Consider removing the faker dependency

The problem

faker uses 7.6M of disk as of writing. For users who aren’t using its features, this is a pretty heavy cost.

It includes a generator for license plates, SSN, ISBN, etc…

Proposed solution

Consider removing the faker dependency, allowing users to plug it if they need it.

This solution would also allow users to control how the faker factory is used and specify its locale, for instance.

[charlax]

I would argue that removing faker is a strictly better design, applying the single responsibility principle. Factoryboy already has a lot of value without using faker at all. This would add a lot of flexibility in choosing how fake values are generated.

But sure, this would definitely be a breaking change. I think this is achievable using setuptools extras, and would only require users to change the package they install without changing any code.

On Thu 1 Aug 2019 at 11:57, Raphaël Barrois [e-mail] wrote:

I’ll try to sum up the current situation:

• Faker takes (currently) 11MB of disk space
• For comparison, an empty virtualenv is already 14MB in size (due to the copies of pip and setuptools there).

What is the actual issue here? factory_boy is intended as a development tool. It’s obviously better to reduce the size when possible, but are those 11MB a problem in development platforms?

If we want to reduce the space used by factory_boy, I see 2 options:

• Find a way to reduce faker’s disk space (for instance by changing the way its provider data is stored, and maybe compressing it; or by making part of the contents optional);
• Provide a way not to install faker alongside factory_boy.

Currently, factory_boy has a hard dependency on faker; making that optional would require every user of factory_boy+faker to change from factory_boy to factory_boy[faker] in their dependencies… which I’d rather avoid. Another option would be to publish two releases (factory_boy & factory_boy_minimal)?

All that is quite a lot of work for the project team and possibly for end users.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/FactoryBoy/factory_boy/issues/632?email_source=notifications&email_token=AAA5NNPG6XCYGBKG67I52J3QCKXSDA5CNFSM4H3FWEDKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD3KBOCA#issuecomment-517216008, or mute the thread https://github.com/notifications/unsubscribe-auth/AAA5NNMVUTIRMVBHW6UQ7PTQCKXSDANCNFSM4H3FWEDA .

[rbarrois]

I’ll try to sum up the current situation:

• Faker takes (currently) 11MB of disk space
• For comparison, an empty virtualenv is already 14MB in size (due to the copies of pip and setuptools there).

What is the actual issue here? factory_boy is intended as a development tool. It’s obviously better to reduce the size when possible, but are those 11MB a problem in development platforms?

If we want to reduce the space used by factory_boy, I see 2 options:

• Find a way to reduce faker’s disk space (for instance by changing the way its provider data is stored, and maybe compressing it; or by making part of the contents optional);
• Provide a way not to install faker alongside factory_boy.

Currently, factory_boy has a hard dependency on faker; making that optional would require every user of factory_boy+faker to change from factory_boy to factory_boy[faker] in their dependencies… which I’d rather avoid. Another option would be to publish two releases (factory_boy & factory_boy_minimal)?

All that is quite a lot of work for the project team and possibly for end users.