new opam features: local switches

---

Among the areas we wanted to improve on for opam 2.0 was the handling of switches. In opam 1.2, they are simply accessed by a name (the OCaml version by default), and are always stored into ~/.opam/<name>. This is fine, but can get a bit cumbersome when many switches are in presence, as there is no way to sort them or associate them with a given project.

A reminder about switches

For those unfamiliar with it, switches, in opam, are independent prefixes with their own compiler and set of installed packages. The opam switch command allows to create and remove switches, as well as select the currently active one, where operations like opam install will operate.

Their uses include easily juggling between versions of OCaml, or of a library, having incompatible packages installed separately but at the same time, running tests without damaging your "main" environment, and, quite often, separation of environment for working on different projects.

You can also select a specific switch for a single command, with

opam install foo --switch other

or even for a single shell session, with

eval $(opam env --switch other)

What opam 2.0 adds to this is the possibility to create so-called local switches, stored below a directory of your choice. This gets users back in control of how switches are organised, and wiping the directory is a safe way to get rid of the switch.

Using within projects

This is the main intended use: the user can define a switch within the source of a project, for use specifically in that project. One nice side-effect to help with this is that, if a "local switch" is detected in the current directory or a parent, opam will select it automatically. Just don't forget to run eval $(opam env) to make the environment up-to-date before running make.

Interface

The interface simply overloads the switch-name arguments, wherever they were present, allowing directory names instead. So for example:

cd ~/src/project
opam switch create ./

will create a local switch in the directory ~/src/project. Then, it is for example equivalent to run opam list from that directory, or opam list --switch=~/src/project from anywhere.

Note that you can bypass the automatic local-switch selection if needed by using the --switch argument, by defining the variable OPAMSWITCH or by using eval $(opam env --switch <name>)

Implementation

In practice, the switch contents are placed in a _opam/ subdirectory. So if you create the switch ~/src/project, you can browse its contents at ~/src/project/_opam. This is the direct prefix for the switch, so e.g. binaries can be found directly at _opam/bin/: easier than searching the opam root! The opam metadata is placed below that directory, in a .opam-switch/ subdirectory.

Local switches still share the opam root, and in particular depend on the repositories defined and cached there. It is now possible, however, to select different repositories for different switches, but that is a subject for another post.

Finally, note that removing that _opam directory is handled transparently by opam, and that if you want to share a local switch between projects, symlinking the _opam directory is allowed.

Current status

This feature has been present in our dev builds for a while, and you can already use it in the current beta.

Limitations and future extensions

It is not, at the moment, possible to move a local switch directory around, mainly due to issues related to relocating the OCaml compiler.

Creating a new switch still implies to recompile all the packages, and even the compiler itself (unless you rely on a system installation). The projected solution is to add a build cache, avoiding the need to recompile the same package with the same dependencies. This should actually be possible with the current opam 2.0 code, by leveraging the new hooks that are made available. Note that relocation of OCaml is also an issue for this, though.

Editing tools like ocp-indent or merlin can also become an annoyance with the multiplication of switches, because they are not automatically found if not installed in the current switch. But the user-setup plugin (run opam user-setup install) already handles this well, and will access ocp-indent or tuareg from their initial switch, if not found in the current one. You will still need to install tools that are tightly bound to a compiler version, like merlin and ocp-index, in the switches where you need them, though.

NOTE: this article is cross-posted on opam.ocaml.org and ocamlpro.com.

Comments

Jeremie Dimino (11 May 2017 at 8 h 27 min):

Thanks, that seems like a useful feature. Regarding relocation of the compiler, shouldn’t it be enough to set the environment variable OCAMLLIB? AFAIK the stdlib directory is the only hardcoded path on the compiler.

Louis Gesbert (11 May 2017 at 8 h 56 min):

Last I checked, there were a few more problematic points, in particular generated bytecode executables statically referring to their interpreter; but yes, in any case, it’s worth experimenting in that direction using the new hooks, to see how it works in practice.

Jeremie Dimino (12 May 2017 at 9 h 13 min):

Indeed, I remember that we had a similar problem in the initial setup to test the public release of Jane Street packages: we were using long paths for the opam roots and the generated #! where too long for the OS… What I did back then is write a program that scanned the tree and rewrote the #! to use “#!/usr/bin/env ocamlrun”.

That could be an option here. The rewriting only need to be done once, since the compiler uses ocamlc -where/camlheader when generating a bytecode executable.

---

About OCamlPro:

OCamlPro is a R&D lab founded in 2011, with the mission to help industrial users benefit from experts with a state-of-the-art knowledge of programming languages theory and practice.

• We provide audit, support, custom developer tools and training for both the most modern languages, such as Rust, Wasm and OCaml, and for legacy languages, such as COBOL or even home-made domain-specific languages;
• We design, create and implement software with great added-value for our clients. High complexity is not a problem for our PhD-level experts. For example, we helped the French Income Tax Administration re-adapt and improve their internally kept M language, we designed a DSL to model and express revenue streams in the Cinema Industry, codename Niagara, and we also developed the prototype of the Tezos proof-of-stake blockchain from 2014 to 2018.
• We have a long history of creating open-source projects, such as the Opam package manager, the LearnOCaml web platform, and contributing to other ones, such as the Flambda optimizing compiler, or the GnuCOBOL compiler.
• We are also experts of Formal Methods, developing tools such as our SMT Solver Alt-Ergo (check our Alt-Ergo Users' Club) and using them to prove safety or security properties of programs.

Please reach out, we'll be delighted to discuss your challenges: contact@ocamlpro.com or book a quick discussion.