add close method

[dominictarr]

I found this sitting uncommited in my dev directory.

I think it's needed because I added close to the benchmark, and the log tests?

[arj03]

This seems to make append a bit slower but make random a lot faster. Are you seeing the same @dominictarr?

Before:

append, 60189.786, 8.994, 602259, 89.996, 10.006
stream, 43594.04, 6.514, 435984, 65.15, 10.001
stream no cache, 50536.246, 7.551, 505413, 75.525, 10.001
stream10, 43835.232, 6.55, 438440, 65.517, 10.002
random, 27497.85, 4.109, 275006, 41.094, 10.001

After:

append, 57312.068, 8.561, 573178, 85.624, 10.001
stream, 44599.44, 6.662, 446039, 66.632, 10.001
stream no cache, 52688.831, 7.87, 526941, 78.717, 10.001
stream10, 44091.79, 6.586, 440962, 65.873, 10.001
random, 34874.812, 5.209, 348783, 52.103, 10.001

Also tested 16 that seems to be almost the same as 8.

[dominictarr]

oops, I didn't mean to change the cache size. I'm gonna update PR to just not change that.

[ahdinosaur]

is this meant to do anything or just provide a consistent api? if it's to make this module consistent with some interface, maybe add a comment saying "nothing to do when closing". anyway, seems innocent, but then there's other changes in this pull request that seem unrelated.

[dominictarr]

merged master into this and removed change to cache:

master:

name, ops/second, mb/second, ops, total-mb, seconds
append, 180069.923, 22.782, 1802680, 228.076, 10.011
stream, 123349.33, 15.606, 1233740, 156.094, 10.002
stream no cache, 147520.547, 18.664, 1475353, 186.663, 10.001
stream10, 151553.446, 19.174, 1517050, 191.934, 10.01
random, 22700.818, 2.872, 227553, 28.79, 10.024

close:

name, ops/second, mb/second, ops, total-mb, seconds
append, 181033.09, 22.904, 1810874, 229.112, 10.003
stream, 120267.692, 15.216, 1203158, 152.224, 10.004
stream no cache, 143455.754, 18.15, 1434701, 181.518, 10.001
stream10, 147191.646, 18.623, 1473094, 186.382, 10.008
random, 24554.644, 3.106, 245571, 31.069, 10.001

[dominictarr]

not a significant difference

[christianbundy]

I think we removed this in master on purpose but I can't remember why. Hmm.

[dominictarr]

I remember! it breaks your delete stuff. unfortunately there is no option that means "open a file for writing and reading and create the file if it does not exist" there is only error if it does not exist

[christianbundy]

Does that mean that we should remove it from here? I don't remember, I can look into it (unless someone else beats me to it).

[dominictarr]

yes, this patch should not change the file open flag, it's not about that.