Function cbm_lsh_query

src/simhash/minhash.c:453–490

}

void cbm_lsh_query(const cbm_lsh_index_t *idx, const cbm_minhash_t *fp,
                   const cbm_lsh_entry_t ***out, int *count) {
    *out = NULL;
    *count = 0;

    if (!idx || !fp) {
        return;
    }

    cbm_lsh_index_t *mut_idx = (cbm_lsh_index_t *)idx;
    mut_idx->result_count = 0;

    /* O(1) dedup via open-addressing hash set */
    seen_set_t seen;
    seen_set_init(&seen);

    for (int b = 0; b < CBM_LSH_BANDS; b++) {
        uint32_t h = band_hash(fp, b);
        const lsh_bucket_t *bucket = &idx->bands[b][h];
        /* Skip oversized buckets — noise from trivially similar utility functions */
        if (bucket->count > MAX_BUCKET_SIZE) {
            continue;
        }
        for (int i = 0; i < bucket->count; i++) {
            const cbm_lsh_entry_t *candidate = &idx->entries[bucket->items[i]];
            if (!seen_set_insert(&seen, candidate->node_id)) {
                continue; /* already seen */
            }
            if (!result_push(mut_idx, candidate)) {
                break;
            }
        }
    }

    seen_set_free(&seen);
    *out = mut_idx->result_buf;
    *count = mut_idx->result_count;
}

int cbm_lsh_query_into(const cbm_lsh_index_t *idx, const cbm_minhash_t *fp,
                       const cbm_lsh_entry_t **out_buf, int out_cap) {