OCaml Weekly News

Hello

Here is the latest OCaml Weekly News, for the week of July 16 to 23, 2024.

A Tour of the Living Library – A Safer FFI

Archive: https://discuss.ocaml.org/t/blog-a-tour-of-the-= living-library-a-safer-ffi/14981/1

Matt Walker announced

I've written a new blog post on the living library I announced= a few days ago. Please give it a read if you're interested in safe use of= Ctypes, or otherwise need lifetime management in OCaml.

Would love to hear your views in this thread!

https://fizzixnerd.com/blog/2024-07-17-touring-the-living-library/

first release of rpmfile

Archive: https://discuss.ocaml.org/t/ann-first-release-of-rpmfile/14985/1=

Mikhail announced

I'm happy to announce the first release of rpmfile, a small library for reading meta information fro= m RPM packages (version 3.0). It uses the Angstrom combinator parser librar= y, which allows it to perform streaming parsing using Lwt or Async.

How to get a package summary:

module Rpm_reader =3D Rpmfile.Reader (Rpmfile.Selector.All)

let metadata =3D Rp=
m_reader.of_file_exn "hello-2.12.1-1=
.7.x86_64.rpm"

Rpmfile.summary metadata
(* - : string =3D "A Friendly Greetin=
g Program" *)

The default reader module can read from a string or file, but has poor perf= ormance. It needs a selector module to select the tags to parse. The exampl= e uses Selector.All to parse all tags.=20

I am developing this library for my pet project to create a self-hosted RPM= repository management solution.=20

Thank you for your attention!

Dune dev meeting

Archive: https://discuss.ocaml.org/t/ann-dune-dev-meeting/14994/1

maiste announced

We are organizing a new public Dune dev meeting on Wednesday, July, 24th= at 10am CET. It will be one hour long.

Whether you are a maintainer, a regular contributor, a new joiner or just c= urious, feel free to join! The goal of these meetings is to provide a place= to discuss the ongoing work together :smile: Below, you can find the agenda for this meeting:

:scroll: Agenda

Quick presentation about the attendees.
Presentation about the ongoing work in Dune.
Questions and Answers.
Information discussions

:chains: Links

Meeting link: zoom
Calendar event: google calendar
Wiki with information and previous notes: GitHub Wiki

Fighting Mutation with Mutation in Living

Archive: https://discuss.ocaml.org/t/blog-fighting-muta= tion-with-mutation-in-living/15003/1

Matt Walker announced

New blog post about fixing the mistakes in the living library.= Please take a look if you're interested in interfacing ocaml with externa= l resources.

https://f= izzixnerd.com/blog/2024-07-21-fixing-living/

In particular, I think the library comes good with its guarantees now that

if every function is properly dependent, and
you only unsafe_free values that are disjoint from their d= ependencies, then

you will obtain a sound program when using the Ctypes ffi, in terms of ther= e being no use-after-free errors.

Please let me know if you disagree!

A small extension of Bigarray.Genarray adding iteration, mappi= ng and folding

Archive: https://discu= ss.ocaml.org/t/ann-a-small-extension-of-bigarray-genarray-adding-iteration-= mapping-and-folding/15005/1

NAlec announced

I needed a few functions which were missing in the OCaml library : Bigarray.Genarr= ay, and decided to write them for my own purpose :=20

Iteration on genarrays
mapping on genarrays
folding on genarrays

Today I believe this can be usefull for others, and may suffer a code inspe= ction as I am not that experienced in OCaml. I am ready to have this piece = of code evolve if it is usefull so even (and maybe first) a feedback on the= usefullness of such code is welcome.

