Elixir v1.18.0 released

Elixir

Elixir Core Team

Full announcement: Elixir v1.18 released: type checking of calls, LSP listeners, built-in JSON, and more - The Elixir programming language

Release notes: Release v1.18.0 · elixir-lang/elixir · GitHub

Changelog

1. Enhancements

Elixir

[CLI] Add experimental PowerShell scripts for elixir, elixirc, and mix on Windows. Those provide a safer entry point for running Elixir from other platforms
[Calendar] Add Duration.to_string/1
[Code] Support several migration options in Code.format_string!/2
[Code] Add parenthesis around -- and --- in Code.format_string!/2 to make precedence clearer
[Code] Include more metadata in Code.string_to_quoted/2 when token_metadata: true to help compute ranges from the AST
[Code.Fragment] Have :capture_arg as its own entry in Code.Fragment.surround_context/2
[Config] Add Config.read_config/1
[Enumerable] Add Enum.product_by/2 and Enum.sum_by/2
[Exception] Add MissingApplicationsError exception to denote missing applications
[JSON] Add a new JSON module with encoding and decoding functionality
[JSON] Implement JSON.Encoder for all Calendar types
[Kernel] Update source code parsing to match UTS #55 latest recommendations. In particular, mixed script is allowed in identifiers as long as they are separate by underscores (_), such as http_сервер. Previously allowed highly restrictive identifiers, which mixed Latin and other scripts, such as the japanese word for t-shirt, Tシャツ, now require the underscore as well
[Kernel] Warn on bidirectional confusability in identifiers
[Kernel] Verify the type of the binary generators
[Kernel] Track the type of tuples in patterns and inside elem/2
[Kernel] Perform validation of root AST nodes in unquote and unquote_splicing to catch bugs earlier
[Kernel] Add source, behaviour, and record information to Docs chunk metadata
[Kernel] Support deterministic builds in tandem with Erlang by setting ERL_COMPILER_OPTIONS=deterministic. Keep in mind deterministic builds strip source and other compile time information, which may be relevant for programs
[Kernel] Allow aliases and imports to be enabled conditionally in module body
[List] Add List.ends_with?/2
[Macro] Improve dbg handling of if/2, with/1 and of code blocks
[Macro] Add Macro.struct_info!/2 to return struct information mirroring mod.__info__(:struct)
[Registry] Add Registry.lock/3 for local locking
[PartitionSupervisor] Add PartitionSupervisor.resize!/2 to resize the number of partitions in a supervisor (up to the limit it was started with)
[Process] Handle arbitrarily high integer values in Process.sleep/1
[Protocol] Add @undefined_impl_description to customize error message when an implementation is undefined
[Protocol] Add __deriving__/1 as optional macro callback to Protocol, no longer requiring empty implementations
[String] Inspect special whitespace and zero-width characters using their Unicode representation
[String] Update Unicode to 16.0

ExUnit

[ExUnit] Support parameterized tests on ExUnit.Case
[ExUnit] Support test groups: tests in the same group never run concurrently
[ExUnit.Case] Add test_pid as a tag

IEx

[IEx] Add IEx.configure(auto_reload: true) to automatically pick up modules recompiled from other operating system processes
[IEx] Add :dot_iex support to IEx.configure/1
[IEx] Add report for normal/shutdown exits in IEx

Mix

[mix compile] Ensure only a single operating system process can compile at a given time
[mix deps.get] Ensure only a single operating system process can fetch deps at a given time
[mix format] Add mix format --migrate to migrate from deprecated functionality
[mix format] Add new options and metadata to improve formatting applying by editors and other environments
[mix test] Taint failure manifest if requiring or compiling tests fail
[Mix.Project] Add a :listeners configuration to listen to compilation events from the current and other operating system processes
[Mix.Task.Compiler] Add API for fetching all persisted compiler diagnostics
[Mix.Task.Compiler] Add API for fetching all compiler tasks

2. Bug fixes

Elixir

[Code] Fix delimiter metadata for single quoted atoms and remote calls in Code.string_to_quoted/2
[Code.Formatter] Fix formatter adding extra escapes to quoted remote calls
[Code.Fragment] Properly handle keyword keys as their own entry
[Inspect.Algebra] Ensure next_break_fits respects line_length
[Kernel] Validate AST on unquote and unquote_splicing to provide better error reports instead of failing too late inside the compiler
[Kernel] Avoid crashes when emitting diagnostics on code using \t for indentation
[Module] Include module attribute line and name when tracing its aliases
[Stream] Do not halt streams twice in Stream.transform/5
[URI] Fix a bug when a schemaless URI is given to URI.merge/2

ExUnit

[ExUnit.Assertions] Raise if guards are used in assert/1 with =
[ExUnit.Assertions] Format inserted/deleted maps in list assertions

IEx

[IEx.Helpers] IEx.Helpers.recompile/0 will reload modules changed by other operating system processes

Mix

[mix compile] Ensure warnings from external resources are emitted with --all-warnings when files do not change
[mix deps.compile] Fix escaping issues when invoking rebar3 in some cases
[mix escript] Fix escript layout and support storing priv directories
[mix release] Make .app files deterministic in releases
[Mix.Shell] Fix Mix.Shell on Windows when outputting non UTF-8 characters

