|
| 1 | +# ============================================================================= |
| 2 | +# Harmonic-distinctive collections: the OMC-only data structures |
| 3 | +# ============================================================================= |
| 4 | +# Three collections that only make sense in a language with first-class |
| 5 | +# Fibonacci attractors and HIM scores: |
| 6 | +# |
| 7 | +# harmonic_set — auto-deduplicates near-fold-equivalent values |
| 8 | +# harmonic_pq — priority queue ranked by HIM score (low = dominant) |
| 9 | +# harmonic_index — Fibonacci-attractor bucketing for sub-linear lookup |
| 10 | +# |
| 11 | +# Pure-functional API (each op returns a fresh structure, mirroring the |
| 12 | +# OMC convention for arrays/dicts that pass by value). Built on top of |
| 13 | +# existing dict + fold + phi.him primitives — no new VM opcodes needed. |
| 14 | +# |
| 15 | +# Run: |
| 16 | +# ./target/release/omnimcode-standalone examples/harmonic_collections.omc |
| 17 | +# OMC_VM=1 ./target/release/omnimcode-standalone examples/harmonic_collections.omc |
| 18 | +# ============================================================================= |
| 19 | + |
| 20 | +# ---- harmonic_set --------------------------------------------------------- |
| 21 | +# Two values are "equivalent" iff their Fibonacci attractors match. So 23 |
| 22 | +# and 21 BOTH live on the 21-attractor; adding both leaves only the first |
| 23 | +# (canonical representative). 89 vs 144 stay distinct. |
| 24 | +# |
| 25 | +# Internal layout: { fold_key_as_string : original_value }. Keeps insertion |
| 26 | +# order via dict's BTreeMap (sorted, deterministic) and lets membership |
| 27 | +# tests run in O(log n) via dict_has. |
| 28 | + |
| 29 | +fn hset_new() { |
| 30 | + return {}; |
| 31 | +} |
| 32 | + |
| 33 | +fn _hset_key(v) { |
| 34 | + # The fold-attractor IS the equivalence class. Stringify because dict |
| 35 | + # keys are strings. |
| 36 | + return concat_many("", fold(v)); |
| 37 | +} |
| 38 | + |
| 39 | +fn hset_add(s, v) { |
| 40 | + h key = _hset_key(v); |
| 41 | + h out = dict_merge(s, {}); # shallow copy |
| 42 | + if dict_has(out, key) == 1 { |
| 43 | + return out; # already represented on this attractor |
| 44 | + } |
| 45 | + dict_set(out, key, v); |
| 46 | + return out; |
| 47 | +} |
| 48 | + |
| 49 | +fn hset_has(s, v) { |
| 50 | + return dict_has(s, _hset_key(v)); |
| 51 | +} |
| 52 | + |
| 53 | +fn hset_len(s) { |
| 54 | + return dict_len(s); |
| 55 | +} |
| 56 | + |
| 57 | +fn hset_items(s) { |
| 58 | + return dict_values(s); |
| 59 | +} |
| 60 | + |
| 61 | +# Set union — right-hand wins on attractor collision (matches dict_merge |
| 62 | +# semantics so users don't have to remember a different convention). |
| 63 | +fn hset_union(a, b) { |
| 64 | + return dict_merge(a, b); |
| 65 | +} |
| 66 | + |
| 67 | +# ---- harmonic_pq ---------------------------------------------------------- |
| 68 | +# Priority queue ranked by HIM score. Lower HIM = more harmonically |
| 69 | +# dominant (a "purer" value, closer to an attractor). pop returns the |
| 70 | +# dominant element. |
| 71 | +# |
| 72 | +# Internal layout: { him_score_str : list_of_values_with_that_score }. |
| 73 | +# A list-of-lists handles HIM ties naturally. Iteration order from |
| 74 | +# dict_keys is sorted-string-ascending — which happens to put "0.000" |
| 75 | +# before "0.382" before "0.459", giving us the right priority order |
| 76 | +# WITHOUT a sort step on each pop. Saves real time on large queues. |
| 77 | + |
| 78 | +fn hpq_new() { |
| 79 | + return {}; |
| 80 | +} |
| 81 | + |
| 82 | +fn _hpq_him_key(v) { |
| 83 | + # Use 6-decimal padding so dict's lexicographic key sort matches |
| 84 | + # numeric order. phi.him returns a float in [0, 1]. |
| 85 | + h h_score = phi.him(v); |
| 86 | + h scaled = to_int(h_score * 1000000); |
