OCaml Weekly News

Previous Week Up Next Week


Here is the latest OCaml Weekly News, for the week of July 10 to 17, 2018.

Table of Contents

Format module from the Standard Library

Continuing this thread, Jesse Haber-Kucharsky said

Check out this paper, whose purpose is to explain and document the module.

Rizo added

I vividly remember how strange it looked the first time I tried using it. Other people recommended some resources to understand how to used it and how the boxing model works. I strongly recommend using the fmt library that abstracts a lot of complexity of the Format module.

Vladimir Keleshev also said

Not a tutorial, but you might find it helpful to see some examples of Format usage.

Here's a pretty-printer for a small subset of XML:


A pretty-printer for a subset of JavaScript:


A pretty-printer for a subset of Python:


Personally, I like the Format module very much and never had to reach for another library, but I remember that it was tricky to wrap my head around when I just started to use it.

Decompress / Checkseum / Optint

Calascibetta Romain announced

I'm happy yo announce a new release of decompress.0.8 available for installation via OPAM. The distribution provides a new package rfc1951 which implements the Request for Comments 1951 - a subset of zlib.

By this change, we put a prefix on previous modules Inflate and Deflate to recognize which format you use. Semantically, we don't have an update of the API. However, decompress needs a new package now: checkseum which depends on optint.

checkseum is a little library à la digestif to provide a shared interface which allows to use an implementation of ADLER-32 or CRC32C. Implementation could be a C implementation or an OCaml implementation, so when the use want to link decompress (or checkseum), he needs to specify which implementation he wants to use: checkseum.c or checkseum.ocaml - at this stage of dune, this dependency needs to be the first of the rest. By the OCaml implementation, we keep the possibility to compile decompress to JavaScript.

Finally, optint is a little library which provide an abstract type to represent an 32-bits integer. From your architecture (x64 or x86), optint will use an unboxed integer (fast) or a boxed int32 (slow). So, at least, you have a integer with 32 bits availables (or more - 63 on x64).

The next release is focus on to implement (finally) GZIP and, may be, LZO. The new internal design of decompress allows to put on top of rfc1951 the GZIP header. Then, we provided an inverted stub of decompress (to use decompress on a C code) to allow a real benchmark with others implementations.

OCaml release 4.07.0

Damien Doligez announced

We have the pleasure of celebrating the birthday of Nikola Tesla and Marcel Proust by announcing the release of OCaml version 4.07.0. This is a major release with a lot of new features, see the changelog below.

It is available as a bunch of OPAM switches, or as a source download here: http://caml.inria.fr/pub/distrib/ocaml-4.07.

Happy hacking, – Damien Doligez for the OCaml team.

OCaml 4.07.0 (10 July 2018)

(Changes that can break existing programs are marked with a "*")

Language features:

  • MPR#6023, GPR#1648: Allow type-based selection of GADT constructors (Thomas Refis and Leo White, review by Jacques Garrigue and Gabriel Scherer)
  • GPR#1546: Allow empty variants (Runhang Li, review by Gabriel Radanne and Jacques Garrigue)

Standard library:

  • MPR#4170, GPR#1674: add the constant Float.pi. (Christophe Troestler, review by Damien Doligez)
  • MPR#6139, GPR#1685: Move the Bigarray module to the standard library. Keep the bigarray library as on overlay adding the deprecated map_file functions (Jérémie Dimino, review by Mark Shinwell)
  • MPR#7690, GPR#1528: fix the float_of_string function for hexadecimal floats with very large values of the exponent. (Olivier Andrieu)
  • GPR#1002: add a new Seq module defining a list-of-thunks style iterator. Also add {to,of}_seq to several standard modules. (Simon Cruanes, review by Alain Frisch and François Bobot)
  • * GPR#1010: pack all standard library modules into a single module Stdlib which is the default opened module (Stdlib itself includes Pervasives) to free up the global namespace for other standard libraries, while still allowing any OCaml standard library module to be referred to as Stdlib.Module). This is implemented efficiently using module aliases (prefixing all modules with Stdlib__, e.g. Stdlib__string). (Jérémie Dimino, David Allsopp and Florian Angeletti, review by David Allsopp and Gabriel Radanne)
  • GPR#1637: String.escaped is faster and does not allocate when called with a string that does not contain any characters needing to be escaped. (Alain Frisch, review by Xavier Leroy and Gabriel Scherer)
  • GPR#1638: add a Float module. (Nicolás Ojeda Bär, review by Alain Frisch and Jeremy Yallop)
  • GPR#1697: Tune [List.init] tailrec threshold so that it does not stack overflow when compiled with the Js_of_ocaml backend. (Hugo Heuzard, reviewed by Gabriel Scherer)

