reseed once per fixture set

[bendavies]

Defaulting the seed to 1 causes Faker randomisation to not really work as expected.
Defaulting the seed to null is probably a better idea?

EDIT: I've just seen this is expected, to produce consistent data across runs.

Anyway, I'd like at least to be able to set seed to null in the LoadFilesCommand
What do you think?

[h4cc]

Having the seed 1 is expected, to produce consistent results.

What would seeding with null result in?
Do you want to "not seed" and use the default random behavior?

[bendavies]

i noticed about 'produce consistent results' intention, sorry.
yeah, the ability to "random seed" (null) in the command would be good i guess

[bendavies]

Ok, actually, I've found the actually cause of my issue.

If i load 2 files with the files command that look like this.

Cat:
  cat{1..100}:
    name: <firstName()>

Dog:
  dog{1..100}:
    name: <firstName()>

This will produce 100 cats and 100 dogs with the same names, because mt_srand is re-seeded (to the same seed) per file,

I'll edit this PR to fix that.

[h4cc]

Changing the default seed to null is not possible, because others are already depending on that.

Having 'seed' => null in your FixtureSets can be done with the effect you want, Faker does use null as it should: https://github.com/fzaninotto/Faker/blob/bcc07d0cb56cf5ec4fb6e66dda5f5fb9e39e5c27/src/Faker/Generator.php#L182

Your problems seems to be only LoadFilesCommand, where a numeric seed parameter is needed ?

[bendavies]

Ok, so the issue was that the reseeding was occurring per file.
I've changed it to per set.

This will keep the consistent results per set.

What do you think?

[h4cc]

It was intended to have a seeding before each file. This way a single file will bring consistent results.

I cant see your remaining problem anymore. What are you trying to achieve?