rscase 1.1.1

Installation

Install with pip using:

pip install rscase

Note: This package requires Rust nightly 2020-02-06 or an equivalent future release.

Usage

The package provides utility functions for generating strings formatted in several different case standards.

The case-standards and their functions are listed below.

All functions are imported and accessed the same way:

>> [in] from rscase import rscase
>> [in] rscase.camel_case('this_is-a_Test')
>> [out] thisIsATest

If you want to use this package, please note that the case functions are written to successfully convert camel case and snake case to the remaining formats. Formatting train case to itself doesn’t really make sense, and the way I would use this would be to, e.g., serialize out response data to a camelCased format.

Benchmarking Performance

This repo is a bit of an experiment, and because the functions contained in this package only do some very simple string manipulation, they seem like they might actually be good candidates for Python vs Rust performance benchmarking.

To try and make this a fair comparison - to make sure we’re comparing apples to apples - I decided to test the Rust function snake_case (see the Rust function here) to an identical Python function. The Python version is shown below:

from rscase import rscase

test_string = "thisIsALongCamelCasedAlphabeticKey"

# Test functions

def original_snake_case():
    string = test_string
    new_string = ""
    dash = "-"
    for index in range(len(string)):
        if index == 0:
            new_string += string[index].lower()
        elif string[index] == dash:
            new_string += '_'
        elif string[index].upper() == string[index]:
            new_string += f'_{string[index]}'
        else:
            new_string += string[index]
    return new_string

def rust_snake_case():
    string = test_string
    return rscase.snake_case(string)

The main difference between the two functions, flow-wise, is only that Rust won’t let you just iterate over a string, so you have to create a vector of char’s instead - or at least that’s how I did it.

Results

After running the tests, the results seems to be pretty promising - in favor of the Rust implementation.

The results are pretty clear: after only 100 reps, the results seem to stabilize, and flatten out at around a 2.3x longer execution time for the Python implementation.

* the 1-rep result seems to show that Python actually outperforms Rust in the scenario that would normally actually matter. Since it makes sense that variance would be high when trying to measure something at the microsecond level I decided to run this individual scenario again, another one million times, to increase the sample size. With a larger sample, the average difference for 1 rep averages to 1.85x slower in Python, and the median is 1.88x. In short, the Rust implementation seems to outperform the Python across the board.

Benchmarking Performance - Update

Thanks to Thomas Hartmann for suggesting a significant performance improvement in the packaged Rust code.

Using some experimental features, we’re able to improve the performance of the Rust code considerably. The snake_case test from above is replicated below, with the performance difference settling at 5x the Python performance.

Running the 1 rep scenario one million times, gives an average Rust execution time of 3.84 us compared to an average Python execution time of 12.61 us (~3.3x slower for Python).

This time around, I also decided to test the camel case implementations, as the logic does behave a bit differently:

Running the 1 rep scenario one million times, gives an average Rust execution time of 3.90 us compared to an average Python execution time of 11.48 us (almost ~3x slower for Python).

In summary, the benchmarked performed similarly, with Rust pulling ahead even more, for these two implementations. At the same time, there’s probably room for improvement for both implementations still, and probably especially for the Python one.