Other libraries:

  • MPR#7745, GPR#1629: Graphics.open_graph displays the correct window title on Windows again (fault introduced by 4.06 Unicode changes). (David Allsopp)
  • * GPR#1406: Unix.isatty now returns true in the native Windows ports when passed a file descriptor connected to a Cygwin PTY. In particular, compiler colors for the native Windows ports now work under Cygwin/MSYS2. (Nicolás Ojeda Bär, review by Gabriel Scherer, David Allsopp, Xavier Leroy)
  • GPR#1451: [getpwuid], [getgrgid], [getpwnam], [getgrnam] now raise Unix error instead of returning [Not_found] when interrupted by a signal. (Arseniy Alekseyev, review by Mark Shinwell and Xavier Leroy)
  • GPR#1477: raw_spacetime_lib can now be used in bytecode. (Nicolás Ojeda Bär, review by Mark Shinwell)
  • GPR#1533: (a) The implementation of Thread.yield for system thread now uses nanosleep(1) for enabling better preemption. (b) Thread.delay is now an alias for Unix.sleepf. (Jacques-Henri Jourdan, review by Xavier Leroy and David Allsopp)

Compiler user-interface and warnings:

  • MPR#7663, GPR#1694: print the whole cycle and add a reference to the manual in the unsafe recursive module evaluation error message. (Florian Angeletti, report by Matej Košík, review by Gabriel Scherer)
  • GPR#1166: In OCAMLPARAM, an alternative separator can be specified as first character (instead of comma) in the set ":|; ," (Fabrice Le Fessant)
  • GPR#1358: Fix usage warnings with no mli file (Leo White, review by Alain Frisch)
  • GPR#1428: give a non dummy location for warning 49 (no cmi found) (Valentin Gatien-Baron)
  • GPR#1491: Improve error reporting for ill-typed applicative functor types, F(M).t. (Valentin Gatien-Baron, review by Florian Angeletti and Gabriel Radanne)
  • GPR#1496: Refactor the code printing explanation for unification type errors, in order to avoid duplicating pattern matches (Armaël Guéneau, review by Florian Angeletti and Gabriel Scherer)
  • GPR#1505: Add specific error messages for unification errors involving functions of type "unit -> _" (Arthur Charguéraud and Armaël Guéneau, with help from Leo White, review by Florian Angeletti and Gabriel Radanne)
  • GPR#1510: Add specific explanation for unification errors caused by type constraints propagated by keywords (such as if, while, for…) (Armaël Guéneau and Gabriel Scherer, original design by Arthur Charguéraud, review by Frédéric Bour, Gabriel Radanne and Alain Frisch)
  • GPR#1515: honor the BUILD_PATH_PREFIX_MAP environment variable to enable reproducible builds (Gabriel Scherer, with help from Ximin Luo, review by Damien Doligez)
  • GPR#1534: Extend the warning printed when (*) is used, adding a hint to suggest using ( * ) instead (Armaël Guéneau, with help and review from Florian Angeletti and Gabriel Scherer)
  • GPR#1552, GPR#1577: do not warn about ambiguous variables in guards (warning 57) when the ambiguous values have been filtered by a previous clause (Gabriel Scherer and Thomas Refis, review by Luc Maranget)
  • GPR#1554: warnings 52 and 57: fix reference to manual detailed explanation (Florian Angeletti, review by Thomas Refis and Gabriel Scherer)
  • GPR#1618: add the -dno-unique-ids and -dunique-ids compiler flags (Sébastien Hinderer, review by Leo White and Damien Doligez)
  • GPR#1649 change compilation order of toplevel definitions, so that some warnings emitted by the bytecode compiler appear more in-order than before. (Luc Maranget, advice and review by Damien Doligez)
  • GPR#1806: add linscan to OCAMLPARAM options (Raja Boujbel)

