* Use the tempfile crate instead of the tempdir crate (which is deprecated)
https://github.com/rust-lang-deprecated/tempdir?tab=readme-ov-file#deprecation-note
* perf: Add benchmark that measures the rejection speed of a large non-zip file
* perf: Speed up non-zip rejection by increasing END_WINDOW_SIZE
I tested several END_WINDOW_SIZEs across 2 machines:
Machine 1: macOS 15.0.1, aarch64 (apfs /tmp)
512: test parse_large_non_zip ... bench: 30,450,608 ns/iter (+/- 673,910)
4096: test parse_large_non_zip ... bench: 7,741,366 ns/iter (+/- 521,101)
8192: test parse_large_non_zip ... bench: 5,807,443 ns/iter (+/- 546,227)
16384: test parse_large_non_zip ... bench: 4,794,314 ns/iter (+/- 419,114)
32768: test parse_large_non_zip ... bench: 4,262,897 ns/iter (+/- 397,582)
65536: test parse_large_non_zip ... bench: 4,060,847 ns/iter (+/- 280,964)
Machine 2: Debian testing, x86_64 (tmpfs /tmp)
512: test parse_large_non_zip ... bench: 65,132,581 ns/iter (+/- 7,429,976)
4096: test parse_large_non_zip ... bench: 14,109,503 ns/iter (+/- 2,892,086)
8192: test parse_large_non_zip ... bench: 9,942,500 ns/iter (+/- 1,886,063)
16384: test parse_large_non_zip ... bench: 8,205,851 ns/iter (+/- 2,902,041)
32768: test parse_large_non_zip ... bench: 7,012,011 ns/iter (+/- 2,222,879)
65536: test parse_large_non_zip ... bench: 6,577,275 ns/iter (+/- 881,546)
In both cases END_WINDOW_SIZE=8192 performed about 6x better than 512 and >8192
didn't make much of a difference on top of that.
* perf: Speed up non-zip rejection by limiting search for EOCDR.
I benchmarked several search sizes across 2 machines
(these benches are using an 8192 END_WINDOW_SIZE):
Machine 1: macOS 15.0.1, aarch64 (apfs /tmp)
whole file: test parse_large_non_zip ... bench: 5,773,801 ns/iter (+/- 411,277)
last 128k: test parse_large_non_zip ... bench: 54,402 ns/iter (+/- 4,126)
last 66,000: test parse_large_non_zip ... bench: 36,152 ns/iter (+/- 4,293)
Machine 2: Debian testing, x86_64 (tmpfs /tmp)
whole file: test parse_large_non_zip ... bench: 9,942,306 ns/iter (+/- 1,963,522)
last 128k: test parse_large_non_zip ... bench: 73,604 ns/iter (+/- 16,662)
last 66,000: test parse_large_non_zip ... bench: 41,349 ns/iter (+/- 16,812)
As you might expect these significantly increase the rejection speed for
large non-zip files.
66,000 was the number previously used by zip-rs. It was changed to zero in
7a55945743.
128K is what Info-Zip uses[1]. This seems like a reasonable (non-zero)
choice for compatibility reasons.
[1] Info-zip is extremely old and doesn't not have an official git repo to
link to. However, an unofficial fork can be found here:
bb0c4755d4/zipfile.c (L4073)
---------
Co-authored-by: Chris Hennick <4961925+Pr0methean@users.noreply.github.com>
- use blocks for reading individual file headers
- remove unnecessary option wrapping for stream entries
- create Block trait
- add coerce method to reduce some boilerplate
- add serialize method to reduce more boilerplate
- use to_le! and from_le!
- add test case
- add some docs
- rename a few structs to clarify zip32-only
