The map build already used a transient map, but each bucket was rebuilt with
a persistent (conj (get ret k []) x) per element — an O(log n) trie path
rebuild + alloc each. A coarse grouping (few large buckets) was bound on that
conj, not the map build. Buckets are now native arrays (transient vectors,
O(1) push) frozen once; distinct keys are tracked in a side vector so the
buckets freeze in place with no second map rebuild. A bucket's first element
stays a cheap persistent [x] and only promotes to a transient on the second,
so an all-singletons grouping pays no transient alloc.
coarse (10/100 buckets, 50k): ~313ms -> ~125ms (~2.5x)
2 buckets (50k): ~322ms -> ~129ms (~2.5x)
all-unique (50k): ~949ms -> ~892ms (no regression)
Surfaced a latent bug: canon-key returns nil for a nil key and Janet tables
drop a nil key, so the canon-keyed transient map silently lost a nil-key
entry — group-by/frequencies/assoc!/into{} dropped the whole nil bucket
((group-by identity [nil nil 1]) gave {1 [1]}, not {nil [nil nil], 1 [1]}).
Route nil through a sentinel (tbl-key) at the transient-map keying sites;
persistent!/count/dissoc! work unchanged since the real [nil v] pair is kept
as the stored value, and phm already has its own has-nil slot. The transient
set has the analogous bug (needs phs nil support) — filed separately.
Co-authored-by: Yogthos <yogthos@gmail.com>
into {}, frequencies, group-by, set, into #{} and persistent! all built
their result by folding an immutable assoc/conj per element — each call
rebuilt the O(log32 n) trie path and allocated a fresh wrapper. Add a
one-pass bottom-up HAMT builder (phm-from-pairs) and route the builders
through it, the map/set analog of the pvec bulk build in #153.
phm-from-pairs partitions entries by hash and constructs the bin/array/
collision nodes directly, with the same bin<=16 / array-node>=17 promotion
the incremental path uses — so the trie is byte-identical to one built by
phm-assoc (validated across the size and branching boundaries, including
hash collisions, duplicate keys and the nil key). persistent! map/set and
the set constructor bulk-build; into {} keeps the small-scalar-map-stays-a
-struct rule via bulk-map-from-pairs; frequencies/group-by switch to the
canonical transient form and ride the fast persistent!.
50k A/B: into {} 704->270ms, frequencies 582->160, set 615->241,
into #{} 702->240, group-by 1358->919 (bound on persistent vector conj).
Gate: conformance x3, full suite (4718 >= baseline), new maps/sets bulk
boundary specs.
Co-authored-by: Yogthos <yogthos@gmail.com>
recompile-defns! is the defn analog of recompile-macros!: pre/at-kernel
overlay defns (00-syntax's destructure and friends; the kernel tier too in
interpret mode) load as interpreted closures, the evaluator stashes their fn
source on the var (:defn-src, scoped by a flag only api/load-core-overlay!
sets), and the end-of-init pass compiles them and swaps the var root. With
that in place, keys/vals/empty? — the fns the 00-syntax expanders call at
expansion time — move to the top of 00-syntax as raw fn* defs (canonical:
keys/vals project (seq m), so sorted maps come back in comparator order and
(keys {}) is nil; empty? keeps O(1) count dispatch with seq's cell check only
for the lazy/list fallback). The sorted tier drops its now-dead :keys/:vals
ops.
Correctness fixes that surfaced once the gate was run with a REAL exit code
(the previous 'jpm test | grep' gates reported grep's exit and masked spec
failures across #48-#50):
- map conj is strict again: a non-nil/non-map arg must be a 2-element vector
('Vector arg to map conj must be a pair'), and merge inherits it — the
batch-2 canonical merge had silently dropped the validation
- conj onto a lazy seq prepends (it fell into the MAP fallback); upstream
clojure.data/diff relies on (conj seq x) via set/union over keys, so diff
now matches Clojure exactly
- (seq {}) / (seq #{}) / empty phm are nil, not ()
- key/val are strict (a plain vector is not an entry); find mints a REAL
entry as the first entry of a one-entry map, nil values intact
- the sci avoid-method-too-large stub passes its registry map through
instead of returning a raw host table (strict conj rejected it; sci's
clojure-core registry is also no longer discarded)
Test updates: lazy-infinite pins take-nth realization at 5 (was 7 — the
canonical lazy impl realizes fewer); self-host asserts the analyzer IS loaded
in interpret mode (compiled expanders, PR #50) and is NOT in the
:compile-macros? false oracle. 18 new maps-spec rows.
Gate: jpm test exit 0 (verified directly, not through a pipe), conformance
326x3, suite 4698 >= 4660.