Code generation and optimizations:

  • MPR#7630, GPR#1401: Faster compilation of large modules with Flambda. (Pierre Chambart, report by Emilio Jesús Gallego Arias, Pierre-Marie Pédrot and Paul Steckler, review by Gabriel Scherer and Leo White)
  • MPR#7630, GPR#1455: Disable CSE for the initialization function (Pierre Chambart, report by Emilio Jesús Gallego Arias, review by Gabriel Scherer and Xavier Leroy)
  • GPR#1370: Fix code duplication in Cmmgen (Vincent Laviron, with help from Pierre Chambart, reviews by Gabriel Scherer and Luc Maranget)
  • GPR#1486: ARM 32-bit port: add support for ARMv8 in 32-bit mode, a.k.a. AArch32. For this platform, avoid ITE conditional instruction blocks and use simpler IT blocks instead (Xavier Leroy, review by Mark Shinwell)
  • GPR#1487: Treat negated float comparisons more directly (Leo White, review by Xavier Leroy)
  • GPR#1573: emitcode: merge events after instructions reordering (Thomas Refis and Leo White, with help from David Allsopp, review by Frédéric Bour)
  • GPR#1606: Simplify the semantics of Lambda.free_variables and Lambda.subst, including some API changes in bytecomp/lambda.mli (Pierre Chambart, review by Gabriel Scherer)
  • GPR#1613: ensure that set-of-closures are processed first so that other entries in the let-rec symbol do not get dummy approximations (Leo White and Xavier Clerc, review by Pierre Chambart)
  • * GPR#1617: Make string/bytes distinguishable in the bytecode. (Hugo Heuzard, reviewed by Nicolás Ojeda Bär)
  • GPR#1627: Reduce cmx sizes by sharing variable names (Flambda only) (Fuyong Quah, Leo White, review by Xavier Clerc)
  • GPR#1665: reduce the size of cmx files in classic mode by droping the bodies of functions that will not be inlined (Fuyong Quah, review by Leo White and Pierre Chambart)
  • GPR#1666: reduce the size of cmx files in classic mode by droping the bodies of functions that cannot be reached from the module block (Fuyong Quah, review by Leo White and Pierre Chambart)
  • GPR#1686: Turn off by default flambda invariants checks. (Pierre Chambart)
  • GPR#1707: Add [Closure_origin.t] to trace inlined functions to prevent infinite loops from repeatedly inlining copies of the same function. (Fu Yong Quah)
  • GPR#1740: make sure startup.o is always linked in when using "-output-complete-obj". Previously, it was always linked in only on some platforms, making this option unusable on platforms where it wasn't (Jérémie Dimino, review by Sébastien Hinderer and Xavier Leroy)

Runtime system:

  • MPR#6411, GPR#1535: don't compile everything with -static-libgcc on mingw32, only dllbigarray.dll and libbigarray.a. Allows the use of C++ libraries which raise exceptions. (David Allsopp)
  • MPR#7100, GPR#1476: trigger a minor GC when custom blocks accumulate in minor heap (Alain Frisch, report by talex, review by Damien Doligez, Leo White, Gabriel Scherer)
  • GPR#1431: remove ocamlrun dependencies on curses/terminfo/termcap C library (Xavier Leroy, review by Daniel Bünzli)
  • GPR#1478: The Spacetime profiler now works under Windows (but it is not yet able to collect profiling information from C stubs). (Nicolás Ojeda Bär, review by Xavier Leroy, Mark Shinwell)
  • GPR#1483: fix GC freelist accounting for chunks larger than the maximum block size. (David Allsopp and Damien Doligez)
  • GPR#1526: install the debug and instrumented runtimes (lib{caml,asm}run{d,i}.a) (Gabriel Scherer, reminded by Julia Lawall)
  • GPR#1563: simplify implementation of LSRINT and ASRINT (Max Mouratov, review by Frédéric Bour)
  • GPR#1644: remove caml_alloc_float_array from the bytecode primitives list (it's a native code primitive) (David Allsopp)
  • GPR#1701: fix missing root bug in GPR#1476 (Mark Shinwell)
  • GPR#1752: do not alias function arguments to sigprocmask (Anil Madhavapeddy)
  • GPR#1753: avoid potential off-by-one overflow in debugger socket path length (Anil Madhavapeddy)