The only alternative I was given was the famous Owl library, which was way = to heavy for my needs, and not easily usable (if not understandable). This = extension is very simple, it is its strenght. Ultimately, it could be merge= d in the standard library … maybe after some work indeed : you tell = me.=20

There is a clean documentation I guess, hope this can help : GenArrayIter

Looking forward to hearing from you all.

cudajit: Bindings to the `cuda` and `nvrtc libraries`




Archive: https://discuss.ocaml.org/t/ann-cudajit-=
bindings-to-the-cuda-and-nvrtc-libraries/15010/1


Lukasz Stafiniak announced

Hi! I'm happy to share cudajit 0.4.0: manually-selected bindings for Nvidia= GPU programming. cudajit should soon propagate to the opam repository. Bindings to the cuda and nvrtc libraries with a unified interface API documentation Currently supported:

Compiling a kernel with conversion to PTX, launching a kernel.
Synchronous and asynchronous memory copying.
Contexts and streams.
(GPU) device attributes.

Currently not supported:

Events.
CUDA graph features.
Cooperative kernel launch.

Let me know your needs so I can prioritize. PRs are also welcome!

Rpmfile 0.2.0 - changelog

Archive: https://discuss.ocaml.org/t/ann-first-release-of-rpmfile/14985

Mikhail announced

Hello again, everyone. :wave: Today I want to tell you about what has chang= ed in a new version of my rpm= file library (previous topic) for reading meta-information from RPM= packages.Should I post this in the forum? I'm sorry.

Changes

Fixed incorrect string parsing. I just forgot to make advance after take_till (commit);

angstrom-unix is used by default to read files in the Reader module functions. Previously, a RPM package was read entire= ly into memory;
Optimized partial parsing of header= sections. Reduced unnecessary memory allocations (c= ommit);
Decoding integers (int8/int16/int32/int64) to native int in acce= ss functions[^1] (like Rpmfile.payload_size). You can also use= get to get "raw" values;
Improved compatibility with 4.0 version of RPM format by using nativ= e int;
Added a module Selector.Base to select only basic package = info (commit);
Some new access functions and output fields of the CLI utility.

rpmfile vs rpm -qi

Not a real "benchmark" for parsing 1.5 GB packages.

$ time rpm -qi repo/*.rpm
Executed in  226.82 millis    fish           external
   usr time  212.74 millis    1.06 millis  211.68 millis
   sys time   13.23 millis    0.00 millis   13.23 millis

$ time rpmfile repo/*.rpm
Executed in  153.97 millis    fish           external
   usr time  116.74 millis    0.00 millis  116.74 millis
   sys time   30.65 millis    1.47 millis   29.18 millis

Rpmfile doesn't verify signatures, which is why it is "faster".

What's next?

This is enough for my tasks, so there probably won't be a next release :col= d_face:

To-Do: functionality to work with signatures, read payload, implement write= r module for create packages.

Thank you for your attention!

P.S. I also want to apologize for my terrible English.=20

[^1]: The access function gets and decodes values from a metadata record.







Exploring the Docusaurus+Odoc combo


Archive: https://discuss.ocaml.org/t/exploring-the-docusaurus-odoc=
-combo/15012/1




Mathieu Barbin announced


To OCaml & Docusaurus enthusiasts out there :camel:+ :sauropod:



Some time ago, I shared my experience using Docusaurus to document an OCaml=
 project, highlighting the integration between Docusaurus, ocaml-mdx, and t=
he dune workflow (previous post here).



Today I wanted to share that I've resumed this exploration in documentation=
 tools to try and integrate odoc-generated pages into Docusaurus, with the =
aim of creating a somewhat minimal template/example for this.



I've published my experiment here: https://mbarbin.github.io/doc-experiment-docusauru=
s/.



Integrating odoc posed challenges - I've written about the (pragmatic) appr=
oach I took here. I'm linking this odoc issue too, for reference about exploring more nat=
ive solutions for this interop.



Have you too tried this "magic combo" of Docusaurus, Odoc, and OCaml tools?=
 And if so, how did you approach it? Do you have insights or suggestions? I=
f this sparks your curiosity, please don't hesitate to engage with the repo=
sitory.









