blissfulness be with me.
Thursday, 25 August 2016
hsenv - Virtual Haskell Environment builder
What is it?
hsenv is a tool (inspired by Python's virtualenv) to create isolated Haskell environments.
What does it do?
It creates a sandboxed environment in a .hsenv/ subdirectory of your project, which, when activated, allows you to use regular Haskell tools (ghc, ghci, ghc-pkg, cabal) to manage your Haskell code and environment. It's possible to create an environment, that uses a different GHC version than your currently installed system GHC version. Very simple emacs integration mode is included.
First, choose a directory where you want to keep your sandboxed Haskell environment, usually a good choice is a directory containing your cabalized project (if you want to work on a few projects (perhaps an app and its dependent library), just choose any of them, it doesn't really matter). Enter that directory:
Next, create your new isolated Haskell environment (this is a one-time-only (per environment) step):
Now, every time you want to use this environment, you have to activate it:
That's it! Now it's possible to use all regular Haskell tools like usual, but it won't affect your global/system Haskell environment, and also your per-user environment (from ~/.cabal and ~/.ghc) will stay the same. All cabal-installed packages will be private to this environment, and the external environments (global and user) will not affect it (this environment will only inherit very basic packages, mostly ghc and Cabal and their deps).
When you're done working with this environment, enter command deactivate_hsenv, or just close the current shell (with exit).
Named vs Unnamed Environments
By default, hsenv creates an "unnamed" environment, but sometimes for particular use cases you might want to create different environments under the same directory. This is achievable creating a "named" environment. To do that, just pass the flag --name=<ENVIRONMENT_NAME> to hsenv:
This will make hsenv generate a folder of the form .hsenv_<ENVIRONMENT_NAME>.
If you want to customize activation and deactivation, create one or more of the following files in ~/.hsenv/bin/: pre-activate, post-activate, pre-deactivate, post-deactivate. These shell scripts will be sourced automatically by the main activation script.
Here's the most advanced usage of hsenv. Let's say you want to:
Hack on a json library
Do so comfortably
Use your own version of the parsec library
And do all this using the nightly version of GHC
First, download the binary distribution of GHC for your platform (e.g. ghc-7.3.20111105-i386-unknown-linux.tar.bz2).
Download a copy of json library and your private version of parsec:
darcs get http://patch-tag.com/r/Paczesiowa/parsec
cabal unpack json
Install the rest of the json deps:
cabal install --only-dependencies
Now, let's say you want to hack on Parsec module of json library. Open it in emacs:
Activate the virtual environment (hsenv must be required earlier):
M-x hsenv-activate <RET> /tmp/test/ <RET>
Edit some code and load it in ghci using 'C-c C-l'. If it type checks, you can play around with the code using nightly version of ghci running in your virtual environment. When you're happy with the code, exit emacs and install your edited json library:
And that's it.
Fetching and downloading a remote version of GHC
Recent versions of hsenv include the possibility to automatically download and install a GHC version directly from the GHC main repository. To do that, all you need to do is to pass the desired version of GHC you want to install:
Or a valid URL pointing to a compressed GHC archive:
hsenv has been tested on Linux, Mac OS X, and FreeBSD systems, but it should work on any POSIX platform. The external (from tarball) GHC feature requires a binary GHC distribution compiled for your platform which that can be extracted with tar and installed with "./configure --prefix=PATH; make install".
Q: Can I use it together with tools like cabal-dev or capri?
A: No. All these tools work more or less the same (wrapping cabal command, setting GHC_PACKAGE_PATH env variable), so something will probably break.
Q: Using GHC from tarball fails with a bunch of make tool gibberish on FreeBSD. What do I do?
A: Try the '--make-cmd=gmake' switch.
Q: Can I use hsenv inside hsenv?
A: No. It may be supported in future versions.
Q: Does it work on x64 systems?
Q: Will it work on Mac?
Q: Will it work on Windows?
A: I really doubt it would even compile. I don't have access to any Windows machines, so you're on your own, but patches/ideas/questions are welcome. Maybe it would work on Cygwin.
Q: Does it require Bash?
A: No, it should work with any POSIX-compliant shell. It's been tested with bash, bash --posix, dash, zsh and ksh.
Q: Can I use it with a different haskell package repository than hackage?
A: Yes, just adjust the url in .hsenv/cabal/config file.
Q: How do I remove the whole virtual environment?
A: If it's activated - 'deactivate_hsenv' it. Then, delete the .hsenv/ directory.
Q: Is every environment completely separate from other environments and the system environment?
A: Yes. The only (minor) exception is ghci history - there's only one per user history file. Also, if you alter your system's GHC, then virtual environments using system's GHC copy will probably break. Virtual environments using GHC from a tarball should continue to work.
Q: Can I share one cabalized project directory among multiple environments (e.g. build a cabalized project in the same directory using different environments)?
A: Yes. hsenv also overrides cabal with a wrapper, that will force using different build directories, so there shouldn't be even any recompilation between different environments.
Q: Is it possible to activate an environment upon entering its directory?
A: Yes, if you really know what you're doing. Here's a snippet for bash, which will activate both named and unnamed environments:
Now we need to get the latest hakyll. I decided to check out the hakyll4 branch because from what I can tell this version allows you to do painless syntax highlighting with pandoc. You could just use the regular master branch (hakyll 3.x) but syntax highlighting support on code blocks is a pain; plus, you’re going to have to migrate to hakyll 4 later anyway, so just use hakyll4.
The hakyll4 repo has several working sample blog sites. Now go into the just-cloned hakyll repo, and copy the contents of the data/example folder to your new blog’s directory.
cp ~/hakyll/data/example ~/myblog
Now go to your myblog folder and instantiate a hsenv sandbox, and then install hakyll into this sandbox.
The hsenv basically changes your shell’s path parameters and also installs everything into the .hsenv_myblog folder, thus giving your shell a different “vision”, so to speak, of what packages are installed. We use the sandbox environment inside hsenv whenever we want to generate our blog’s static generator binary (remember, hakyll is just a library; we need to write a static site generator program that makes use of it to generate the static html/css).
To deactivate the sandbox, invoke deactivate_hsenv or just close your shell. Just remember to call source .hsenv_myblog/bin/activate when you want to make changes to your blog or edit your blog’s binary. A sample creating/edition session would look something like this:
# edit some files, including site.hs, the static generator binary# compile site.hs./site rebuild
Your static site will be available at ~/myblog/_site.