  • MPR#7643, GPR#1377: ocamldep, fix an exponential blowup in presence of nested structures and signatures (e.g. "include struct … include(struct … end) … end") (Florian Angeletti, review by Gabriel Scherer, report by Christophe Raffalli)
  • MPR#7687, GPR#1653: deprecate -thread option, which is equivalent to -I +threads. (Nicolás Ojeda Bär, report by Daniel Bünzli)
  • MPR#7710: ocamldep -sort should exit with nonzero code in case of cyclic dependencies (Xavier Leroy, report by Mantis user baileyparker)
  • GPR#1537: boot/ocamldep is no longer included in the source distribution; boot/ocamlc -depend can be used in its place. (Nicolás Ojeda Bär, review by Xavier Leroy and Damien Doligez)
  • GPR#1585: optimize output of "ocamllex -ml" (Alain Frisch, review by Frédéric Bour and Gabriel Scherer)
  • GPR#1667: add command-line options -no-propt, -no-version, -no-time, -no-breakpoint and -topdirs-path to ocamldebug (Sébastien Hinderer, review by Damien Doligez)
  • GPR#1695: add the -null-crc command-line option to ocamlobjinfo. (Sébastien Hinderer, review by David Allsopp and Gabriel Scherer)
  • GPR#1710: ocamldoc, improve the 'man' rendering of subscripts and superscripts. (Gabriel Scherer)
  • GPR#1771: ocamdebug, avoid out of bound access (Thomas Refis)

Manual and documentation:

  • MPR#7613: minor reword of the "refutation cases" paragraph (Florian Angeletti, review by Jacques Garrigue)
  • PR#7647, GPR#1384: emphasize ocaml.org website and forum in README (Yawar Amin, review by Gabriel Scherer)
  • PR#7698, GPR#1545: improve wording in OCaml manual in several places, mostly in Chapter 1. This addresses the easier changes suggested in the PR. (Jim Fehrle, review by Florian Angeletti and David Allsopp)
  • GPR#1540: manual, decouple verbatim and toplevel style in code examples (Florian Angeletti, review by Gabriel Scherer)
  • GPR#1556: manual, add a consistency test for manual references inside the compiler source code. (Florian Angeletti, review by Gabriel Scherer)
  • GPR#1647: manual, subsection on record and variant disambiguation (Florian Angeletti, review by Alain Frisch and Gabriel Scherer)
  • GPR#1702: manual, add a signature mode for code examples (Florian Angeletti, review by Gabriel Scherer)
  • GPR#1741: manual, improve typesetting and legibility in HTML output (steinuil, review by Gabriel Scherer)
  • GPR#1757: style the html manual, changing type and layout (Charles Chamberlain, review by Florian Angeletti, Xavier Leroy, Gabriel Radanne, Perry E. Metzger, and Gabriel Scherer)
  • GPR#1765: manual, ellipsis in code examples (Florian Angeletti, review and suggestion by Gabriel Scherer)
  • GPR#1767: change html manual to use relative font sizes (Charles Chamberlain, review by Daniel Bünzli, Perry E. Metzger, Josh Berdine, and Gabriel Scherer)
  • GPR#1779: integrate the Bigarray documentation into the main manual. (Perry E. Metzger, review by Florian Angeletti and Xavier Clerc)

Type system:

