-
Notifications
You must be signed in to change notification settings - Fork 9
2014 Blog Post Draft
The OCaml Labs initiative within the Cambridge Computer Laboratory is now just over two years old, and it is time for an update about our activities since the last update at the [end of 2013](http://www.cl.cam.ac.uk/projects/ocamllabs/news/index.html#Dec 2013) and 2012.
The theme of our group was not to be pure research, but rather a hybrid group that takes on some of the load of day-to-day OCaml maintenance from INRIA, as well as help grow the wider community and meet our own research agendas around topics such as unikernels. To this end, all of our projects have been highly collaborative, often involving colleagues from OCamlPro, INRIA, Jane Street, Lexifi and Citrix.
This post covers our progress in tooling, the compiler and language, community efforts, research projects and concludes with our priorities for 2015.
At the start of 2014, we had just helped to release OPAM 1.1.1 with our colleagues at OCamlPro, and serious OCaml users had just started moving over to using it.
Our overall goal at OCaml Labs is to deliver a modular set of of development tools around OCaml that we dub the OCaml Platform. The remainder of 2014 was thus spent polishing this nascent OPAM release into a solid base (both as a command-line tool and as a library) that we could use as the basis for documentation, testing and build infrastructure, all the while making sure that bigger OCaml projects continued to migrate over to it. Things have been busy; here are the highlights of this effort.
The central OPAM repository that contains the package descriptions has grown tremendously in 2014, with over 280 contributors committing almost 10000 changesets across 3800 pull requests on GitHub. The front line of incoming testing has been continuous integration by the wonderful Travis CI, who also granted us access to their experimental MacOS X build pool. The OPAM package team also to expanded to give David Sheets, Jeremy Yallop, Peter Zotov and Damien Doligez commit rights, and they have all been busily triaging new packages as they come in.
Several large projects such as Xapi, Ocsigen and our own MirageOS switched over to using OPAM for day-to-day development, as well as prolific individual developers such as Daniel Buenzli and Markus Mottl. Jane Street continued to send regular monthly updates of their Core/Async suite, and releases appeared from the Facebook open-source team as well (who develop Hack, Flow and Pfff in OCaml).
We used feedback from the users to smooth away many of the rough edges, with:
- a redesigned development workflow that lets developers quickly grab a development version of a library recompile all dependent packages automatically, and quickly publish results to GitHub.
- binary distributions for common OS distributions via their native packaging, as well as 0install and Vagrant boxes.
- a unified way of cloning the source of any package via
opam source. This handles any supported OPAM archive, including Git, Mercurial or Darcs remotes. - a richer package metadata, including source code, development archives and bug report URLs.
These changes were all incorporated into the OPAM 1.2, along with backwards compatibility shims to keep the old 1.1 metadata format working until the migration is complete. The 1.2.x series has been a solid and usable development manager, and last week's release of OPAM 1.2.1 has further polished the core scripting engine.
One of the more notable developments during 2014 was the adoption of OPAM further up the ecosystem by the Coq theorem prover. This broadening of the community prompted us to create an official OPAM blog to give us a central place for new and tips, and we've had posts about XenServer developments, the Merlin IDE tool and the modern UTop interactive REPL. If you are using OPAM in an interesting or production capacity, please do get in touch so that we can work with you to write about it for the wider community.
The goal of the blog is also to start bringing together the various components that form the OCaml Platform. These are designed to be modular tools (so that you can pick and choose which ones are necessary for your particular use of OCaml). There are more details available from the OCaml Workshop presentation at ICFP 2014 (abstract, slides, video).
OPAM has also been adopted now by several big universities (including us at Cambridge!) for undergraduate and graduate Computer Science courses. The demands increased for an out-of-the-box solution that makes it as easy possible for new users to get started with minimum hassle. We created a dedicated teaching list to aid collaboration, and a list of teaching resources on ocaml.org and supported several initiatives in collaboration with Louis Gesbert at OCamlPro, as usual with OPAM development).
The easiest way to make things "just work" are via regular binary builds of the latest releases of OCaml and OPAM on Debian, Ubuntu, CentOS and Fedora, via Ubuntu PPAs and the OpenSUSE Build Service repositories. Our industrial collaborators from Citrix, Jon Ludlam and Dave Scott began an upstreaming initiative to Fedora and sponsored the creation of a CentOS SIG to ensure that binary packages remain up-to-date. We also contribute to the hardworking packagers on MacOS X, Debian, FreeBSD, NetBSD and OpenBSD where possible as well to ensure that binary builds are well rounded out. Richard Mortier also assembled Vagrant boxes that contain OCaml, for use with VirtualBox.
