DOC: add section on setting random number seeds #9123
Conversation
|
Minor remark FWIW, shouldn´t we promote fixing the PRNG seed in examples as well? I think that @NelleV (or maybe @choldgraf, I don´t remember clearly 🐑) did some work a while ago to ensure that 19680801 was used as the seed for the PRNGs in every files where a call to a random function was made (not only the test-related ones). Do we want to make this clear in the docs too? |
|
I think the samples are uniformly change but most of the tests are not (because it would require re-generating test images). I decided against mentioning the examples here as this is a section on tests. |
doc/devel/testing.rst
Outdated
|
|
||
| Random data can is a very convenient way to generate data for examples, | ||
| however the randomness is problematic for testing (as the tests | ||
| must be deterministic!). To work around this set the seed in the test. |
There was a problem hiding this comment.
"in the tests ?": typo or correct English?
There was a problem hiding this comment.
tweaked a bit. The seed needs to be set in each test, just just once.
|
Fair enough about restricting the mention to the tests in a doc section about testing. |
ad5450d to
3833ff0
Compare
Document using 19680801 as the preferred random seed.