  • MPR#7611, GPR#1491: reject the use of generative functors as applicative (Valentin Gatien-Baron)
  • MPR#7706, GPR#1565: in recursive value declarations, track static size of locally-defined variables (Gabriel Scherer, review by Jeremy Yallop and Leo White, report by Leo White)
  • MPR#7717, GPR#1593: in recursive value declarations, don't treat unboxed constructor size as statically known (Jeremy Yallop, report by Pierre Chambart, review by Gabriel Scherer)
  • MPR#7767, GPR#1712: restore legacy treatment of partially-applied labeled functions in 'let rec' bindings. (Jeremy Yallop, report by Ivan Gotovchits, review by Gabriel Scherer)
  • * MPR#7787, GPR#1652, GPR#1743: Don't remove module aliases in module type of and with module. The old behaviour can be obtained using the [@remove_aliases] attribute. (Leo White and Thomas Refis, review by Jacques Garrigue)
  • GPR#1468: Do not enrich type_decls with incoherent manifests (Thomas Refis and Leo White, review by Jacques Garrigue)
  • GPR#1469: Use the information from [@@immediate] annotations when computing whether a type can be [@@unboxed] (Damien Doligez, report by Stephan Muenzel, review by Alain Frisch)
  • GPR#1513: Allow compilation units to shadow sub-modules of Pervasives. For instance users can now use a largeFile.ml file in their project. (Jérémie Dimino, review by Nicolas Ojeda Bar, Alain Frisch and Gabriel Radanne)
  • GPR#1516: Allow float array construction in recursive bindings when configured with -no-flat-float-array (Jeremy Yallop, report by Gabriel Scherer)
  • GPR#1583: propagate refined ty_arg to Parmatch checks (Thomas Refis, review by Jacques Garrigue)
  • GPR#1609: Changes to ambivalence scope tracking (Thomas Refis and Leo White, review by Jacques Garrigue)
  • GPR#1628: Treat reraise and raise_notrace as nonexpansive. (Leo White, review by Alain Frisch)
  • * GPR#1778: Fix Soundness bug with non-generalized type variable and local modules. This is the same bug as MPR#7414, but using local modules instead of non-local ones. (Leo White, review by Jacques Garrigue)

Compiler distribution build system

  • MPR#5219, GPR#1680, GPR#1877: use 'install' instead of 'cp' in install scripts. (Gabriel Scherer, review by Sébastien Hinderer and Valentin Gatien-Baron)
  • MPR#7679: make sure .a files are erased before calling ar rc, otherwise leftover .a files from an earlier compilation may contain unwanted modules (Xavier Leroy)
  • GPR#1571: do not perform architecture tests on 32-bit platforms, allowing 64-bit back-ends to use 64-bit specific constructs (Xavier Clerc, review by Damien Doligez)

Internal/compiler-libs changes:

  • MPR#7738, GPR#1624: Asmlink.reset also resets lib_ccobjs/ccopts (Cedric Cellier, review by Gabriel Scherer)
  • GPR#1488, GPR#1560: Refreshing parmatch (Gabriel Scherer and Thomas Refis, review by Luc Maranget)
  • GPR#1502: more command line options for expect tests (Florian Angeletti, review by Gabriel Scherer)
  • GPR#1511: show code at error location in expect-style tests, using new Location.show_code_at_location function (Gabriel Scherer and Armaël Guéneau, review by Valentin Gatien-Baron and Damien Doligez)
  • GPR#1519, GPR#1532, GRP#1570: migrate tests to ocamltest (Sébastien Hinderer, review by Gabriel Scherer, Valentin Gatien-Baron and Nicolás Ojeda Bär)
  • GPR#1520: more robust implementation of Misc.no_overflow_mul (Max Mouratov, review by Xavier Leroy)
  • GPR#1557: Organise and simplify translation of primitives (Leo White, review by François Bobot and Nicolás Ojeda Bär)
  • GPR#1567: register all idents relevant for reraise (Thomas Refis, review by Alain Frisch and Frédéric Bour)
  • GPR#1586: testsuite: 'make promote' for ocamltest tests (The new "-promote" option for ocamltest is experimental and subject to change/removal). (Gabriel Scherer)
  • GPR#1619: expect_test: print all the exceptions, even the unexpected ones (Thomas Refis, review by Jérémie Dimino)
  • GPR#1621: expect_test: make sure to not use the installed stdlib (Jérémie Dimino, review by Thomas Refis)
  • GPR#1646 : add ocamldoc test to ocamltest and migrate ocamldoc tests to ocamltest (Florian Angeletti, review by Sébastien Hinderer)
  • GPR#1663: refactor flambda specialise/inlining handling (Leo White and Xavier Clerc, review by Pierre Chambart)
  • GPR#1679 : remove Pbittest from primitives in lambda (Hugo Heuzard, review by Mark Shinwell)
  • * GPR#1704: Make Ident.t abstract and immutable. (Gabriel Radanne, review by Mark Shinwell)

