Wrapper for the OmNomNum c project, which implements most of the logic from numerizer.
gem install omnomnum
require 'omnomnum'
OmNomNum.normalize('first')
=> "1st"
OmNomNum.normalize('second')
=> "second"
OmNomNum.normalize('second', parse_second: true)
=> "2nd"
| Library | Bytes at exit | Total # of Allocations | Total Frees | Total # of Bytes Allocated | Savings |
|---|---|---|---|---|---|
| Numerizer1 | 2,746,272 | 59,243 | 35,161 | 11,055,638 | - |
| Numerizer2 | 3,719,364 | 10,852,981 | 10,821,658 | 1,115,915,077 | - |
| Numerizer3 | 973,092 | 10,793,738 | 10,786,497 | 1,104,859,439 | 1× |
| OmNomNum1 | 2,657,613 | 57,967 | 34,681 | 10,797,999 | - |
| OmNomNum2 | 2,659,473 | 62,006 | 38,700 | 10,839,010 | - |
| OmNomNum3 | 1,860 | 4,039 | 4,019 | 41,011 | 25,000+× |
1: loads the library and does nothing else
2: runs the string "two hundred" 2,000 times
3: difference between 1 and 2
Note: Benchmark done using valgrind.
valgrind ruby -w Ilib test/memory_test_*.rb
| Library | Iterations per second | Speedup |
|---|---|---|
| Numerizer | 323.271 | 1× |
| OmNomNum | 2.516M | 7,750+× |
The OmNomNum actually has two methods: one that returns a normalized string, and one that returns a list containing the locations of the numbers in the string and their values. ruby_omnomnum currently only supports the first function, but the second could eventually be exposed.