Persistent Isolated Development Environments

Use docker to create fully persistent and isolated development environments. Think virtualenv from the python world, only for a whole system.

Installation

If you have rust + cargo installed: cargo install pide-rs.

Binaries coming soon.

Dependencies

• docker (tested with v17.03)

Usage

First, create a dockerfile somewhere like ~/dockerfiles/debian-jessie-vim. The contents should look something like this:

FROM debian:jessie

RUN apt-get update

RUN apt-get install -y \
    vim

Next, move to the working directory of your project. Here I will use ~/demo. Initialize pide to your local working directory. This step may take some time, depending on how long it takes to build your dockerfile for the first time.

~ cd ~/demo
~ pide init ~/dockerfiles/debian-jessie-vim
Building dockerfile (if necessary)...
Initialized pidefile

Now you have a persistent working environment! jump into it by calling pide resume.

~ ls -a
.  ..  a.txt  b.txt  c.txt  .pide
~ pide resume
Running `debian-jessie-vim` for the first time...
root@ae6da64ee5e2:/# ls /host
a.txt  b.txt  c.txt
root@ae6da64ee5e2:/# echo Hello World > hello_file
root@ae6da64ee5e2:/# cat hello_file
Hello World
root@ae6da64ee5e2:/# exit
Committing container history...

Note that the location you call pide resume from is mapped as /host in the docker environment. This is dangerous and bad. Be careful with this.

In this example, we created a file called hello_file. If we resume again, it will still be there! This is also true for any system changes made, for example installing packages. Lets see what happens when we run it again:

~ pide resume
Resuming `debian-jessie-vim` where you left off...
root@9691f706c0e5:/# ls -l | grep hello_file
-rw-r--r--   1 root root    12 Mar 19 18:39 hello_file

Multiple Instances

If you re-use the same dockerfile in separate places, they will not interact with each other (though the init stage will run much faster, as the original cached docker image is used). Example:

~ mkdir demo2
~ cd demo2
~ pide init ~/dockerfiles/debian-jessie-vim
Building dockerfile (if necessary)...
Initialized pidefile
~ pide resume
Running `debian-jessie-vim` for the first time...
root@b64b1d8e24da:/# ls
bin  boot  dev  etc  home  host  lib  lib64  media  mnt  opt  proc  root  run  sbin  srv  sys  tmp  usr  var
root@b64b1d8e24da:/# exit
Committing container history...
~

Using multiple dockerfiles in one working directory is not currently supported.

TODOs, Notes, Other

First of all, this is a really hacky tool meant to support my workflow. It should NOT be used for production, and may misbehave in any number of unsafe or dangerous ways. Use at your own risk.

Known issues/problems

• This will create a lot of orphaned docker images. This generally wastes space on your machine. From docker v1.13+, you can run docker system prune to clean up a bit. This is a bigger problem on Mac/Windows, where docker runs within a virtual machine with fairly limited total space. In those environments, I suggest using vagrant instead of this tool (you're running a VM anyway)
• This tool maps your current working directory to docker. The root user can then do all sorts of bad things with UID 0, including making things executable, deleting the mapped directory, etc. Assume anything you do inside docker could have impact outside of docker. Use this only for dependency isolation, NOT security isolation
• Basically nothing is configurable yet. Eventually I would like to support more functionality from tools like docker-compose to map ports, volumes, networks, etc.

License

This tool is licensed under the MIT license.