Skip to content
This repository has been archived by the owner on Jul 26, 2024. It is now read-only.

Latest commit

 

History

History
623 lines (468 loc) · 23.3 KB

README.md

File metadata and controls

623 lines (468 loc) · 23.3 KB

Warning

This repository has been moved to Codeberg: OCaml-Buck-2-Examples

OCaml-Buck-2-Examples

Test

This contains documentation and examples on how to use Buck 2 to build OCaml projects.

Installation

You need the following:

Install Buck 2 like documented at Buck 2 - Getting Started.

Copy the scripts meta2json.py, rules.py and dromedary.py from the ocaml-scripts repository to some place you can use it in your project(s), no "real" installation needed. Warning: all three must be in the same directory.

This Repository

To actually clone the Buck 2 prelude into the prelude subdirectories of this repository, you have to update them:

  • Run git submodule update --init in the root directory ./.

Optional Tools and Extensions

Buck 2 has an integrated (Starlark) LSP, which can be started using buck2 lsp in the root directory of a Buck 2 project. To use this to help with writing BUCK (Starlark) files, you need a plugin for your editor of choice:

Buck 2 Usage with OCaml

To generate a new OCaml project with Buck 2:

  1. Generate the Buck 2 project structure using buck2 init
  2. Generate a new Opam switch (a sandbox) and a Buck 2 configuration file for the Opam packages, using dromedary.py and a dromedary configuration file from Meta's ocaml-scripts.
  3. Edit the Buck 2 configuration files for the build targets

To use Buck 2 for an existing OCaml project:

  1. Generate the Buck 2 project structure using buck2 init
  2. Generate a Buck 2 configuration file for the Opam packages of an existing Opam switch (sandbox), using dromedary.py from Meta's ocaml-scripts.
  3. Edit the Buck 2 configuration files for the build targets

Initialize a Buck 2 Project

The following steps are needed to setup a Buck 2 project. These are the same for new and already existing OCaml projects.

  • buck2 init --git to generate all needed files and directories and add the Buck 2 prelude directory as a git submodule, so you are able to update it using git. The Prelude of Buck 2 contains the build rules for every language, for example the ones for OCaml.

  • Building the generated default target using buck2 build //.. should work without errors now:

    % buck2 build //...
    
    Build ID: a125ba60-835a-4afb-813f-cede508c833f
    Jobs completed: 6. Time elapsed: 0.0s.
    Cache hits: 0%. Commands: 1 (cached: 0, remote: 0, local: 1)
    BUILD SUCCEEDED
    
  • Set the execution platform in ./.buckconfig to the default one of the prelude.

    [build]
    execution_platforms = prelude//platforms:default
  • Add the needed toolchains to ./toolchains/BUCK:. edit the file ./toolchains/BUCK as follows, to include the Python and C++ toolchains, which are always needed, and the OCaml toolchain:

load("@prelude//toolchains:cxx.bzl", "system_cxx_toolchain")
load("@prelude//toolchains:genrule.bzl", "system_genrule_toolchain")
load("@prelude//toolchains:ocaml.bzl", "system_ocaml_toolchain")
load("@prelude//toolchains:python.bzl", "system_python_bootstrap_toolchain")

system_cxx_toolchain(
    name = "cxx",
    visibility = ["PUBLIC"],
)

system_genrule_toolchain(
    name = "genrule",
    visibility = ["PUBLIC"],
)

system_ocaml_toolchain(
    name = "ocaml",
    visibility = ["PUBLIC"],
)

system_python_bootstrap_toolchain(
    name = "python_bootstrap",
    visibility = ["PUBLIC"],
)
  • Generate a subdirectory ./third-party in the prohect root directory to hold the information about third-party packages. The name can be any you like, but third-party is the canonical one, that's also used in all examples. If you've chosen another one, substitute third-party with your directory name in all of the documentation.

Generate a New Opam Switch and Install Packages

To generate a new switch and install Opam packages in this switch, do the following:

  • Edit the JSON configuration file ./third-party/dromedary.json.

