darkskyapp / string-hash Goto Github PK

I was using this module to get a hash and mod by 16 to get a random-ish but constant index for a given string. I was about to ensure the string was lowercase so that the hash was case insensitive, when I noticed that all my indexes for a string were the same no matter what the case was!

Here's an example on tonic to show this: https://tonicdev.com/rahat/string-hash-case-insensitivity

I noticed that this property holds true if you mod a hash by a power of 2 up to 6432 (though 64 does have a lot of matching indexes, it will rarely have a different index for a different case).

I don't know if this a bug per-se, but it is a very interesting property. Any clue as to why it behaves like this?

I suspect that (x >>> 0) is a quicker way to convert the hash result to an unsigned integer.

Ie, replace:

hash >= 0 ? hash : (hash & 0x7FFFFFFF) + 0x80000000

with:

hash >>> 0

Differences include iterating over the string backwards (as that is faster in JavaScript)

This requires some proof, because for long strings backward iteration may break any data caching/prefetching, therefore access to string characters may be significantly slower.

https://jsperf.com/string-hash-speed forward direction is faster in Chrome

https://gist.github.com/victor-homyakov/bcb7d7911e4a388b1c810f8c3ce17bcf shows me that in Node 8.6.0 forward iteration is twice as fast.

The readme has you importing a stringHash function but the code clearly calls the function hash. either refactor or adjust the readme.

source code:

const glob = require('glob')
const hash = require('string-hash')

const names = glob.sync('**')
console.log('Size: %s', names.length)

const used = {}
for (const item of names) {
  const value = hash(item)
  if (used[value]) {
    console.log('Found collision, hash: %s, a: %s, b: %s', value, used[value], item)
  }
  used[value] = item
}

The output is:

Size: 69791
Found collision, hash: 413020853, a: node_modules/@shimo/ui/icons/symbol/plus_circle.svg, b: node_modules/aliyun-sdk/apis/dm-2015-11-23.json

Thanks for creating this. I was curious, what is the chance of two hashes colliding within this algorithm. I was thinking of using it to hash emails.

so that your changes since 2012 will be available on npm =]

It is possible to unhash the resultet value?

I do some automated checks, which licenses my application is using. That is mainly based on the license field in package.json. If you put a CC0 there, it would be fine.

[user@likely html]$ sudo npm install string-hash -save
npm ERR! path /var/www/html/node_modules/npm/node_modules/dezalgo
npm ERR! code ENOENT
npm ERR! errno -2
npm ERR! syscall rename
npm ERR! enoent ENOENT: no such file or directory, rename '/var/www/html/node_modules/npm/node_modules/dezalgo' -> '/var/www/html/node_modules/npm/node_modules/.dezalgo.DELETE'
npm ERR! enoent This is related to npm not being able to find a file.
npm ERR! enoent

npm ERR! A complete log of this run can be found in:
npm ERR! /root/.npm/_logs/2017-12-10T05_45_35_903Z-debug.log