Within OPAM itself, we applied polish to the handling of external dependencies to automate checking that the system libraries required by OPAM are present. Two emerging tools that should help further in 2015 are the opam-user-setup and OPAM-in-a-box plugins that automate first-time configuration. These last two are primarily developed at OCamlPro, with design input and support from OCaml Labs.
We do have a lot of work left to do with making the new user experience really seamless, and help is very welcome from anyone who is interested. It often helps to get the perspective of a newcomer to find out where the stumbling blocks are, and we value any such advice. Just mail [email protected] with your thoughts, or create an issue on how we can improve. A particularly good example of such an initiative was started by Jordan Walke, who prototyped CommonML with a NodeJS-style development workflow, and wrote up his design document for the mailing list. (Your questions or ideas do not need to be as well developed as Jordan's prototype!)
The public Travis CI testing does come with some limitations, since it only checks that the latest package sets install, but not if any transitive dependencies fail due to interface changes. It also doesn't test all the optional dependency combinations due to the 50 minute time limit.
We expanded the OPAM repository testing in several ways to get around this:
-
Individual Repositories: Thomas Gazagnaire built centralised Travis scripts that can be used on any OCaml GitHub repository to easily test code before it is released into OPAM. These scripts are sourced from a central repository and support external, optional and reverse dependency checking across multiple revisions of the compiler. For instance, it just needs one file to test all the supported permutations of the CoHTTP library.
-
Bulk Builds: Damien Doligez and I independently started doing large-scale bulk builds of the repository to ensure that a single snapshot of the package repository can automatically build as many packages as possible. My implementation used the Docker container manager to spawn off 1000s of package builds in parallel and commit the results into a filesystem This required building a Dockerfile eDSL, and the results are now online at https://opam.ocaml.org/builds.
-
OCamlot: An ongoing piece of infrastructure work is to take the bulk build logs (which are around 7GB per daily run), and to store and render them using our Irmin Git store. Expect to see more around this soon; it has the awesome feature of letting any developer clone the build logs for their project locally, to make triage of foreign operating systems as simple as possible.
This ability to do unattended builds of the package repository has also improved the decision making process within the core compiler team. Since we now have a large (3000+ package) corpus of OCaml code, it became a regular occurrence in the 4.02 development cycle to "ask OPAM" whether a particular feature or new syntax would break any existing code. This in turn provides an incentive for commercial users to provide representative samples of their code; for instance, the Jane Street Core releases in OPAM (with their very modular style) act as an open-source canary without needing access to any closed source code.
One good example in 2014 was the decoupling of the Camlp4 macro preprocessor from the main OCaml distribution. Since Camlp4 has been used for over a decade and there are some very commonly used syntax extensions such as type_conv, a simple removal would break a lot of packages. We used OPAM to perform a gradual movement that most users hopefully never noticed by the time OCaml 4.02 was released. First, we added a dummy package in OPAM for earlier versions of the compiler that had Camlp4 built-in, and then used the OPAM constraint engine to compile it as an external tool for the newer compiler revisions. Then we just had to triage the bulk build logs to find build failures from packages that were missing a Camlp4 dependency, and add them to the package metadata.
An interesting comment from Vincent Hanquez
about OPAM is that "OCaml's OPAM is a post-GitHub design". This is very true, as much of the
workflow for pinning git:// URLs emerged out of being early adopters of GitHub for hosting the
MirageOS. OCaml Labs supported two pieces of infrastructure integration around GitHub in 2014:
-
OPAM has a compiler switch feature that lets you run simultaneous OCaml installations and swap between them easily. I used my GitHub API bindings to regularly convert every GitHub pull request into a custom compiler switch. This lets users reporting bugs try out a patched compiler almost immediately upon a fix becoming available.
-
The motivation behind this feature was our collaborator Gabriel Scherer's experiment to enable patch review of OCaml on GitHub, alongside the venerable Mantis bug tracker. We supported this via adding Travis CI support to the main compiler, and also helped to migrate a number of support libraries to GitHub, such as camlp4. These can all be found on the ocaml organisation on GitHub.