  • Fields, these are the arguments you would pass to opam create:

    • name, a string: the path to the Opam sandbox or the name of the global Opam switch to create.
    • compiler, a string: the name of the compiler to use.
    • packages, a list of strings: the name and optionally versions of the packages to install in the Opam switch.
  • The only mandatory field is packages, the default for name is ./ and the default for compiler is ocaml-variants.

  • Example file ./third-party/dromedary.json:

    {
        "name": "./",
        "compiler": "ocaml-variants",
        "packages": [
            "menhirLib",
            "sedlex=3.2",
            "alcotest>=1.7.0"
        ]
    }

    results in the following opam command in the project root: opam switch create ./third-party/ ocaml-variants 'menhirLib' 'sedlex=3.2' 'alcotest>=1.7.0'

  • Run python3 dromedary.py -o ./third-party/BUCK ./third-party/dromedary.json. Instead of dromedary.py you need to use the actual path to the Python script. Substitute third-party with the name you have chosen for your subdirectory.

  • The file ./third-party/BUCK should now have been created.

  • The symlink ./third-party/opam should link to the generated Opam switch.

  • Running buck2 targets //third-party: (or the name of your subdirectory instead of third-party) should now list all installed Opam packages of the switch.

% buck2 targets //third-party:
Build ID: ed8ad382-6848-44c7-842c-a537fa916411
Jobs completed: 4. Time elapsed: 0.3s.
root//third-party:alcotest
root//third-party:alcotest.alcotest-plugin
root//third-party:alcotest.engine
root//third-party:alcotest.engine.alcotest_engine-plugin
root//third-party:alcotest.runtime.js
root//third-party:alcotest.stdlib_ext
root//third-party:alcotest.stdlib_ext.alcotest_stdlib_ext-plugin
root//third-party:astring
root//third-party:astring.astring-plugin
root//third-party:astring.top
root//third-party:astring.top.astring_top-plugin
root//third-party:base
root//third-party:base.base-plugin
...

Use the Packages of an Existing Opam Switch

To use all opam packages of an existing Opam switch (or sandbox) with name or directory OPAM_SWITCH, do the following:

  • Run python3 dromedary.py --switch OPAM_SWITCH -o ./third-party/BUCK. Instead of dromedary.py you need to use the actual path to the Python script. Substitute third-party with the name you have chosen for your subdirectory.
  • The file ./third-party/BUCK should now have been created.
  • The symlink ./third-party/opam should link to the path of the Opam switch.
  • Running buck2 targets //third-party: (or the name of your subdirectory instead of third-party) should now list all installed Opam packages of the switch.
% buck2 targets //third-party:
Build ID: ed8ad382-6848-44c7-842c-a537fa916411
Jobs completed: 4. Time elapsed: 0.3s.
root//third-party:alcotest
root//third-party:alcotest.alcotest-plugin
root//third-party:alcotest.engine
root//third-party:alcotest.engine.alcotest_engine-plugin
root//third-party:alcotest.runtime.js
root//third-party:alcotest.stdlib_ext
root//third-party:alcotest.stdlib_ext.alcotest_stdlib_ext-plugin
root//third-party:astring
root//third-party:astring.astring-plugin
root//third-party:astring.top
root//third-party:astring.top.astring_top-plugin
root//third-party:base
root//third-party:base.base-plugin
...

Buck 2 Target Configuration

Libraries

File ./lib/BUCK

This is the easiest case, just compile all ml files into a library named my_project. deps would hold library names the library depends on.

ocaml_library(
    name = "my_project",
    srcs = glob(["./*.ml"]),
    deps = [],
    visibility = ["PUBLIC"],
) if not host_info().os.is_windows else None

But this library does not contain the files in a single module. For example the module Bar of the file bar.ml is named Bar, not My_project.Bar.