| 87 | + # Zero-pad to 7 digits so lexicographic order matches numeric. |
| 88 | + h s = concat_many("", scaled); |
| 89 | + h n = str_len(s); |
| 90 | + h pad_count = 7 - n; |
| 91 | + h padded = ""; |
| 92 | + h i = 0; |
| 93 | + while i < pad_count { |
| 94 | + padded = padded + "0"; |
| 95 | + i = i + 1; |
| 96 | + } |
| 97 | + return padded + s; |
| 98 | +} |
| 99 | + |
| 100 | +fn hpq_push(q, v) { |
| 101 | + h key = _hpq_him_key(v); |
| 102 | + h out = dict_merge(q, {}); |
| 103 | + h existing = dict_get(out, key, []); |
| 104 | + h fresh = arr_concat(existing, [v]); |
| 105 | + dict_set(out, key, fresh); |
| 106 | + return out; |
| 107 | +} |
| 108 | + |
| 109 | +# Returns [value, new_queue]. If empty: [null, q]. |
| 110 | +fn hpq_pop(q) { |
| 111 | + h ks = dict_keys(q); |
| 112 | + if arr_len(ks) == 0 { |
| 113 | + return [null, q]; |
| 114 | + } |
| 115 | + # First key is the lowest HIM score — that's the dominant element. |
| 116 | + h k = arr_get(ks, 0); |
| 117 | + h bucket = dict_get(q, k); |
| 118 | + h v = arr_get(bucket, 0); |
| 119 | + h rest = arr_slice(bucket, 1, arr_len(bucket)); |
| 120 | + h out = dict_merge(q, {}); |
| 121 | + if arr_len(rest) == 0 { |
| 122 | + dict_del(out, k); |
| 123 | + } else { |
| 124 | + dict_set(out, k, rest); |
| 125 | + } |
| 126 | + return [v, out]; |
| 127 | +} |
| 128 | + |
| 129 | +fn hpq_peek(q) { |
| 130 | + h ks = dict_keys(q); |
| 131 | + if arr_len(ks) == 0 { return null; } |
| 132 | + h bucket = dict_get(q, arr_get(ks, 0)); |
| 133 | + return arr_get(bucket, 0); |
| 134 | +} |
| 135 | + |
| 136 | +fn hpq_len(q) { |
| 137 | + h ks = dict_keys(q); |
| 138 | + h total = 0; |
| 139 | + h i = 0; |
| 140 | + while i < arr_len(ks) { |
| 141 | + total = total + arr_len(dict_get(q, arr_get(ks, i))); |
| 142 | + i = i + 1; |
| 143 | + } |
| 144 | + return total; |
| 145 | +} |
| 146 | + |
| 147 | +# ---- harmonic_index ------------------------------------------------------- |
| 148 | +# An index that buckets every key by its Fibonacci attractor. Lookup of |
| 149 | +# a query key folds the query to its attractor and returns the entire |
| 150 | +# bucket — sub-linear in the FULL data set, but linear within an |
| 151 | +# attractor. For data that clusters along the Fibonacci spine |
| 152 | +# (resonance-aligned domains), this is significantly faster than a |
| 153 | +# naive scan AND it surfaces "fuzzy match" semantics for free. |
| 154 | +# |
| 155 | +# Internal layout: { fold_key_str : list_of_(key, value)_pairs }. |
| 156 | + |
| 157 | +fn hidx_new() { |
| 158 | + return {}; |
| 159 | +} |
| 160 | + |
| 161 | +fn hidx_insert(idx, key, value) { |
| 162 | + h k = _hset_key(key); |
| 163 | + h out = dict_merge(idx, {}); |
| 164 | + h existing = dict_get(out, k, []); |
| 165 | + h pair = [key, value]; |
| 166 | + h fresh = arr_concat(existing, [pair]); |
| 167 | + dict_set(out, k, fresh); |
| 168 | + return out; |
| 169 | +} |
| 170 | + |
| 171 | +# Returns the array of (key, value) pairs whose key folds to the same |
| 172 | +# attractor as `query`. Empty array if no match. |
| 173 | +fn hidx_lookup(idx, query) { |
| 174 | + return dict_get(idx, _hset_key(query), []); |
| 175 | +} |
| 176 | + |
| 177 | +# All values across all buckets. Useful for length / iteration. |
| 178 | +fn hidx_size(idx) { |
| 179 | + h ks = dict_keys(idx); |
| 180 | + h total = 0; |
| 181 | + h i = 0; |
| 182 | + while i < arr_len(ks) { |
| 183 | + total = total + arr_len(dict_get(idx, arr_get(ks, i))); |
| 184 | + i = i + 1; |
| 185 | + } |
| 186 | + return total; |
| 187 | +} |
| 188 | + |
| 189 | +# How many distinct attractors hold data? Useful for sparsity diagnostics. |
| 190 | +fn hidx_attractors(idx) { |
| 191 | + return dict_len(idx); |
| 192 | +} |
| 193 | + |
| 194 | +# ============================================================================= |
