speedup

[schochastics]

runtime is ok, but given how fast ada-url is by itself, there is room to improvement at a) the interface R/C++ and b)the URLencoding to fix UTF8 support (see #1)

[chainsawriot]

header only: https://github.com/mmarkeloff/cpp-url-decode

ada itself also has something similar. But the above implementation is simple enough to be understood.

[schochastics]

and here I am writing my own function :D

CharacterVector url_decode_cpp(CharacterVector URLs) {
  return sapply(URLs, [](const String& URL) {
    std::string input = URL;
    std::string output;
    size_t i = 0;

    while (i < input.length()) {
      if (input[i] != '%') {
        output += input[i];
        i++;
      } else {
        int value;
        sscanf(input.substr(i + 1, 2).c_str(), "%x", &value);
        output += static_cast<char>(value);
        i += 3;
      }
    }

    return output;
  });
}

bench::mark(
    url_decode_cpp("https://www.google.co.jp/search?q=\u30c9\u30a4\u30c4"),
    URLdecode("https://www.google.co.jp/search?q=\u30c9\u30a4\u30c4"),
    iterations = 1000, check=FALSE
)
# A tibble: 2 × 13
  expression    min  median `itr/sec` mem_alloc `gc/sec` n_itr  n_gc total_time result
  <bch:expr> <bch:> <bch:t>     <dbl> <bch:byt>    <dbl> <int> <dbl>   <bch:tm> <list>
1 "url_deco…  4.4µs  5.29µs   164215.    2.49KB      0    1000     0     6.09ms <NULL>
2 "URLdecod… 31.3µs 36.53µs    25980.        0B     26.0   999     1    38.45ms <NULL>
# ℹ 3 more variables: memory <list>, time <list>, gc <list>

But I will check out the header only thing

[schochastics]

@chainsawriot ok this is surprising to me that my implementation is as fast. I would take that but will do some robustness checks

urls <- rep("https://www.google.co.jp/search?q=\u30c9\u30a4\u30c4",5000)
bench::mark(
    cpp_david = url_decode_cpp(urls),
    URLdecode = URLdecode(urls),
    cpp_header = url_decode_header(urls),
    iterations = 1, check=FALSE
)
# A tibble: 3 × 13
  expression      min   median `itr/sec` mem_alloc `gc/sec` n_itr  n_gc total_time result memory    
  <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl> <int> <dbl>   <bch:tm> <list> <list>    
1 cpp_david    4.38ms   4.38ms    228.      41.6KB      0       1     0     4.38ms <NULL> <Rprofmem>
2 URLdecode  153.79ms 153.79ms      6.50    39.1KB     13.0     1     2   153.79ms <NULL> <Rprofmem>
3 cpp_header   4.98ms   4.98ms    201.      41.6KB      0       1     0     4.98ms <NULL> <Rprofmem>
# ℹ 2 more variables: time <list>, gc <list>

[schochastics]

URLdecode speedup done with #25

ref #13

urls <- rep("https://www.google.co.jp/search?q=\u30c9\u30a4\u30c4",10000)
bench::mark(
    cpp = url_decode(urls),
    URLdecode = URLdecode(urls),
    iterations = 1, check=FALSE
)
# A tibble: 2 × 13
  expression      min   median `itr/sec` mem_alloc `gc/sec` n_itr  n_gc total_time result
  <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl> <int> <dbl>   <bch:tm> <list>
1 cpp          10.3ms   10.3ms     96.9     80.7KB      0       1     0     10.3ms <NULL>
2 URLdecode   390.6ms  390.6ms      2.56    78.2KB     12.8     1     5    390.6ms <NULL>
# ℹ 3 more variables: memory <list>, time <list>, gc <list>

[schochastics]

Benchmark datasets:

• https://github.com/ada-url/url-various-datasets
• data("testdt_tracking", package = "webtrackR")

[schochastics]

remove utf8 (need to set R>=4.2)

[schochastics]

public_suffix() at least in c++?

[chainsawriot]

• Experiment: Create the data frame first Rcpp_ada_parse

won't speed up