Bug fixes

  • MPR#4499, GPR#1479: Use native Windows API to implement Sys.getenv, Unix.getenv and Unix.environment under Windows. (Nicolás Ojeda Bär, report by Alain Frisch, review by David Allsopp, Xavier Leroy)
  • MPR#5250, GPR#1435: on Cygwin, when ocamlrun searches the path for a bytecode executable file, skip directories and other non-regular files, like other Unix variants do. (Xavier Leroy)
  • MPR#6394, GPR#1425: fix fatal_error from Parmatch.get_type_path (Virgile Prevosto, review by David Allsopp, Thomas Refis and Jacques Garrigue)
  • * MPR#6604, GPR#931: Only allow directives with filename and at the beginning of the line (Tadeu Zagallo, report by Roberto Di Cosmo, review by Hongbo Zhang, David Allsopp, Gabriel Scherer, Xavier Leroy)
  • MPR#7138, MPR#7701, GPR#1693: Keep documentation comments even in empty structures and signatures (Leo White, Florian Angeletti, report by Anton Bachin)
  • MPR#7178, MPR#7253, MPR#7796, GPR#1790: Make sure a function registered with "at_exit" is executed only once when the program exits (Nicolás Ojeda Bär and Xavier Leroy, review by Max Mouratov)
  • MPR#7391, GPR#1620: Do not put a dummy method in object types (Thomas Refis, review by Jacques Garrigue)
  • PR#7660, GPR#1445: Use native Windows API to implement Unix.utimes in order to avoid unintended shifts of the argument timestamp depending on DST setting. (Nicolás Ojeda Bär, review by David Allsopp, Xavier Leroy)
  • MPR#7668: -principal is broken with polymorphic variants (Jacques Garrigue, report by Jun Furuse)
  • MPR#7680, GPR#1497: Incorrect interaction between Matching.for_let and Simplif.simplify_exits (Alain Frisch, report and review by Vincent Laviron)
  • MPR#7682, GPR#1495: fix [@@unboxed] for records with 1 polymorphic field (Alain Frisch, report by Stéphane Graham-Lengrand, review by Gabriel Scherer)
  • MPR#7695, GPR#1541: Fatal error: exception Ctype.Unify(_) with field override (Jacques Garrigue, report by Nicolás Ojeda Bär)
  • MPR#7704, GPR#1564: use proper variant tag in non-exhaustiveness warning (Jacques Garrigue, report by Thomas Refis)
  • MPR#7711, GPR#1581: Internal typechecker error triggered by a constraint on self type in a class type (Jacques Garrigue, report and review by Florian Angeletti)
  • MPR#7712, GPR#1576: assertion failure with type abbreviations (Thomas Refis, report by Michael O'Connor, review by Jacques Garrigue)
  • MPR#7747: Type checker can loop infinitly and consumes all computer memory (Jacques Garrigue, report by kantian)
  • MPR#7751, GPR#1657: The toplevel prints some concrete types as abstract (Jacques Garrigue, report by Matej Kosik)
  • MPR#7765, GPR#1718: When unmarshaling bigarrays, protect against integer overflows in size computations (Xavier Leroy, report by Maximilian Tschirschnitz, review by Gabriel Scherer)
  • MPR#7760, GPR#1713: Exact selection of lexing engine, that is correct "Segfault in ocamllex-generated code using 'shortest'" (Luc Maranget, Frédéric Bour, report by Stephen Dolan, review by Gabriel Scherer)
  • MPR#7769, GPR#1714: calls to Stream.junk could, under some conditions, be ignored when used on streams based on input channels. (Nicolás Ojeda Bär, report by Michael Perin, review by Gabriel Scherer)
  • MPR#7793, GPR#1766: the toplevel #use directive now accepts sequences of ';;' tokens. This fixes a bug in which certain files accepted by the compiler were rejected by ocamldep. (Nicolás Ojeda Bär, report by Hugo Heuzard, review by Hugo Heuzard)
  • GPR#1517: More robust handling of type variables in mcomp (Leo White and Thomas Refis, review by Jacques Garrigue)
  • GPR#1530, GPR#1574: testsuite, fix 'make parallel' and 'make one DIR=…' to work on ocamltest-based tests. (Runhang Li and Sébastien Hinderer, review by Gabriel Scherer)
  • GPR#1550, GPR#1555: Make pattern matching warnings more robust to ill-typed columns (Thomas Refis, with help from Gabriel Scherer and Luc Maranget)
  • GPR#1614: consider all bound variables when inlining, fixing a compiler fatal error. (Xavier Clerc, review by Pierre Chambart, Leo White)
  • GPR#1622: fix bug in the expansion of command-line arguments under Windows which could result in some elements of Sys.argv being truncated in some cases. (Nicolás Ojeda Bär, review by Sébastien Hinderer)
  • GPR#1623: Segfault on Windows 64 bits when expanding wildcards in arguments. (Marc Lasson, review by David Allsopp, Alain Frisch, Sébastien Hinderer, Xavier Leroy, Nicolas Ojeda Bar)
  • GPR#1661: more precise principality warning regarding record fields disambiguation (Thomas Refis, review by Leo White)
  • GPR#1687: fix bug in the printing of short functor types "(S1 -> S2) -> S3" (Pieter Goetschalckx, review by Gabriel Scherer)
  • GPR#1722: Scrape types in Typeopt.maybe_pointer (Leo White, review by Thomas Refis)
  • GPR#1755: ensure that a bigarray is never collected while reading complex values (Xavier Clerc, Mark Shinwell and Leo White, report by Chris Hardin, reviews by Stephen Dolan and Xavier Leroy)
  • GPR#1764: in byterun/memory.c, struct pool_block, use C99 flexible arrays if available (Xavier Leroy, review by Max Mouratov)
  • GPR#1774: ocamlopt for ARM could generate VFP loads and stores with bad offsets, rejected by the assembler. (Xavier Leroy, review by Mark Shinwell)
  • GPR#1808: handle [@inlined] attributes under a module constraint (Xavier Clerc, review by Leo White)
  • GPR#1810: use bit-pattern comparison when meeting float approximations (Xavier Clerc, report by Christophe Troestler, review by Nicolás Ojeda Bär and Gabriel Scherer)
  • GPR#1835: Fix off-by-one errors in Weak.get_copy and Weak.blit (KC Sivaramakrishnan)
  • GPR#1849: bug in runtime function generic_final_minor_update() that could lead to crashes when Gc.finalise_last is used (report and fix by Yuriy Vostrikov, review by François Bobot)

