perf(d2p1): further optimization

Author: Alireza Alavi

src/aoc/2025/day2/assets/profiling2.png

@@ -216,7 +216,41 @@
 ; ```
 ; Boy, Oh boy! We got ~87.3% increase in performance after our optimizations!\
 ; These are still cookie-cutter optimizations, tough. I think we can get much further
-; with more in depth opts, but I don't know how to.
+; with more in depth opts.
+
+; Our profiling after the optimizations shows that now our most resource consuming
+; part is the `mapv` function in `num->digits-opt`.
+(clerk/image "src/aoc/2025/day2/assets/profiling2.png")
+
+; Let's see how we can improve it.
+;
+; Now that I look at it, the previous `num->digits-opt` looks absolutely stupid.\
+; There are two calls to `str` and there is no reason to use threading `->>` here.
+; Just makes it confusing.
+; - Convert `n` to string directly, feed it as `coll` to `mapv`.
+; - Use `int` substract the ASCII offset to get the number.
+
+(defn num->digits-opt2
+  [^long n]
+  (mapv #(- (int %) 48) (Long/toString n)))
+
+; ```clojure
+; (with-redefs-fn {#'repeated-twice? repeated-twice-opt?
+;                  #'num->digits num->digits-opt2}
+;  #(c/quick-bench (sum-invalid-ids (parse-input "src/aoc/2025/day2/input.txt"))))
+; =>
+; (out) Evaluation count : 6 in 6 samples of 1 calls.
+; (out)              Execution time mean : 927.130309 ms
+; (out)     Execution time std-deviation : 30.063344 ms
+; (out)    Execution time lower quantile : 898.622286 ms ( 2.5%)
+; (out)    Execution time upper quantile : 972.890815 ms (97.5%)
+; (out)                    Overhead used : 6.691813 ns
+; (out)
+; (out) Found 1 outliers in 6 samples (16.6667 %)
+; (out) 	low-severe	 1 (16.6667 %)
+; (out)  Variance from outliers : 13.8889 % Variance is moderately inflated by outliers
+; ```
+; This make our total optimizations **~90% faster** than our initial solution.
 
 ; ## Running all tests
 (run-tests)