Leo White, David Sheets, Amir Chaudhry and Thomas Gazagnaire led the charge to build a modern documentation generator for OCaml, and published an alpha version of codoc 0.2.0 after a lot of work throughout 2014. In the 2014 OCaml workshop presentation (abstract, slides, video), we mentioned the "module wall" for documentation and this attempts to fix it. To try it out, simply follow the directions in the README on that repository, or browse some samples of the current, default output of the tool. Please do bear in mind codoc and its constituent libraries are still under heavy development and are not feature complete, but we're gathering feedback from early adopters.
codoc's aim is to provide a widely useful set of tools for generating OCaml
documentation. In particular, we are striving to:
- Cover all of OCaml's language features
- Provide accurate name resolution and linking
- Support cross-linking between different packages
- Expose interfaces to the components we've used to build
codoc - Provide a magic-free command-line interface to the tool itself
- Reduce external dependencies and default integration with other tools
We haven't yet achieved all of these at all levels of our tool stack but are
getting close, and the patches are all under discussion for integration into
the mainstream OCaml compiler.
codoc 0.2.0 is usable today (if a little rough in some areas like default CSS),
and there is a blog post that outlines the architecture of the new system to
make it easier to understand the design decisions that went into it.
As the amount of infrastructure built around the ocaml.org domain grows (e.g. mailing lists, file hosting, bulk building), it is important to establish a governance framework to ensure that it is being used as best needed by the wider OCaml community.
Amir Chaudhry took a good look at how other language communities organise themself,
and began putting together a succinct
governance framework
to capture how the community around ocaml.org operates, and how to quickly resolve any conflicts
that may arise in the future. He took care to ensure it had a well-defined scope, is
simple and self-contained, and (crucially) documents the current reality. The result
of this work is circulating privately through all the existing volunteers for a first
round of feedback, and will go live in the next few months as a living document
that explains how our community operates.
One consequence of OCaml's age (close to twenty years old now) is that the tools built around the compiler have evolved fairly independently. While OPAM now handles the high-level package management, there is quite a complex ecosystem of other components that are complex for new users to get to grips with: OASIS, ocamlfind, ocamlbuild, and Merlin to name a few. Each of these components (while individually stable) have their own metadata and namespace formats, further compounding the lack of cohesion of the tools.
Thomas Gazagnaire and Daniel Buenzli embarked on an effort to build an eDSL that unifies OCaml package descriptions, with the short-term aim of generating the support files required by the various support tools, and the long-term goal of being the integration point for the build, test and documentation generation lifecycle of an OCaml/OPAM package. This prototype, dubbed Assemblage has gone through several iterations and design discussions over the summer of 2014. Daniel has since been splitting out portions of it into the Bos OS interaction library.
Assemblage is not released officially yet, but we are committed to resuming work on it this summer when Daniel visits again, with the intention of unifying much of our workflow through this tool. If you are interested in build and packaging systems, now is the time to make your opinion known!
We also spent time in 2014 working on the core OCaml language and compiler, with our work primarily led by Jeremy Yallop and Leo White. These efforts were not looking to make any radical changes in the core language; instead, we generally opted for evolutionary changes that either polish rough edges in the language (such as open type and handler cases), or new features that fit into the ML style of building programs.
The OCaml 4.02 series was primarily developed and released in 2014. The ChangeLog generated much user excitement, and we were also pleased to have contributed several language improvements.
OCaml's try and match constructs are good at dealing with exceptions
and values respectively, but neither constructs can handle both values and exceptions.
Jeremy Yallop investigated
how to handle success
more elegantly, and an elegant unified syntax emerged. A simple example is that
of a stream iterator that uses exceptions for control flow:
let rec iter_stream f s =
match (try Some (MyStream.get s) with End_of_stream -> None) with
| None -> ()
| Some (x, s') -> f x; iter_stream f s'
This code is not only verbose, but it also has to allocate an option value to ensure
that the iter_stream calls remains tail recursive. The new syntax in OCaml 4.02
allows the above to be rewritten succinctly:
let rec iter_stream f s =
match MyStream.get s with
| (x, s') -> f x; iter_stream f s'
| exception End_of_stream -> ()
Read more about the background of this feature in Jeremy's blog post, the associated discussion in the upstream Mantis bug, and the final manual page in the OCaml 4.02 release. For an example of its use in a real library, see the Jane Street usage in the s-expression handling library (which they use widely to reify arbitrary OCaml values and exceptions).
