Produce a sample that does not match a regex

[spacether]

Thank you so much for this package!
It looks great and I am looking forward to seeing if I can use it to generate regex samples for api documentation and test generation.

Is your feature request related to a problem? Please describe.
One can generate strings that do match a regex pattern but can it generate a string that does not match a pattern?
Your code lets us test a positive use case. If you provide a sample_does_not_match_pattern functionality then we can test the negative use case too.

Describe the solution you'd like
Could a method or function be created to produce a sample that does not match a regex pattern?

Describe alternatives you've considered
Randomly generating strings until a non matching sample is found.
But that is non-deterministic and is brute forcing the problem.

Additional context
I am considering using this library in https://github.com/OpenAPITools/openapi-generator

[curious-odd-man]

Hello @spacether .

Thank you for your ticket.
Let me describe what solution I may be able to provide and please let me know if that solution will satisfy you.

Solution:

Suppose we have a regex: abc[def] - a simple one.

I would then separate this regex into 2 parts: asd and [def].

To produce string that does not match the initial pattern I will;

1. Select 1 to N parts (N is total number of parts - 2 in sample case)
2. Invert each selected part. For example:
   1. For abc part I will produce random 3 character long string, that is not abc
   2. For [def] part I will change to [^def]
3. Generate a regex mixing "normal" parts and "inverted" parts.

This approach should produce strings that will partially match the regex.
I will also allow user to specify fraction of parts that should be inverted.
For example if user specifies 1 - then all parts will be inverted. If user will specify 0 - then only 1 part will be inverted, and if user will specify 0.5 - then half of the parts will be inverted.

I have not yet deep dived into this approach, so there might be some underwater stones, but generally it looks promising for me.

Is that approach clear for you? Do you think that will suite your needs?

[spacether]

Yup, I think that that approach will suit my needs.
For now your library generating samples that match is the most important use case for us.

These not matching negative use cases are a nice to have for the future.
The most important need that we would have both for the matching and non matching samples is that wen we get the first sample that does or doesn't match, that it is constant.

If our pattern is:
^def
and we get:
first_match = def
first_non_match = fed
as long as those samples will always be those values that works for us.
We need this determinism because the samples are used to generate documentation and tests.
We don't want values changing with every sample generation invocation.

Will your solution always give the same result for the first non matching sample?
The random 3 character long string does not sound deterministic.

[curious-odd-man]

In current implementation you can provide initialized Random to get same generated random values: public String generate(Random random);
In not matching generation I will use same approach - if identical Random will be provided - identical results will be returned.

You can refer to this test case.

Random 3 character long string will be deterministic because randomness of this string will be defined by some of the pseudo-random values generators, e.g. Random, which is deterministic. Refer to this post for explanation.

[spacether]

That works great; thank you!

[curious-odd-man]

Hello. I've completed implementation for the feature requested.
Please refer to README for usage.
It is available as a snapshot release.

If something is not working or any additional features are needed - please open another ticket.

I did not implement any customization for now - from your description i understand it is not very important and this will require big overhead also.

[spacether]

Thank you! Openapi generator is now using your package to generate matching regex examples for the python-experimental generator OpenAPITools/openapi-generator#7157

[curious-odd-man]

That is great! Thank you. It's always nice to know that someone uses your work 👍