Dune 1.0.0

Jérémie Dimino announced

It is my pleasure to announce the 1.0.0 release of Dune!

Dune is a built system for OCaml and Reason projects. It aims at building things fast and providing a smooth user experience.

Dune 1.0.0 is now available on github and in opam.

This is a big milestone for the project as it is the first release using the new Dune name that the community helped to choose. Dune started about a year and a half ago as the more narrow scoped Jbuilder and as since then proven itself to be the right tool for the job.

A big part of the work for this release was preparing the project for long-term support. Dune is the kind of project that will pretty much continue to evolve forever: it needs to keep up with the OCaml compiler evolving, development workflows changing, new kinds of tools with specific needs, etc… While doing so, it also needs to offer stability so that projects released now will continue to build with future versions of Dune.

Meeting these two constraints is a challenge. However, we had to do this for the renaming: we changed many things in Dune using what we learned from the Jbuilder experience. Still, Dune is able to understand and build Jbuilder projects. The mechanism used to support both Jbuilder and Dune projects is fully generic and will be the basis for future evolution.

You can find the full list of changes and bug fixes since Jbuilder 1.0+beta20 on the github release page. Following is an overview of some the new features:

  • General cleanup of the syntax: less parentheses, better behaved variables
  • Introduction of a dune.configurator library to help projects with C stubs detect and query the environment
  • General improvement of the default behavior: more parallelism by default, development mode by default
  • Better support for changing the defaults: it is now possible to set the default flags or change what dune build does by default
  • Support for setting the build directory
  • Better support for expectation tests on Windows

Release of Stdcompat 5

Thierry Martinez announced

It is my pleasure to announce a new release (#5) of Stdcompat!

Stdcompat is a compatibility module for OCaml standard library, available from OCaml 3.07 to OCaml 4.07.0 (that is to say, the whole spectrum of OCaml compilers available in OPAM).

Stdcompat.* redefines (most of the) symbols available in OCaml 4.07.0 standard library, such as List.find_opt, floatarray, Uchar.t, Seq.t, result, etc.

Stdcompat has no dependency! Former versions used to rely on cppo but the preprocessing is now done by a ./configure script (generated with autoconf).