A long-standing trick to build universal containers
in OCaml has been to encode them using the exception exn type.
There is a similar concept of a universal type in Standard ML,
and they were described in the "Open Data Types and Open Functions"
paper by Andres Löh and Ralf Hinze in 2006.
Leo White designed, implemented and upstreamed support for extensible variant types in OCaml 4.02. Extensible variant types are variant types that can be extended with new variant constructors. They can be defined as follows:
type attr = ..
type attr += Str of string
type attr +=
| Int of int
| Float of float
Pattern matching on an extensible variant type requires a default case to handle unknown variant constructors, just as is required for pattern matching on exceptions (extensible types use the exception memory representation at runtime).
With this feature added, the OCaml exn type simply becomes a special case of open extensible
types. Exception constructors can be declared using the type extension syntax:
type exn += Exc of int
You can read more about the discussion behind open extensible types in the upstream Mantis bug. If you'd like to see another example of their use, they have been adopted by the latest releases of the Jane Street Core libraries in the Type_equal module.
A common criticism of OCaml is its lack of support for ad-hoc polymorphism.
The classic example of this is OCaml's separate addition operators for
integers (+) and floating-point numbers (+.).
Another example is the need for type-specific printing functions (print_int, print_string, etc.)
rather than a single print function which works across multiple types.
Taking inspiration from Scala's implicits and Modular Type Classes by Dreyer et al., Leo White designed a system for ad-hoc polymorphism in OCaml based on using modules as type-directed implicit parameters. The design not only supports implicit modules, but also implicit functors (that is, modules parameterised by other module types) to permit the expression of generic modular implicits in exactly the same way that functors are used to build abstract data structures.
Frederic Bour joined us as a summer intern and dove straight into the implementation, resulting in an online demo and ML Workshop presentation (abstract, video and paper). Another innovation in how we've been trialling this feature is the use of Andy Ray's IOCamlJS to publish an interactive, online notebook that is fully hosted in the browser. You can follow the examples of modular implicits online, or try them out on your own computer via an OPAM switch:
opam switch 4.02.0+modular-implicits
eval `opam config env`
opam install utop
utop
Some of the early feedback on modular implicits from industrial users was interesting.
Jane Street commented that although this would be a big usability leap, it would be dangerous to
lose control over exactly what goes into the implicit environment (i.e. the programmer should always
know what (a + b) represents by locally reasoning about the code). The current design thus follows
the ML discipline of maintaining explicit control over the namespace, with any ambiguities in resolving
an implicit module type resulting in a type error.
In addition to ad-hoc polymorphism, support for parallel execution on multicore CPUs is undoubtedly the most common feature request for OCaml. This has been high on our list after improving tooling support, and Stephen Dolan and Leo White made solid progress in 2014 on the core runtime plumbing required.
Stephen initially added thread-local support to the OCaml compiler. This design avoided the need to make the entire OCaml runtime preemptive (and thus a huge patch) by allocating thread-local state per core.
We are now deep into the design and implementation of the programming abstractions built over these low-level primitives. One exciting aspect of our implementation is much of the scheduling logic for multicore OCaml can be written in (single-threaded) OCaml, making the design very flexible with respect to heterogenous hardware and variable IPC performance.
To get feedback on the overall design of multicore OCaml, we presented at OCaml 2014 (slides, video and abstract), and Stephen visited INRIA to consult with the development team and Arthur Chargueraud (the author of PASL). Towards the end of the year, KC Sivaramakrishnan finished his PhD studies at Purdue and joined our OCaml Labs group. He is the author of MultiMlton, and is now driving the completion of the OCaml multicore work along with Stephen Dolan, Leo White and Mark Shinwell. Stay tuned for updates from us when there is more to show later this year!
The Ctypes library started as an experiment with GADTs by Jeremy Yallop, and has since ballooned in a robust, comprehensive library for safely interacting with the OCaml foreign function interface. The first release came out in time to be included in Real World OCaml in lieu of the low-level FFI (which I was not particularly enamoured with having to explain in a tight page limit).
Throughout 2014, Jeremy expanded support for a number of features requested by users (both industrial and academic) who adopted the library in preference to manually writing C code to interface with the runtime, and issued several updated releases.