To have all the files of the library contained in a parent module My_project, you need to do a mapping:

File ./lib/my_project.mli

(* for file foo.ml *)
module Foo = Foo

(* for file bar.ml *)
module Bar = Bar

File ./lib/my_project.ml

(* for file foo.ml *)
module Foo = Foo

(* for file bar.ml *)
module Bar = Bar

File ./lib/BUCK

Give a name to the file my_project.mli, so we can use that in other rules.

export_file(
    name = "my_project.mli",
    src = "my_project.mli",
    visibility = [
        ":my_project",
        ":my_project__",
    ],
)

Add the mapping as an extra target.

# buildifier: disable=no-effect
ocaml_library(
    name = "my_project__",
    srcs = [
        "my_project.ml",
        ":my_project.mli",
    ],
    compiler_flags = [
        "-no-alias-deps",
    ],
    visibility = [":my_project"],
) if not host_info().os.is_windows else None

Compiler the library, using the mapping. ocamldep_flags are flags to be passed to ocamldep.

# buildifier: disable=no-effect
ocaml_library(
    name = "my_project",
    srcs = ["./foo.ml", "./bar.ml"],
    compiler_flags = [
        "-no-alias-deps",
        "-open",
        "My_project",
    ],
    ocamldep_flags = [
        "-open",
        "My_project",
        "-map",
        "$(location :my_project.mli)",
    ],
    deps = [":my_project__"],
    visibility = ["PUBLIC"],
) if not host_info().os.is_windows else None

Executables

File ./bin/BUCK

This binary named my_project is contained in one source file, ./main.ml and uses the "internal" library my_project located in the subdirectory lib of the project root.

ocaml_binary(
    name = "my_project",
    srcs = ["./main.ml"],
    deps = [
        "//lib:my_project",
    ],
    visibility = ["PUBLIC"],
) if not host_info().os.is_windows else None

File ./test/BUCK

This test binary named my_project is contained in one source file, ./my_project_test.ml and uses the "internal" library my_project located in the subdirectory lib of the project root and the libraries alcotest and qcheck-alcotest from the third-party subdirectory.

The binary is missing some libraries when linking, we have to pass -cclib -lunixbyt -cclib -lunixnat to the OCaml compiler.

ocaml_binary(
    name = "my_project",
    srcs = ["./my_project_test.ml"],
    deps = [
        "//lib:my_project",
        "//third-party:alcotest",
        "//third-party:qcheck-alcotest"
    ],
    compiler_flags = [
        "-cclib",
        "-lunixbyt",
        "-cclib",
        "-lunixnat",
    ],
    visibility = ["PUBLIC"],
) if not host_info().os.is_windows else None

PPX

To compile a library using a PPX library, we have to first generate an executable from the PPX. Than use this executable to parse the the source file and generate the actual OCaml source file to be compiled by the OCaml compiler.

File ./lib/driver.ml

This file (yes, that is all) is compiled to an executable, that later parses the source files containing the PPX syntax.

let () = Ppxlib.Driver.standalone ()

File ./lib/BUCK

This rule generates the executable named ppx from the ./lib/driver.ml source file and the PPX libraries.

ocaml_binary(
    name = "ppx",
    srcs = ["./driver.ml"],
    compiler_flags = [
        "-linkall",
    ],
    deps = [
        "//third-party:sedlex",
        "//third-party:sedlex.ppx",
    ],
) if not host_info().os.is_windows else None

File ./lib/BUCK

Now we use the PPX executable named ppx from above to parse the PPX syntax. It is passed to the OCaml compiler using -ppx $(exe_target :ppx) --as-ppx.

ocaml_library(
    name = "my_project",
    srcs = ["./ast.ml", "./interpreter.ml", "./lexer_ex.ml"],
    compiler_flags = [
        "-no-alias-deps",
        "-ppx",
        "$(exe_target :ppx) --as-ppx",
    ],
    deps = ["//third-party:sedlex",
            "//third-party:sedlex.ppx"],
    visibility = ["PUBLIC"],
) if not host_info().os.is_windows else None

Inline Tests (Runners)

To be able to use inline tests, we have to also compile a test runner executable for these to be called. This needs an extra ocaml_binary target to run the inline tests.

