Agile/fix/random string generator producing duplicates #86
Conversation
Update the `RandomStringGenerator` class to use a static `Random` variable to ensure that when multiple instances of this class are created, they do not all use the same initial seed (based on system time). By using a static instance, the random generator is shared and will always use a single seed. This issue has been observed on a project that uses this library, but was not reproduceable as a test in this project.
Update the CHANGELOG file to include a reference to the random string generator fix.
|
Fix looks good. The time seed will be per program instance rather than per object instance, and no term in the sequence will be shared by multiple objects. So you should get random behaviour across multiple program instances and object instances within a program. The only potential issue I can think of is multi-threading. What if two objects try to access the static random variable at the same time, and somehow both end up getting the same term in the sequence? My guess is that C# has that covered by only allowing one thread concurrent access to the object at a time. So I wouldn't worry about this. Just a thought. Just a note on the issue. It must have only been happening when multiple random objects were created in such rapid succession that there was no measurable time difference between them. If there's any sort of pause, the random sequences will get a different seed: https://dotnetfiddle.net/L8SSuJ But if there's no pause, they'll probably get the same seed: https://dotnetfiddle.net/TwlZmc Explicitly seeding the date-time doesn't help:
https://dotnetfiddle.net/dB4Joy So yeah, the best solution I can think of is the one you implemented. |
Update how the code creats instances of the `Random` class to ensure that the code is thread-safe. The new code uses a static global `Random` instance that generates seeds for each thread-specific instance.
|
Very good point on the thread-safety of this solution. I hadn't thought of that. I did a bit of investigation and I updated the PR with an updated implementation that is thread-safe. It works by creating a global static instance that is used to generate seeds for thread-specific instances of the |
|
Interesting. Clever use of a global random variable for locking & seeding. |
hctrdev
deleted the
agile/fix/random-string-generator-producing-duplicates
branch
Description
Update the
RandomStringGeneratorclass to use a staticRandomvariable to ensure that when multiple instances of this class are created, they do not all use the same initial seed (based on system time). By using a static instance, the random generator is shared and will always use a single seed.This issue has been observed on a project that uses this library, but was not reproduceable as a test in this project.
TODO