avoid double check !isBuffer

[hex7c0]

Hi,
if you want to use var buf, are you free to close this PR :)

[dougwilson]

So the main reason it was var buf was that under certain conditions, writing to the variable that is mapped to an argument can kill the function's optimizations. I'm not 100% if that's the case here, but just giving some history of why it didn't just overwrite entity before (because, well, I didn't want to have to figure out if it was a penalty case or not :) )

[hex7c0]

gotcha 👍

tl;dr

commit, md4

> node benchmark/index.js

> node benchmark/body0-100b.js

  100B body

  4 tests completed.

  buffer - strong x   307,236 ops/sec ±1.89% (188 runs sampled)
  buffer - weak   x 1,146,018 ops/sec ±0.55% (196 runs sampled)
  string - strong x   178,177 ops/sec ±1.78% (188 runs sampled)
  string - weak   x   341,098 ops/sec ±0.61% (195 runs sampled)

> node benchmark/body1-1kb.js

  1KB body

  4 tests completed.

  buffer - strong x 222,709 ops/sec ±1.72% (191 runs sampled)
  buffer - weak   x 294,112 ops/sec ±0.77% (196 runs sampled)
  string - strong x 134,743 ops/sec ±2.50% (187 runs sampled)
  string - weak   x 159,173 ops/sec ±1.78% (188 runs sampled)

> node benchmark/body2-5kb.js

  5KB body

  4 tests completed.

  buffer - strong x 94,987 ops/sec ±1.19% (192 runs sampled)
  buffer - weak   x 96,372 ops/sec ±1.12% (195 runs sampled)
  string - strong x 48,524 ops/sec ±6.49% (172 runs sampled)
  string - weak   x 49,302 ops/sec ±6.64% (173 runs sampled)

> node benchmark/body3-10kb.js

  10KB body

  4 tests completed.

  buffer - strong x 55,977 ops/sec ±0.77% (195 runs sampled)
  buffer - weak   x 56,101 ops/sec ±0.93% (194 runs sampled)
  string - strong x 34,054 ops/sec ±5.53% (179 runs sampled)
  string - weak   x 34,195 ops/sec ±5.61% (180 runs sampled)

> node benchmark/body4-100kb.js

  100KB body

  4 tests completed.

  buffer - strong x 6,617 ops/sec ±0.60% (197 runs sampled)
  buffer - weak   x 6,630 ops/sec ±0.52% (196 runs sampled)
  string - strong x 4,862 ops/sec ±3.07% (183 runs sampled)
  string - weak   x 4,855 ops/sec ±3.11% (182 runs sampled)

orig, md5

> node benchmark/index.js

> node benchmark/body0-100b.js

  100B body

  4 tests completed.

  buffer - strong x   303,749 ops/sec ±1.88% (185 runs sampled)
  buffer - weak   x 1,133,652 ops/sec ±0.51% (196 runs sampled)
  string - strong x   174,857 ops/sec ±1.71% (190 runs sampled)
  string - weak   x   341,315 ops/sec ±1.47% (192 runs sampled)

> node benchmark/body1-1kb.js

  1KB body

  4 tests completed.

  buffer - strong x 205,874 ops/sec ±1.63% (191 runs sampled)
  buffer - weak   x 294,107 ops/sec ±0.66% (196 runs sampled)
  string - strong x 127,550 ops/sec ±2.64% (187 runs sampled)
  string - weak   x 159,006 ops/sec ±1.72% (188 runs sampled)

> node benchmark/body2-5kb.js

  5KB body

  4 tests completed.

  buffer - strong x 86,839 ops/sec ±1.14% (192 runs sampled)
  buffer - weak   x 87,445 ops/sec ±1.19% (193 runs sampled)
  string - strong x 46,031 ops/sec ±6.19% (175 runs sampled)
  string - weak   x 45,623 ops/sec ±6.65% (175 runs sampled)

> node benchmark/body3-10kb.js

  10KB body

  4 tests completed.

  buffer - strong x 50,551 ops/sec ±0.75% (195 runs sampled)
  buffer - weak   x 50,261 ops/sec ±1.08% (195 runs sampled)
  string - strong x 32,204 ops/sec ±5.24% (179 runs sampled)
  string - weak   x 32,043 ops/sec ±5.40% (180 runs sampled)

> node benchmark/body4-100kb.js

  100KB body

  4 tests completed.

  buffer - strong x 5,914 ops/sec ±0.65% (196 runs sampled)
  buffer - weak   x 5,904 ops/sec ±0.58% (196 runs sampled)
  string - strong x 4,478 ops/sec ±2.93% (186 runs sampled)
  string - weak   x 4,495 ops/sec ±3.03% (184 runs sampled)

[dougwilson]

I have to think about changing MD5 to MD4 for strong ETags, though, but I can certainly change the large content weak ETags to use MD4 right away, since collisions don't particularly matter.