Feature #13943: Use unpack1 instead of unpack(template)[0] in erb.rb

• Status: Closed
• Assignee: znz (Kazuhiro NISHIYAMA)
• Target version: -

[znz (Kazuhiro NISHIYAMA)]

I think unpack1(template) is faster than unpack(template)[0].

Attached patch for lib/erb.rb.

[k0kubun (Takashi Kokubun)]

Could you put benchmark result?

[wyhaines (Kirk Haines)]

It should implicitly be faster, as it avoids the array creation, insertion, and dereferencing that occurs with:

m.unpack("C")[0]

However, a benchmark to quantify the difference would be nice to have.

[shevegen (Robert A. Heiler)]

I think a benchmark would be great. There are some ruby hackers/devs that almost always add benchmarks to their changes, which I think is nice.

[znz (Kazuhiro NISHIYAMA)]

Using benchmark_driver gem, unpack1(template) is about 1.07x faster than unpack(template)[0].

% cat url_encode.yml
prelude: |
  require 'erb'
  str = "Programming Ruby:  The Pragmatic Programmer's Guide"
benchmark: ERB::Util.url_encode(str)
% benchmark_driver url_encode.yml -e ruby-base::$(rbenv root)/versions/git/bin/ruby -e ruby-unpack1::$(rbenv root)/versions/git2/bin/ruby
benchmark results:
Execution time (sec)
name             ruby-base ruby-unpack1
url_encode       1.793    1.665

Speedup ratio: compare with the result of `ruby-base' (greater is better)
name             ruby-unpack1
url_encode       1.077

[k0kubun (Takashi Kokubun)]

• Assignee changed from k0kubun (Takashi Kokubun) to znz (Kazuhiro NISHIYAMA)

Thank you for showing benchmark result.
While the impact is not so large with that benchmark, this would be expected usage of #unpack1 and it's good for maintainability too. So I think "not being slower in benchmark" is enough for adding this change.
I'm +1 for committing this.

• Status changed from Assigned to Closed

Applied in changeset trunk|r60061.

---

Use unpack1 instead of unpack and [0]

[Feature #13943][ruby-core:83027]