Bijection from string to int aka reversible hash

Question

I need to convert a set of strings similar to /azurite/spot00 to integers in order to use in ML libraries. Hand-rolling an enumerating algorithm (assign i++ to each next label) sounds easy enough. But nowhere nearly as elegant as a bidirectional hash between std::string and int (not sure if I need int64 or something else).

std::hash doesn't seem to state it's reversible. Anything in the standard library?

https://stackoverflow.com/questions/27694153/perfect-hash-function-for-strings-known-in-advance — President James K. Polk, May 31 '21 at 20:24

score 3 · Accepted Answer · answered May 31 '21 at 20:40

3

There's no general-purpose way to find a bijection from std::string to int for the simple but mundane reason that there are more possible std::strings than there are ints. (Specifically, there's effectively an unbounded number of possible std::strings, and there are only 2³² or 2⁶⁴ distinct possible integers).

There are ways to construct perfect hash functions from strings to integers if you have a fixed set of strings you want to work with, but in your case if the goal is just to label all the strings with distinct values your initial idea of just having a counter and assigning each string the next available number is probably just fine.

answered May 31 '21 at 20:40

templatetypedef

362,284
104
897
1,065

there is `std::string::max_size`, but yeah practically only bounded by available memory – 463035818_is_not_an_ai May 31 '21 at 20:44
@463035818_is_not_a_number Yep, hence "effectively" an unbounded number of strings :-) – templatetypedef May 31 '21 at 20:45
1

Or just limit your program to only supporting strings whose length is less than `sizeof(int)` :) – Jeremy Friesner May 31 '21 at 20:58
Now that I think of it, my alphabet is "lowercase latin letters and / and digits". So the size is 39. In 64 bits I think one can encode 12 characters from that alphabet. Which is almost enough. But not worth the hassle. – Vorac Jun 01 '21 at 08:25

Bijection from string to int aka reversible hash

1 Answers1