| 1 | #!/usr/bin/env bash
|
| 2 | #
|
| 3 | # Test awk vs Python speed.
|
| 4 | #
|
| 5 | # On this hash table benchmark, Python is maybe 10% slower than gawk. mawk is
|
| 6 | # twice is fast as gawk (and bwk).
|
| 7 | #
|
| 8 | # Python has much more functionality, so it's not exactly a fair comparison,
|
| 9 | # but it's instructive.
|
| 10 | #
|
| 11 | # Update: simply adding tolower() makes gawk much slower than Python (555 ms
|
| 12 | # vs. 280 ms), and mawk is still much faster at 138 ms.
|
| 13 | #
|
| 14 | # Mawk is known to be fast? Faster than Java on this benchmark.
|
| 15 | # https://brenocon.com/blog/2009/09/dont-mawk-awk-the-fastest-and-most-elegant-big-data-munging-language/
|
| 16 | #
|
| 17 | # Usage:
|
| 18 | # ./awk-python.sh <function name>
|
| 19 |
|
| 20 | set -o nounset
|
| 21 | set -o pipefail
|
| 22 | set -o errexit
|
| 23 |
|
| 24 | readonly FILES=(../*.sh ../*/*.sh ../*.py ../*/*.py ../*/*/*.py)
|
| 25 |
|
| 26 | # Test out hash table implementations
|
| 27 | # mawk is faster: 77ms vs 155ms for 10 iterations.
|
| 28 | test-awk() {
|
| 29 | for awk in gawk mawk ~/git/bwk/bwk; do
|
| 30 | echo ---
|
| 31 | echo $awk
|
| 32 | echo ---
|
| 33 | time for i in {1..10}; do
|
| 34 | $awk '
|
| 35 | {
|
| 36 | line = tolower($0)
|
| 37 | num_lines += 1
|
| 38 |
|
| 39 | # NOTE: gawk has length(); mawk does not
|
| 40 | if (!(line in unique)) {
|
| 41 | num_unique += 1
|
| 42 | }
|
| 43 | unique[line] += 1
|
| 44 | }
|
| 45 | END {
|
| 46 | print "unique lines: " num_unique
|
| 47 | print "total lines: " num_lines
|
| 48 | }
|
| 49 | ' "${FILES[@]}"
|
| 50 |
|
| 51 | done
|
| 52 | done
|
| 53 | }
|
| 54 |
|
| 55 | # Python VM is slower: 160-170 ms. Oops.
|
| 56 | #
|
| 57 | # Well Python has more general dictionaries -- they take more than strings.
|
| 58 | test-python() {
|
| 59 | time for i in {1..10}; do
|
| 60 | python -S -c '
|
| 61 | import collections
|
| 62 | import sys
|
| 63 |
|
| 64 | num_lines = 0
|
| 65 | num_unique = 0
|
| 66 | unique = collections.defaultdict(int)
|
| 67 |
|
| 68 | for path in sys.argv[1:]:
|
| 69 | with open(path) as f:
|
| 70 | for line in f:
|
| 71 | line = line.lower()
|
| 72 | num_lines += 1
|
| 73 |
|
| 74 | if line not in unique:
|
| 75 | num_unique += 1
|
| 76 | unique[line] += 1
|
| 77 |
|
| 78 | print "unique lines: ", num_unique
|
| 79 | print "total lines: ", num_lines
|
| 80 | ' "${FILES[@]}"
|
| 81 |
|
| 82 | done
|
| 83 | }
|
| 84 |
|
| 85 | # Only 10-30 ms. We are doing real work.
|
| 86 | test-wc() {
|
| 87 | time for i in {1..10}; do
|
| 88 | cat "${FILES[@]}" | wc -c
|
| 89 | done
|
| 90 | }
|
| 91 |
|
| 92 | files() {
|
| 93 | echo "${FILES[@]}"
|
| 94 | echo "${#FILES[@]} files"
|
| 95 | }
|
| 96 |
|
| 97 | "$@"
|