
-            entry (in (fiber/getenv (fiber/current)) (symbol nm))]
-        (if (not (nil? entry))
-          {:kind :host :name nm}
-          {:kind :unresolved :name nm})))))
+  # Unqualified syms must resolve against the COMPILE ns (h-current-ns), not
+  # ctx-current-ns: the analyzer runs interpreted in jolt.analyzer, so during
+  # analysis resolve-var's unqualified branch looks in the wrong namespace.
+  # Before jolt-2o7.3 the analyzer's lenient fallthrough masked this (it
+  # emitted a var-ref against compile-ns for everything 'unresolved').
+  (def v (if (sym :ns)
+           (resolve-var ctx @{} sym)
+           (let [cns (ctx-find-ns ctx (h-current-ns ctx))]
+             (or (and cns (ns-find cns (sym :name)))
+                 (ns-find (ctx-find-ns ctx "clojure.core") (sym :name))))))
+  (if v
+    {:kind :var :ns (var-ns v) :name (var-name v)}
+    (let [nm (sym :name)
+          entry (in (fiber/getenv (fiber/current)) (symbol nm))]
+      (if (not (nil? entry))
+        {:kind :host :name nm}
+        {:kind :unresolved :name nm}))))
 
 (defn h-intern! [ctx ns-name nm]
   (ns-intern (ctx-find-ns ctx ns-name) nm)

src/jolt/loader.janet

@@ -24,12 +24,7 @@
     (let [ff (first form)]
       (when (and (struct? ff) (= :symbol (ff :jolt/type))) (ff :name)))))
 
-(defn eval-toplevel
-  "Evaluate one top-level form for ctx, honoring :compile?. Stateful forms always
-  interpret; otherwise the form runs through the self-hosted compile pipeline
-  (portable Clojure analyzer -> IR -> Janet back end), which falls back to the
-  interpreter for forms it can't compile. Only the compile step is guarded —
-  runtime errors in compiled code propagate (no double-eval, no hidden errors)."
+(defn- eval-toplevel-1
   [ctx form]
   # Repair point for the interpreted-fn ns swap: a body runs with current-ns
   # rebound to its defining ns and restores it on normal return; an UNWINDING
@@ -69,6 +64,26 @@
       # fiber's stack
       (propagate err fib))))
 
+(defn eval-toplevel
+  "Evaluate one top-level form for ctx, honoring :compile?. Stateful forms always
+  interpret; otherwise the form runs through the self-hosted compile pipeline
+  (portable Clojure analyzer -> IR -> Janet back end), which falls back to the
+  interpreter for forms it can't compile. Only the compile step is guarded —
+  runtime errors in compiled code propagate (no double-eval, no hidden errors)."
+  [ctx form]
+  # Clojure's top-level `do` rule: children are compiled AND evaluated one at
+  # a time, so a child's runtime effects (defmulti's var intern, requires, …)
+  # are visible while the NEXT child compiles. Without the split, (do
+  # (defmulti area …) (area …)) can't analyze — `area` only exists once the
+  # defmulti has RUN, and unresolved symbols are analysis errors now
+  # (jolt-2o7.3).
+  (if (and (array? form) (= "do" (form-head-name form)))
+    (do
+      (var res nil)
+      (each child (array/slice form 1) (set res (eval-toplevel ctx child)))
+      res)
+    (eval-toplevel-1 ctx form)))
+
 (defn load-ns
   "Load a Clojure namespace from a .clj file. Per-form routing (compile-or-
   interpret, stateful forms interpret) is shared with eval-one via eval-toplevel.

test/integration/cli-test.janet

@@ -50,9 +50,16 @@
 (check "arity error names the fn"
        (run-err "-e" "(defn afn [x] x) (afn 1 2)")
        (has "Wrong number of args (2) passed to: user/afn"))
-(check "nil-call hints at undefined symbol"
+(check "nil-call (nil value) keeps the hint"
+       (run-err "-e" "(def x nil) (x 1)")
+       (has "Cannot call nil as a function"))
+# round 3: typos die at resolve time with Clojure's message, not as nil-calls
+(check "unresolved symbol named at resolve time"
        (run-err "-e" "(undefined-fn 1)")
-       (has "undefined (misspelled?) symbol"))
+       (has "Unable to resolve symbol: undefined-fn in this context"))
+(check "typo inside fn body also resolves to the message"
+       (run-err "-e" "(defn f [] (no-such 1)) (f)")
+       (has "Unable to resolve symbol: no-such"))
 (check "trace shows the user's call chain"
        (run-err "-e" "(defn inner [x] (let [r (+ x :k)] r)) (defn outer [x] (let [v (inner x)] v)) (outer 1)")
        (fn [s] (and (string/find "at user/inner" s) (string/find "at user/outer" s))))

test/integration/compile-mode-test.janet

@@ -124,12 +124,15 @@
           "runtime error in compiled code propagates"))
 
 # Context isolation: a def in one compiled context is invisible in another. With
-# var-indirection each context has its own var cells, so b's `secret` is a
-# distinct, unbound var (nil) rather than a's 7.
+# var-indirection each context has its own var cells, so in b `secret` does not
+# resolve at all ('Unable to resolve symbol', jolt-2o7.3) rather than seeing a's 7.
 (let [a (init-cached {:compile? true}) b (init-cached {:compile? true})]
   (eval-string a "(def secret 7)")
   (assert (= 7 (ct-eval a "secret")) "def visible in its own ctx")
-  (assert (nil? (ct-eval b "secret")) "def isolated to its ctx"))
+  (def r (protect (ct-eval b "secret")))
+  (assert (and (not (r 0))
+               (string/find "Unable to resolve symbol: secret" (string/format "%q" (r 1))))
+          "def isolated to its ctx (unresolved there)"))
 
 # Redefinition is visible to already-compiled callers (var-indirection).
 (let [c (init-cached {:compile? true})]

test/integration/ir-passes-test.janet

@@ -20,12 +20,18 @@
 (check-const "(quot 7 2)" 3)
 (check-const "(mod -7 3)" 2)
 (check-const "(if (< 1 2) :yes :no)" :yes)
-# dead-branch elimination: the untaken branch never evaluates
-(check-const "(if false (this-would-not-resolve) 2)" 2)
+# dead-branch elimination: the untaken branch never evaluates (it must still
+# RESOLVE — unresolved symbols are analysis errors as in Clojure, jolt-2o7.3)
+(check-const "(if false (throw (ex-info \"boom\" {})) 2)" 2)
+# an unresolvable symbol errors even in a dead branch (analysis precedes folding)
+(assert (not ((protect (ir "(if false (this-would-not-resolve) 2)")) 0))
+        "unresolved symbol errors even in a dead branch")
 
-# non-constants stay calls; folding must be conservative
-(assert (= :invoke ((ir "(+ x 2)") :op)) "free var stays a call")
-(assert (= :invoke ((ir "(mod x 0)") :op)) "non-const args stay calls")
+# non-constants stay calls; folding must be conservative. `xq` is a real var
+# (defined here) so it resolves, but its VALUE must not be folded in.
+(api/eval-string ctx "(def xq 1)")
+(assert (= :invoke ((ir "(+ xq 2)") :op)) "var ref stays a call")
+(assert (= :invoke ((ir "(mod xq 0)") :op)) "non-const args stay calls")
 # a fold that would THROW is left for runtime
 (assert (= :invoke ((ir "(mod 5 0)") :op)) "throwing fold left to runtime")