Mopsa 1.0 – Modular Open Platform for Static Analysis


Archive: https://discuss.ocaml.org/t/ann-mopsa=
-1-0-modular-open-platform-for-static-analysis/15013/1




Rapha=C3=ABl Monat announced


On behalf of all its developers, I am glad to announce the release of Mopsa 1.0! You can just=
 opam install mopsa.



Mopsa stands for Modular and Open Platform for Static Analysis. It aims at =
easing the development and use of static analyzers. More specifically, Mops=
a is a generic framework for building sound static analyzer based on the th=
eory of abstract interpretation. Mopsa is independent of language and abstr=
action choices. Developers are free to add arbitrary abstractions (numeric,=
 pointer, memory, etc.) and syntax iterators for new languages. Mopsa encou=
rages the development of independent abstractions which can cooperate or be=
 combined to improve precision.



Mopsa currently support the analysis of Python, C and Python+C programs. It=
 reports run-time errors on C programs and uncaught exceptions on Python pr=
ograms. Our benchmarks provide an illustrative overview of what Mopsa can c=
urrently analyze. All analyses currently provided are flow and context-sens=
itive (i.e, control-flow operators are taken into account by the analysis, =
and functions are analyzed by virtual inlining). The C analysis is actively=
 developed and maintained. The Python and Python+C analyses work on real-wo=
rld examples, but are not actively developed.



Please note that Mopsa is an academic tool under development. Feel free to =
submit issues if you=
 encounter any bug!



Additional resources:=20


user ma=
nual
=
demo of our abstract debugger
=
academic overview of Mopsa, and in a PhD thesis
co=
reutils benchmarks on which Mopsa can run









OCaml 5 performance


Archive: https://discuss.ocaml.org/t/ocaml-5-performance/15014/1




Thomas Leonard announced


I've been trying out some tools to investigate performance problems in my O=
Caml programs and I've written up my experiences here in case other people =
find it useful:



http=
s://roscidus.com/blog/blog/2024/07/22/performance/
ht=
tps://roscidus.com/blog/blog/2024/07/22/performance-2/



The first post examines a case of slow IO in a concurrent Eio program, and =
the second looks at poor GC performance in a multicore app.



In particular, it seems that minor GC performance is very sensitive to othe=
r work running on the machine, since any domain being late will trigger the=
 others to sleep, e.g.









I'd be interested to know if others can shed more light on this, or have ot=
her profiling tools they've found useful.









Other OCaml News



From the ocaml.org blog


Here are links from many OCaml blogs aggregated at the ocaml.org blog.



OC=
aml 5 performance part 2
OCam=
l 5 performance problems
OCaml Compiler Manual HTML Generation









Old CWN


If you happen to miss a CWN, you can send me a message and I'll mail it to you, or go take a loo=
k at the archive or the <=
a href=3D"https://alan.petitepomme.net/cwn/cwn.rss">RSS feed of the archive=
s.



If you also wish to receive it every week by mail, you may subscribe to the=
 caml-list.




Alan Schmitt

OCaml Weekly News

Table of Contents

A Tour of the Living Library – A Safer FFI

Matt Walker announced

first release of rpmfile

Mikhail announced

Dune dev meeting

maiste announced

:scroll: Agenda

:chains: Links

Fighting Mutation with Mutation in Living

Matt Walker announced

A small extension of Bigarray.Genarray adding iteration, mappi= ng and folding

NAlec announced

cudajit: Bindings to the cuda and nvrtc libraries

Lukasz Stafiniak announced

Rpmfile 0.2.0 - changelog

Mikhail announced

Changes

rpmfile vs rpm -qi

What's next?

Exploring the Docusaurus+Odoc combo

Mathieu Barbin announced

Rapha=C3=ABl Monat announced

OCaml 5 performance

Thomas Leonard announced

Other OCaml News

From the ocaml.org blog

Old CWN

cudajit: Bindings to the `cuda` and `nvrtc libraries`