| 195 | +# Demo |
| 196 | +# ============================================================================= |
| 197 | + |
| 198 | +println("=== harmonic_set: dedup-by-attractor ======================"); |
| 199 | +h s = hset_new(); |
| 200 | +s = hset_add(s, 21); |
| 201 | +s = hset_add(s, 23); # folds to 21 — duplicate, ignored |
| 202 | +s = hset_add(s, 89); |
| 203 | +s = hset_add(s, 91); # folds to 89 — duplicate, ignored |
| 204 | +s = hset_add(s, 144); |
| 205 | +s = hset_add(s, 200); # folds to 233 — distinct |
| 206 | +println(concat_many(" added 6 values, len = ", hset_len(s), |
| 207 | + " (because 23 folds to 21, 91 folds to 89)")); |
| 208 | +println(concat_many(" members: ", hset_items(s))); |
| 209 | +println(concat_many(" has(20)? ", hset_has(s, 20), |
| 210 | + " (folds to 21, present)")); |
| 211 | +println(concat_many(" has(50)? ", hset_has(s, 50), |
| 212 | + " (folds to 55, absent)")); |
| 213 | +println(""); |
| 214 | + |
| 215 | +println("=== harmonic_pq: HIM-ranked priority queue ================"); |
| 216 | +h q = hpq_new(); |
| 217 | +q = hpq_push(q, 100); # off-attractor → high HIM |
| 218 | +q = hpq_push(q, 89); # exact Fibonacci → HIM = 0 |
| 219 | +q = hpq_push(q, 50); # near 55 → middling HIM |
| 220 | +q = hpq_push(q, 144); # exact Fibonacci → HIM = 0 |
| 221 | +q = hpq_push(q, 23); # near 21 |
| 222 | +println(concat_many(" pushed 5 values, len = ", hpq_len(q))); |
| 223 | +println(concat_many(" peek (most dominant) = ", hpq_peek(q))); |
| 224 | +println(" popping in HIM order:"); |
| 225 | +h still = q; |
| 226 | +while hpq_len(still) > 0 { |
| 227 | + h pair = hpq_pop(still); |
| 228 | + h v = arr_get(pair, 0); |
| 229 | + still = arr_get(pair, 1); |
| 230 | + println(concat_many(" pop -> ", v, " (HIM=", phi.him(v), ")")); |
| 231 | +} |
| 232 | +println(""); |
| 233 | + |
| 234 | +println("=== harmonic_index: attractor-bucketed lookup ============="); |
| 235 | +h idx = hidx_new(); |
| 236 | +# Insert 12 user records keyed by user_id (the harmonic value). |
| 237 | +idx = hidx_insert(idx, 21, "alice"); |
| 238 | +idx = hidx_insert(idx, 23, "bob"); # same attractor as alice (21) |
| 239 | +idx = hidx_insert(idx, 19, "carol"); # also folds to 21 |
| 240 | +idx = hidx_insert(idx, 89, "dave"); |
| 241 | +idx = hidx_insert(idx, 90, "eve"); # folds to 89 |
| 242 | +idx = hidx_insert(idx, 144, "frank"); |
| 243 | +idx = hidx_insert(idx, 142, "grace"); # folds to 144 |
| 244 | +idx = hidx_insert(idx, 233, "heidi"); |
| 245 | +idx = hidx_insert(idx, 55, "ivan"); |
| 246 | +idx = hidx_insert(idx, 8, "judy"); |
| 247 | +idx = hidx_insert(idx, 3, "kim"); |
| 248 | +idx = hidx_insert(idx, 377, "leo"); |
| 249 | + |
| 250 | +println(concat_many(" total records: ", hidx_size(idx), |
| 251 | + " across ", hidx_attractors(idx), " attractors")); |
| 252 | + |
| 253 | +println(" query: who is near user_id 22?"); |
| 254 | +h hits = hidx_lookup(idx, 22); |
| 255 | +h k = 0; |
| 256 | +while k < arr_len(hits) { |
| 257 | + h pair = arr_get(hits, k); |
| 258 | + println(concat_many(" user_id=", arr_get(pair, 0), |
| 259 | + " -> ", arr_get(pair, 1))); |
| 260 | + k = k + 1; |
| 261 | +} |
| 262 | + |
| 263 | +println(" query: who is near user_id 91?"); |
| 264 | +h hits2 = hidx_lookup(idx, 91); |
| 265 | +h k2 = 0; |
| 266 | +while k2 < arr_len(hits2) { |
| 267 | + h pair = arr_get(hits2, k2); |
| 268 | + println(concat_many(" user_id=", arr_get(pair, 0), |
| 269 | + " -> ", arr_get(pair, 1))); |
| 270 | + k2 = k2 + 1; |
| 271 | +} |
| 272 | + |
| 273 | +println(" query: who is near user_id 999? (folds off-spine to 610)"); |
| 274 | +h hits3 = hidx_lookup(idx, 999); |
| 275 | +println(concat_many(" -> ", arr_len(hits3), " hits")); |
| 276 | + |
| 277 | +println(""); |
| 278 | +println("=== Done ===================================================="); |
0 commit comments