Alcotest Inline Runner for a Library

File ./lib/BUCK

We have to generate an inline test runner executable and link all libraries containing inline tests and the Alcotest inline runner library ppx_inline_alcotest.runner with it.

ocaml_binary(
    name = "inline_runner",
    srcs = ["./inline_runner.ml"],
    compiler_flags = [
        "-linkall",
        "-cclib",
        "-lunixbyt",
        "-cclib",
        "-lunixnat",
    ],
    deps = [":inline_test_runners",
            "//third-party:ppx_inline_alcotest.runner",],
    visibility = ["PUBLIC"],
) if not host_info().os.is_windows else None

The processing of the library's sources works the same as above, with a PPX driver executable :ppx.

File ./lib/inline_runner.ml

The test runner executable's source.

let () = Ppx_inline_alcotest_runner.run ()
Alcotest Inline Runner for a Test Executable

File ./test/BUCK

We have to generate an inline test runner executable and link all libraries containing inline tests and the Alcotest inline runner library ppx_inline_alcotest.runner with it.

ocaml_binary( name = "ppx", srcs = ["./driver.ml"], compiler_flags = [ "-linkall", ], deps = ["//third-party:ppx_inline_alcotest", ], ) if not host_info().os.is_windows else None

File ./test/inline_tests.ml

The test runner executable's source, add the stanza let () = Ppx_inline_alcotest_runner.run () to run the inline tests.

(* ... *)

let%test "1 is 1" = Alcotest.(check int) "same ints" 1 1

let () = Ppx_inline_alcotest_runner.run ()

The processing of the executable's sources works the same as above, with a PPX driver executable :ppx and by adding the PPX driver to the test executable's config:

ocaml_binary(
    name = "inline_test_runner",
    srcs = [YOUR_SOURCES],
    deps = [
        "//lib:LIBRARY_TO_TEST",
        "//third-party:ppx_inline_alcotest",
        "//third-party:ppx_inline_alcotest.runner",
        "//third-party:alcotest"
    ],
    compiler_flags = [
        "-cclib",
        "-lunixbyt",
        "-cclib",
        "-lunixnat",
         "-ppx",
        "$(exe_target :ppx) --as-ppx",
    ],
    visibility = ["PUBLIC"],
) if not host_info().os.is_windows else None

Aliases

You can use aliases for targets, for example in subdirectories.

If you have 3 subdirectories, bin, lib and test, and the target in each subdirectory has the name my_project defined in its BUCK file, you can define the following aliases:

File ./BUCK in the project root defines the aliases:

alias(
    name="bin",
    actual="//bin:my_project",
    visibility=["PUBLIC"],
)

alias(
    name="lib",
    actual="//lib:my_project",
    visibility=["PUBLIC"],
)

alias(
    name="test",
    actual="//test:my_project",
    visibility=["PUBLIC"],
)

So, instead of buck2 run //test:my_project you can now use buck2 run //:test

See Examples.

Using Buck 2 with OCaml-LSP or Merlin

You need to generate a .merlin file containing the paths to the source and build directories and the list of Opam packages (but using ocamlfind names!), see Merlin - Project Configuration

Example:

S bin/**
S lib/**
S test/**

B ./buck-out/v2/gen/root/*/bin/__ppx_usage_example__/_nativeobj_
B ./buck-out/v2/gen/root/*/lib/__ppx__/_nativeobj_
B ./buck-out/v2/gen/root/*/lib/__ppx_usage_example__/_nativeobj_
B ./buck-out/v2/gen/root/*/lib/__ppx_usage_example__/_nativeobj_/_native_gen_
B ./buck-out/v2/gen/root/*/lib/__ppx_usage_example____/_nativeobj_
B ./buck-out/v2/gen/root/*/test/__ppx_usage_example__/_nativeobj_