The first release of Ctypes required the use of libffi
to dynamically load shared libraries and dynamically construct function call stack
frames whenever a foreign function is called. While this works for simple libraries,
it cannot cover all usecases, since interfacing with C demands an understanding
of struct memory layout, C preprocessor macros, and other platform-dependent quirks
which are more easily dealt with by invoking a C compiler. Finally, the performance
of a libffi-based API will necessarily be slower than writing direct C stub code.
While many other language FFIs provide separate libraries for dynamic and static FFI libraries, we decided to have a go at building a modular version of Ctypes that could handle both cases from a single description of the foreign function interface. The result (dubbed "Cmeleon") remained surprisingly succinct and usable, and now covers almost every use of the OCaml foreign function interface. We submitted a paper to ICFP 2015 dubbed "A modular foreign function interface" that describes it in detail. Here is a highlight of how simple a generic binding looks:
module Bindings(F : FOREIGN) = struct
open F
let gettimeofday = foreign "gettimeofday"
(ptr timeval @-> ptr timezone @-> returning int)
end
The FOREIGN module type completely abstracts the details of whether or not dynamic
or static binding is used, and handles C complexities such as computing the struct
layout on the local machine architecture.
The other nice result from functorising the foreign function interface emerged when we tried to invert the FFI and serve a C interface from OCaml code (for example, by compiling the OCaml code as a shared library). This would let us begin swapping out C libraries that we don't trust with safer equivalents written in OCaml.
You can see an example
of how inverted stubs work via a simple C XML parsing exposed
from the Xmlm library. We can define a C struct
by:
(* Define a struct of callbacks (C function pointers) *)
let handlers : [`handlers] structure typ = structure "handlers"
let (--) s f = field handlers s (funptr f)
let on_data = "on_data" -- (string @-> returning void)
let on_start_tag = "on_start_tag" -- (string @-> string @-> returning void)
let on_end_tag = "on_end_tag" -- (void @-> returning void)
let on_dtd = "on_dtd" -- (string @-> returning void)
let on_error = "on_error" -- (int @-> int @-> string @-> returning void)
let () = seal handlers
and then expose this via C functions:
module Stubs(I : Cstubs_inverted.INTERNAL) = struct
(* Expose the type 'struct handlers' to C. *)
let () = I.structure handlers
(* We expose just a single function to C. The first argument is a (pointer
to a) struct of callbacks, and the second argument is a string
representing a filename to parse. *)
let () = I.internal "parse_xml"
(ptr handlers @-> string @-> returning void) parse
end
You can find the full source code to these snippets on the ocaml-ctypes-inverted-stubs-example repository on GitHub.
We'll be exploring this aspect of Ctypes further in 2015 for SSL/TLS with David Kaloper and Hannes Mehnert, and Microsoft Research has generously funded a PhD studentship to facilitate the work.
Ctypes benefited enormously from several external contributions from the OCaml community. From a portability perspective, A. Hauptmann contributed Windows support, and Thomas Leonard added Xen support to allow Ctypes bindings to work with MirageOS unikernels (which opens up the intriguing possibility of accessing shared libraries across virtual machine boundaries in the future). C language support was fleshed out by Edwin Torok contributing typedef support, Ramkumar Ramachandra adding C99 bools and Peter Zotov integrating native strings.
The winner of "most enthusiastic use of OCaml Labs code" goes to Thomas Braibant of Cryptosense, who used every feature of the Ctypes library (consider multi-threaded, inverted, staged and marshalled bindings) in their effort to hack the hackers. David Sheets comes a close second with his implementation of the FUSE binary protocol, parameterised by version quirks.
If you're using Ctypes, we would love to hear about your particular use. A search on GitHub and OPAM reveals over 20 projects using it already, including industrial use at Cryptosense and Jane Street, and ports to Windows, *BSD, MacOS X and even iPhone and Android. There's a getting started guide, and a mailing list available.
In addition to the online community building, we also participated in a number of conferences and face-to-face events to promote education about functional programming.
### Conferences and Talks
There has been a huge growth in the number of quality conferences in recent years, making
it tough to choose which ones to attend.
ICFP is the academic meeting point that predates most of them,
and we participated extensively in 2014 via talks, tutorials and a keynote at the Haskell Symposium.