Fifteen vars from the spec coverage gap (docs/spec/coverage.md):
parse-long/parse-double/parse-boolean (strict validation; scan-number alone
accepts 0x10), newline, current-time-ms (host clock for time), update-keys/
update-vals (PHM base, collisions last-wins), partitionv/partitionv-all/
splitv-at (lazy seqs of vectors; splitv-at's tail stays a seq, matching the
reference), with-redefs/with-redefs-fn (roots restored on throw), time,
macroexpand (expand-1 to fixpoint), alias/ns-unalias (write BOTH alias stores
— require :as uses string-keyed :imports while ns-aliases reads :aliases;
split filed), ns-publics (symbol-keyed map; publics == interns, no privacy).
Three pre-existing bugs fixed along the way:
- (partition n step pad coll) misparsed pad as the coll and returned ()
- (require 'bare.symbol) rejected — only vector specs were accepted
- analyze-form leaked the interpreted analyzer's ns on a punt: a throw out of
the analyzer left current-ns=jolt.analyzer and :compile-ns set, so the
fallback interpretation resolved user vars against the wrong namespace
(bit (var user-sym) under compile mode)
And one overlay-authoring landmine documented in-code: a 20-coll fn must not
use 30-macros macros (with-redefs-fn's dotimes compiled as a forward ref that
resolved to the macro fn at runtime) — loop/recur instead.
Gate: conformance 316x3 (+14 rows), suite 4324 pass / 78 clean (was 4081/72;
parse_*/update_* files now contribute), baselines raised, all specs+unit,
fixpoint, self-host, sci, staged. Coverage: missing-portable 35 -> 20.
adds self-hosted compiler is functionally:
- The default compile path is the portable pipeline using jolt.analyzer (Clojure) → host-neutral IR → backend.janet.
- The analyzer is itself Clojure, compiled by jolt for true self-hosting.
- bootstrap-fixpoint passes (stage1 == stage2 == stage3): rebuilding the compiler on its own output.
- clojure.core is now self-hosted in the overlay.
- Stateful forms (defmacro/ns/deftype/defmulti/require/in-ns) are interpreted by design.
- A map entry is a 2-element tuple (Jolt produces tuples only from map iteration;
vector literals are pvecs, lists are arrays). key/val/map-entry? now accept a
2-tuple and reject a plain vector, matching Clojure's MapEntry-vs-vector
distinction — no metadata needed, the representations already differ.
- min-key/max-key reproduce Clojure's NaN-aware folding (2-arg strict </>, then
<=/>=) and require numeric keys (NaN allowed, strings throw).
- subvec coerces float/NaN indices like (int ...) (truncate, NaN->0) then
bounds-checks, instead of throwing on non-integers.
min_key 35/14 -> 49/0 (clean); key/val recover the 2-vector cases; subvec floats
fixed. clojure-test-suite pass 3898->3921. Updated conformance-test (key/val now
needs a real entry). spec: map/map-entry-&-key-ordering (14). jpm test green.
merge now throws when a non-first arg is a scalar, a set, a list, or a
wrong-length vector (a length-2 vector or a map still merge). Other map-like
tables (records/sorted-maps/host tables, e.g. SCI's namespaces) keep the lenient
conj path so the SCI bootstrap still loads.
merge 29/11/2 -> 35/3/4. clojure-test-suite pass 3874->3880.
spec: 5 merge strictness cases. jpm test green.
- assoc on a vector bounds-checks the index (0..count); out-of-range/negative throw
- dissoc throws on non-maps (numbers/sequences/sets/scalars); nil ok; records/
sorted-maps/meta-maps still handled
- count throws on scalars (numbers/keywords/symbols/booleans/chars)
- subvec validates vector type and 0<=start<=end<=count
- numerator/denominator always throw (Jolt has no ratio type)
- min-key/max-key throw on no values
count 18/2; dissoc 19/3; assoc 36/6; subvec 26/3/5; numerator/denominator throw
cases pass. clojure-test-suite pass 3840->3864. spec: map/strictness (16). jpm test green.
Close jolt-do7. Maps/sets keyed by a collection (a map, vector, ...) now compare
by value instead of Janet identity:
- PHM/PHS hash and compare keys through an injected canonicalizer (collection
keys -> value-hashable struct/tuple); keys are still stored as-is
- map literals and core-assoc promote to a phm when a key is a collection
- frequencies/group-by use a phm base so collection elements/keys dedup by value
- set equality is value-based (from earlier)
Real bugs found and fixed along the way:
- set literals #{(inc 1)} did not evaluate their elements (stored raw forms!)
- the REPL printer rendered phm maps as {} (they hit the deftype branch); now a
phm branch prints entries
Added spec cases (maps/collection-keys, sets/literals & value elements).
conformance 218/218, jpm test green.
Add behavioral spec suites for the collection API (~180 cases). Writing them
surfaced and fixed real bugs:
- (count nil) errored; now 0 (Clojure semantics)
- (repeat x) and (repeatedly f) — the infinite 1-arg arities — were unsupported;
now return lazy infinite seqs
- set equality used deep= (representation-sensitive), so a set of map literals
could differ from an equal set built differently; now value-based (each
element value-equal to some element of the other)
Known limitation filed: phm-typed maps (from hash-map) used as set elements /
map keys hash by identity (map *literals* are structs and work).
jpm test green.
