Memory footprint of StringZilla

[MalteEbner]

Do you have any benchmarks and comparisons of the memory footprint of Stringzilla?

As some examples:

• saving a string with 1Mio chars as stringzilla.Str vs a python str
• saving many strings, e.g. 1Mio strings with 50 chars each as
  • python list[str]
  • list[stringzilla.Str]
  • stringzilla.Strs

In the case of stringzilla.Strs: What is their behavior when reading from them with multiprocessing? Will they be copied on read or not?

[ashvardanian]

Valid concerns! In the first case you won’t feel the difference. In the second case, the behavior will vary depending on the platform, and may be suboptimal, depending on the C++ shared_ptr implementation. As for the last part - the strings won’t be copied.

[MalteEbner]

Thanks for the answer!