I also served on the program committee and
industrial relations chair
and took over as the steering committee
chair of CUFP. Jeremy Yallop, Thomas Gazagnaire and Leo White all
served program committees on workshops, with Jeremy also chairing this year's ML Workshop.
Outside of academic conferences, we participated in a number of non-academic conferences such as QCon, OSCON, CCC, New Directions in OS, FunctionalConf, FPX and FOSDEM. The vast majority of these talks were about the MirageOS, and slides can be found at decks.openmirage.org.
Yaron Minsky and I have run OCaml tutorials for ICFP for a few years, and we finally hung up our boots in favour of a new crowd.
Jeremy Yallop and Leo White stepped up to the mark with their ICFP/CUFP 2014 Introduction to OCaml tutorial, which had the additional twist of being taught entirely in a web browser by virtue of using the js_of_ocaml and IOCamlJS. They decided that a good practical target was the popular 2048 game that has wasted many programmer hours here at OCaml Labs. They hacked on it over the summertime, assisted by our visitor Daniel Buenzli who also released useful libraries such as Vg, React, Useri, and Gg.
The end result is satisfyingly playable online, with the source code available at ocamllabs/2048-tutorial.
Thomas Gazagnaire got invited to Bangalore for Functional Conf later in the year, and hee extended the interactive tutorial notebook and also ran an OCaml tutorial to a packed room. We were very happy to support the first functional programming conference in India, and hope to see many more such events spring up! Amir Chaudhry then went to Belgium to FOSDEM 2015 where he showed off the 2048 game running as an ARM unikernel to a crowd of attendees at the Xen booth.
Jeremy Yallop and Leo White (with assistance from Alan Mycroft and myself) also led the design of a new graduate course on Advanced Functional Programming at the Computer Laboratory. This ran in the Lent Term and was over-subscribed by three times the number who pre-registered (due to a number of PhD students and our collaborators from Citrix also attending).
The course materials are freely available online and cover the theory behind functional programming, and then move onto type inference, abstraction and parametricity, GADTs, rows, monads, and staging. We will be running this again in future years, and the lecture materials are already proving useful to answer mailing list questions.
We also had the pleasure of mentoring up-and-coming functional programmers via several outreach programs, both face-to-face and remote.
We started the Cambridge Compiler Hacking sessions in a small way towards the end of 2013 in order to provide a local, friendly place to assist people who wanted to dip their toes into the unnecessarily mysterious world of programming language hacking. The plan was simple: provide drinks, pizza, network and a bug list of varying difficulty for attendees to choose from and work on for the evening, with mentoring from the experienced OCaml contributors.
We continued this bi-monthly tradition in 2014, with a regular attendance of 15-30 people, and even cross-pollinated communities with our local F# and Haskell colleagues. We rotated locations from the Cambridge Computer Laboratory to Citrix, Makespace, and the new Cambridge Postdoc Centre. We posted some highlights from sessions towards the start of the year, and are very happy with how it's going. There has even been uptake of the bug list across the water in France, thanks to Gabriel Scherer.
In 2015, we'd like to branch out further and host some sessions in London. If you have a suggestion for a venue or theme, please get in touch!
There has been a laudable rise in summer programs designed to encourage diversity in our community, and we of course leap at the opportunity to participate in these when we find them.
- The GNOME Outreach Program (now also known as Outreachy) had one funded place for Xen and MirageOS. Mindy Preston did a spectacular blog series about her experiences and motivations behind learning OCaml.
- The Google Summer of Code 2014 also had us participating via MirageOS, and Jyotsna Prakash took on the challenging job of building OCaml bindings for Amazon EC2, also detailed on her blog.
- Amir Chaudhry began the Mirage Pioneer Projects initiative to give beginners an easier onramp, and this has taken off very effectively as a way to advertise interesting projects for beginners at varying levels of difficulties.
Our own students also had the chance to participate in such workshops to get out of Cambridge in the summer! Heidi Howard liveblogged her experiences at the PLMW workshop in Mumbai. Meanwhile, David Sheets got to travel to the slightly less exotic London to liveblog OSIO, and Leonhard Markert covered ICFP 2014 as a student volunteer.
Our blog roll maintains the ongoing stream of activity from the OCaml Labs crew, but there were some particular highlights throughout 2014.
