Bootstrap numpy installation

[cgrimal]

Hi guys,

I have a private repository with C extensions. The setup.py file requires Numpy, so you need to have it installed on your system before building the package.
I run into a problem when using the dockerizePip option, because Numpy is not installed in the docker container.

I found 2 possible workarounds:

1. Change the setup.py of my project to bootstrap installation of Numpy (see this stackoverflow post).
   PROS: project is more straightforward to install, and nothing has to be changed to this plugin.
   CONS: Numpy is built at every deploy, and it is very long.
2. Add a bootstrapNumpy option to this plugin, that if true, will do a pip install numpy before running the install of the requirements.
   PROS: fast and flexible.
   CONS: Yet another option...

To be honest, the first option is not viable for me, because IMO building Numpy at every deploy is killing it. So I am using a dirty local version to implement the second option.
That being said, if someone run into the same issue, and if you think it might be useful to others, I can submit a PR.

Thanks!

[dschep]

I would prefer the third option which is to implement #114 (dockerfile build support), then you could just have:

Dockerfile:

FROM lambci/lambda:build-python3.6
RUN pip install numpy

sls.yml:

custom:
  dockerizePip: true
  dockerFile: Dockerfile

[cgrimal]

Agreed! This is a much more flexible and elegant solution.
Will try to look into it.