PKG sedlex menhirLib qcheck-alcotest alcotest sedlex.ppx

You can generate the B stanzas using find, the cmi files are located at _nativeobj_ directories in buck-out.

dirname $(find ./buck-out -name "*.cmi") | sort | uniq

To be able to use the .merlin file with OCaml-LSP, you need to add the command line argument --fallback-read-dot-merlin to the ocaml-lsp invocation and need the Opam package dot-merlin-reader installed. OCaml-LSP documentation: Merlin configuration

For VS Code or Codium the setting to add the command line argument is

    "ocaml.server.args": [
        "--fallback-read-dot-merlin"
    ],

Examples

Each example project directory contains a README.md file explaining its Buck 2 configuration files.

  • ./basic_example/ - a OCaml project which uses Opam package dependencies, but no PPX. Contains a library, an executable that uses the library and tests of the library.
  • ./ocamllex_menhir_example/ - a OCaml project which uses OCamlLex and Menhir to generate OCaml code. Contains a library, an executable that uses the library and tests of the library.
  • ./ppx_usage_example/ - a OCaml project which uses Sedlex and Menhir as code generators and PPX for Sedlex to show the usage of PPX libraries. Contains a library, an executable that uses the library and tests of the library.
  • ./inline_test_runners - a OCaml project which uses PPX inline tests, the library uses Alcotest inline PPX. Contains a library, an executable that uses the library and tests and inline tests of the library.

All of these examples also use Dune configuration files, so you can compare them to the BUCK files.

GitHub Actions

In the file ./.github/workflows/test.yml there are working Github actions for MacOS and Linux. The Windows part does not work, as passing the PATH into dromedary.py (which calls other Python scripts, which call ocamlfind) fails.

Other Buck 2 with OCaml Resources

Questions and Answers

Does Buck2 support OCaml on Windows?

No. Buck2 itself works on Windows, but OCaml's build rules do not support Windows so far. But you can change this: see OCaml rules in Buck 2 Prelude at GitHub.

Can I use Dune (or any other build system) too?

The short answer is "yes". The longer answer is: yes, but you need to manually add changes to dune files to BUCK files and vice versa. There does not exist a script (so far) that does this work for you.

Why Buck 2 and not Dune (or other build systems)?

If you have to ask, you may not need it ;). It's mostly useful if you use more than one language (more than just OCaml) that you want to build.

Why Buck 2 and not Bazel?

Buck2 has OCaml support from Meta itself, not a third party and it is written in Rust, which I use myself, so there is no additional dependency for me on a JVM (which Bazel uses). And, last but not least, Bazel is made by Google, which sometimes (ahem) abruptly stops supporting some of their projects. But Bazel has more mature support as Buck 2 is much younger and less used.

There is Bazel support for OCaml by OBazl, documentation: The OBazl Book. I have never tried Bazel (or the OCaml support), so I cannot say anything about it. Try it out for yourself.

Error Message "inconsistent assumptions over implementation"

If you are getting an error message like

Error: Files third-party/opam/lib/uutf/uutf.cmxa
       and /Users/roland/.opam/5.0.0/lib/ocaml/stdlib.cmxa
       make inconsistent assumptions over implementation Stdlib__String

The Opam environment of your shell and the one used by Buck 2 may be off. Set the Opam environment to the one of the right switch using opam switch BUCK2_SWICH_NAME and call opam env afterwards. The output of opam switch BUCK2_SWICH_NAME tells you how. Then clean your project: buck2 clean and rebuild everything: buck2 build //...

Buck 2 Examples for other Languages

I've also made example projects using C++ with vcpkg as package manager: Cxx-Buck2-vcpkg-Examples and with Conan (the C++ package manager) at Cxx-Buck2-Conan-Examples.

Contributions

If you want to add tips or tricks on using Buck 2, examples, a link to other examples, blog or forum posts, or found an error, please open an issue or pull request with your changes.

License

All files in this repository are licensed under the MIT license, see file ./LICENSE.