- Thomas Leonard began writing about his experiences with switching his 0install installation system from Python to OCaml and what you gain with OCaml. This series led to a bunch of interesting feedback on social networking sites, and Thomas joined the group full-time to work on our research into unikernels.
- Magnus Skjegstad returned from Norway to Cambridge to work on MirageOS, and came up with some crazy experiements, as well as helping to build Vagrant images of the OCaml development environment.
- Amir Chaudhry began his quest to port his website website to a Jekyll unikernel.
- The Mirage 2.0 release in the summer of 2014 saw a slew of blogs posts about the surge in MirageOS activity.
It wasn't all just blogging though, and Jeremy Yallop and Leo White in particular participated in some epic OCaml bug threads about new features, and explanations about OCaml semantics on the mailing list.
Amir Chaudhry also continued to curate and develop the content on the ocaml.org website with our external collaborators Ashish Agarwal, Christophe Troestler and Phillippe Wang. Notably, it is now the recommended site for OCaml (with the INRIA site being infrequently updated), and also hosts the ACM OCaml Workshop pages. One addition that highlighted the userbase of OCaml in the teaching community came from building a map of all of the universities where the language is taught, and this was Yan Shvartzshnaider's first contribution to the site.
Finally, a really important part of any community is hanging out with each other to chat over ideas in a friendly environment. As usual, we had a very steady stream of visitors and interns throughout 2014 to facilitate this.
Frederic Bour, Benjamin Farinier and Matthieu Journault joined us as summer interns from their respective universities in France as part of their Masters programs. Frederic worked on modular implicits and gave a great talk at the OCaml Users group. Benjamin and Matthieu worked on Irmin data structures and complexity (and merge-queues and merge-ropes), and Benjamin had his paper on "Mergeable Persistent Data Structures" accepted to JFLA 2015, while Matthieu's work on efficient algorithms for synchronising Irmin DAGs is being integrated into the upstream source code.
Daniel Buenzli repeated his visit from 2013 and spent a productive summer with us, commenting on almost every project we're working on. In his own words (edited for brevity):
I started by implementing and releasing Uucp, a library to provide efficient access to a selection of the properties of the latest Unicode Character database (UCD). [...] As a side effect of the previous point I took time to write an absolute minimal introduction to Unicode. [...] Since I was in this Unicode business I took the opportunity to propose a 31 loc patch to the standard library for a type to represent Unicode scalar values (an Unicode character to be imprecise) to improve interoperability.
The usual yearly update to OpenGL was announced at the Siggraph conference. This prompted me to update the ctypes-based tgls library for supporting the latest entry point of OpenGL 4.5 and OpenGL ES 3.1. Since the bindings are automatically generated from the OpenGL XML registry the work is not too involved but there's always the odd function signature you don't/can't handle automatically yet.
Spend quite a bit (too much) time on useri, a small multi-platform abstraction for setting up a drawing surface and gather user input (not usury) as React events. Useri started this winter as a layer on top of SDL to implement a CT scan app and it felt like this could be the basis for adding interactivity and animation to Vg/Vz visualizations – js viz libraries simply rely on the support provided by the browser or SVG support but Vg/Vz strives for backend independence and clear separations of concern (up to which limit remains an open question). Unfortunately I couldn't bring it to a release and got a little bit lost in browser compatibility issues and trying to reconcile what browser and SDL give us in terms of functionality and way of operating, so that a maximum of client code can be shared among the supported platforms. But despite this non-release it still managed to be useful in some way, see the next point.
Helped Jeremy and Leo to implement the rendering and interaction for their ICFP tutorial 2048 js_of_ocaml implementation. This featured the use of Gg, Vg, Useri and React and I was quite pleased with the result (despite some performance problems in certain browsers, but hey composable rendering and animation without a single assignement in client code). It's nice to see that all these pains at trying to design good APIs eventually fit together [...]
A couple of visitors joined us from sunny Morocco, where Hannes Mehnert and David Kaloper had gone to work on a clean-slate TLS stack. They found the MirageOS effort online, and got in touch about visiting. After a very fun summer of hacking, their stack is now the standard TLS option in MirageOS and resulted in the Bitcoin Pinata challenge being issued! Hannes and David have since moved to Cambridge to work on this stack full-time in 2015, but the internships served as a great way for everyone to get to know each other.