OPAM packages result, seq, uchar are optional dependencies: if they are installed, types declared by stdcompat are compatible with the types these packages declare.

Stdcompat can be opened in the beginning of a module: its signature follows the signature of the standard library, including the new Stdlib module.

Github repository: https://github.com/thierry-martinez/stdcompat Available in OPAM: opam install stdcompat

Happy hacking!

JS_of_OCaml vs Bucklescript

Volodymyr Melnyk asked

Currently we have 2 options to write frontend-related code in OCaml: js_of_ocaml and bucklescript. I have no experience with this two libraries and I want to pick one. Has anybody tried both of them or made some analysis with comparison? Could you please share your thoughts and experience about both libraries if so. I'm interested in: performance of generation, source-mapping, performance, quality, safety and readability of generated JS, simplicity of use, pros and cons of both solutions, etc. Thank you very much for your responses. OCaml community is the best!

Yaron Minsky replied

We've made extensive use of js_of_ocaml for internal apps at Jane Street. I can't give a detailed comparison with Bucklescript, but I can tell you what I know of the tradeoffs.

  • First, js_of_ocaml runs pretty fast, but I've heard tell that Bucklescript is faster. js_of_ocaml now supports separate compilation of Javascript, so subsequent recompilations are quite zippy, in my experience. That said, the initial compilation takes material time. Dune does separate compilation for js_of_ocaml by default, and does a single, more compact javascript executable when run in production mode. Anyway, we haven't found performance of the compiler to be an obstacle with js_of_ocaml.
  • js_of_ocaml is highly compatible with OCaml's semantics. Advanced libraries like Async and Incremental that make fairly aggressive use of OCaml's memory model work under js_of_ocaml without modification, which is great. I believe you have to be a bit more careful when compiling with Bucklescript. (See incr_dom for an interesting application of Incremental to the browser.)
  • js_of_ocaml is highly compatible in a another way: it is essentially always fully up to date with the latest OCaml. That's because it's easier to maintain, by virtue of operating only on OCaml bytecode. Bucklescript is a more fullsome set of patches to the compiler, and so it typically lags a few versions behind. That alone is for us a sufficiently compelling reason to stick to js_of_ocaml.
  • Bucklescript seems to have a more active web-dev community, associated with the Reason community. I think this is mostly because Bucklescript generates easy to read javascript output. I don't care much about readable JavaScript output (especially in a world with sourcemaps), but the community surely has value. For example, the OCaml React bindings are currently Bucklescript-only (though @jordwalke has suggested that porting to js_of_ocaml would be totally doable.)

Anyway, good luck in picking! I think people have had good experiences with both, so you can't go too far wrong.

Yaron Minsky later added

Oh, and for people who are concerned about bindings, this project from LexiFi looks pretty cool. I think we haven't started using it yet, but some folk internally are currently kicking the tires. It looks pretty cool to me.


Sundials/ML 3.1.1

Timothy Bourke announced

We are pleased to announce a new release of Sundials/ML, an OCaml interface to the Sundials suite of numerical solvers (CVODE, CVODES, IDA, IDAS, ARKODE, KINSOL).

This release adds support for Sundials 3.1.x, and most notably for the new generic matrix and linear solver interfaces. The OCaml interface changes but the library is backward compatible with the older versions of Sundials (2.5.0 onwards) that are still current in some packaging systems.

OCaml 4.02.3 or greater is now required and optionally OCamlMPI 1.03.

Information and documentation: http://inria-parkas.github.io/sundialsml/
Source code (BSD): https://github.com/inria-parkas/sundialsml

opam install sundialsml # (requires Sundials 2.5.0 or above)

There is a draft paper that describes our implementation. It is still undergoing peer review, but we would be happy to share it by private mail.

We gratefully acknowledge the original authors of Sundials, and the support of the ITEA 3 project 11004 MODRIO (Model driven physical systems operation), Inria, and the Departement d'Informatique de l'ENS.

Other OCaml News

From the ocamlcore planet blog

Here are links from many OCaml blogs aggregated at OCaml Planet.


If you happen to miss a CWN, you can send me a message and I'll mail it to you, or go take a look at the archive or the RSS feed of the archives.

If you also wish to receive it every week by mail, you may subscribe online.