We also had the pleasure of visits from several of our usually remote collaborators. Christophe Troestler, Yaron Minsky, Jeremie Diminio and Andy Ray all visited for the annual OCaml Labs review meeting in Christ's College. There were also many academic talks from foreign visitors in our SRG seminar series, ranging from Uday Khedkar from IIT to Oleg Kiselyov deliver multiple talks on staging and optimisation (as well as making a celebrity appearance at the compiler hacking session, and Yaron Minsky delivering an Emacs-driven departmental seminar on his experiences with Incremental computation.
The OCaml Labs are of course based in the Cambridge Computer Laboratory, where our day job is to do academic research. Balancing the demands of open source coding, community efforts and top-tier research has be a tricky one, but an effort that has been worthwhile.
Our research efforts are broadly unchanged from 2013 (it takes time to craft good ideas!), and this will not be an exhaustive recap. Instead, we'll summarise them here and point to our papers that describe the work in detail.
-
The MirageOS really found its own feet in 2014, with a summer 2.0 release and an extensive end-of-year recap. The most notable thing has been how well the MirageOS research work has melded with the core OCaml Labs efforts, since much of it has been constructing good quality OCaml libraries to plug holes in the ecosystem. It also served to make us use OPAM on a day-to-day basis for our own work, thus creating an effective feedback loop between open-source and research.
-
In the Trilogy2 and UCN EU projects, we built out MirageOS features such as the Jitsu toolstack for the "just-in-time" summoning of unikernels in response to DNS requests. This paper will be presented next month at UlSENIX NSDI. It also drove the development of the ARMv7 port, an architecture for which OCaml has an excellent native code generator, as well as more experimental forays into BitCoin incentive schemes for distributed systems.
-
The Irmin Git-like branchable store created by Thomas Gazagnaire matured, with Dave Scott prototyping a complex port of the XenStore database to Irmin, thus letting us show off debugging systems with Git. We had a paper accepted on some early datastructures accepted at JFLA, and Thomas Leonard is building the JavaScript backend for running in-browser, while Yan Schvartzshnaider is experimenting with graph processing over the DAG representation for privacy-friendly queries. KC is investigating how to adapt his PLDI 2015 paper on Quelea into using Irmin as a backend as well.
-
The Higher kinded polymorphism library written by Jeremy Yallop and Leo White was published in FLOPS 2014, forming a basis for building more complex use-cases that need the flexibility of higher kinded types without requiring functorising code.
-
Our long standing research into personal online privacy led to our next system target that uses unikernels: the Databox paper outlines the architecture, and was covered in the Guardian newspaper. Jon Crowcroft led the establishment of the Cambridge wing of the Microsoft Cloud Computing Research Center to consider the legal aspect of things, and so we have made forays outside of technology into considering the implications of region-specific clouds as well.
Some of the most exciting work done in the group as part of the REMS and NaaS projects came towards the end of 2014 and start of 2015, with multiple submissions going into top conferences. Unfortunately, due to most of them being double blind reviewed, we cannot link to the papers yet. Keep an eye on the blog and published paper set, or ask us directly about what's been going on!
As spring breaks and the weather (almost) becomes bearable again, we're setting our work priorities for the remainder of the year.
-
Tooling Cohesion: The entire core team is focussed on fusing together the individual tools that have been created last year into a cohesive OCaml Platform release that covers the lifecycle of documentation, testing and build. This is being managed by Amir Chaudhry. OPAM remains at the heart of this strategy, and Louis Gesbert and Thomas Gazagnaire have settled on the OPAM 1.3 roadmap (summary).
-
Multicore: KC Sivaramakrishnan has joined the core OCaml Labs fulltime to drive the multicore work into a publically testable form. Leo White recently departed after many productive years in Cambridge to head into a career in industry (but still remains very much involved with OCaml development!).
-
Language Evolution: Jeremy Yallop continues to drive our efforts on staged programming, modular implicits, and a macro system for OCaml, all of which are key features that make building complex, reliable systems more tractable than ever.
I'd like to thank the entire team and wider community for a wonderfully enjoyable 2014 and start of 2015, and am very thankful to the funding and support from Jane Street, Citrix, British Telecom, RCUK, EPSRC, DARPA and the EU FP7 that made it all possible. As always, please feel free to contact any of us directly with questions, or reach out to me personally with any queries, concerns or bars of chocolate as encouragement.