3. Soft deprecations (no warnings emitted)

Elixir

[Inspect.Algebra] color/3 is deprecated in favor of color_doc/3
[Inspect.Algebra] fold_doc/2 is deprecated in favor of fold/2
[Kernel] Deprecate unless in favor of if. Use mix format --migrate to automate the migration
[Macro] Macro.struct!/2 is deprecated in favor of Macro.struct_info!/2
[Protocol] Defining __deriving__/3 inside the Any implementation is deprecated, derive it inside the protocol definition itself

4. Hard deprecations

EEx

[EEx] <%# is deprecated in favor of <%!-- or <% #
[EEx] c:EEx.handle_text/2 is deprecated in favor of c:EEx.handle_text/3

Elixir

[Code] Setting :warnings_as_errors is deprecated via Code.put_compiler_option/2. This must not affect developers, as the :warnings_as_errors option is managed by Mix tasks, and not directly used via the Code module
[Enumerable] Deprecate returning a two-arity function in Enumerable.slice/1
[List] List.zip/1 is deprecated in favor of Enum.zip/1
[Module] Deprecate Module.eval_quoted/3 in favor of Code.eval_quoted/3
[Range] Deprecate inferring negative ranges on Range.new/2
[Tuple] Tuple.append/2 is deprecated, use Tuple.insert_at/3 instead

Mix

[mix cmd] Deprecate mix cmd --app APP in favor of mix do --app APP
[mix compile] :warnings_as_errors configuration in :elixirc_options is deprecated. Instead pass the --warnings-as-errors flag to mix compile. Alternatively, you might alias the task: aliases: [compile: "compile --warnings-as-errors"]
[mix test] :warnings_as_errors configuration in :test_elixirc_options is deprecated. Instead pass the --warnings-as-errors flag to mix test. Alternatively, you might alias the task: aliases: [test: "test --warnings-as-errors"]
[Mix.Tasks.Compile] Deprecate compilers/0 in favor of Mix.Task.Compiler.compilers/0

v1.17

The CHANGELOG for v1.17 releases can be found in the v1.17 branch.

51 comments

#elixir-release

97 5575 51

2025-01-29 14:40:19 UTC

Showing Posts 51 to 42

Show Best Posts
Show All (oldest first)
Show All (newest first)

barttenbrinke

I was looking at Module — Elixir v1.18.2 - but I will try compile — OTP 29.0.2 (compiler 10.0.1)

Post #52

hubertlepicki

@barttenbrinke just a quick check, you used @compile :nowarn_unused_vars i.e. there is no space in :nowarn ?

Post #51

barttenbrinke

Nice! It would be nice if you could squelch some warnings easier - things like warning: this clause of defp subject_for_env/2 is never used. I’ve tried @compile {:no_warn_unused, subject_for_env/2} but that does not seem to solve it with mix compile --warnings-as-errors for some reason.

Post #50

barryp

@josevalim Thanks. I think the new message would have made my mistake clear to me. BTW really enjoying Elixir, thanks for the work you’re all doing. HNY

Post #49

harrisi

I compiled zigler with 1.18.0 (and 1.18.1) and got a pretty difficult to parse type warning (Type issue with argument error prong causing unmatched case branch · Issue #502 · E-xyza/zigler · GitHub). At roughly 600 lines long for a single value, I thought it might be interesting.

Post #48

josevalim

Creator of Elixir

The error message will be improved in Elixir v1.18.2.

Post #47

mjrusso

For the Nix users: Elixir 1.18.1 is now available on the nixpkgs-unstable and nixos-unstable channels, with OTP 27 by default.

Here’s what the elixir package gives you out-of-the-box now:

$ elixir --version
Erlang/OTP 27 [erts-15.2] [source] [64-bit] [smp:8:8] [ds:8:8:10] [async-threads:1] [jit]

Elixir 1.18.1 (compiled with Erlang/OTP 27)

Post #46

wojtekmach

Hex Core Team

This seems to fix it:

diff --git a/lib/ex_unit/lib/ex_unit/case.ex b/lib/ex_unit/lib/ex_unit/case.ex
index 21701d870..09a22449f 100644
--- a/lib/ex_unit/lib/ex_unit/case.ex
+++ b/lib/ex_unit/lib/ex_unit/case.ex
@@ -562,7 +562,7 @@ defmacro __before_compile__(%{module: module}) do
       |> Map.new()

     opts = Module.get_attribute(module, :ex_unit_module, [])
-    async? = Keyword.get(opts, :async, false)
+    async? = !!Keyword.get(opts, :async, false)
     group = Keyword.get(opts, :group, nil)
     parameterize = Keyword.get(opts, :parameterize, nil)

I haven’t gone through all relevant code yet so there’s probably a better way to fix this. If that seems interesting, please feel free to take over and send a PR!

Post #45

barryp

@BartOtten There’s no reason except my incompetence. The code was written early in my Elixir journey. At the time my mindset was that use could be doing anything, and it’s in the test code so not released and therefore I don’t need to worry about understanding what it’s doing, or how it works, I just need to know that my tests run. I interpreted async: as a flag, and the tests ran fine. Of course I copied the pattern many times…

The issue is that when I upgraded to 1.18.x my code simply broke, and the message was not especially helpful in tracking down what was wrong. Maybe the macro should insist on a :true or :false value? It is off putting when a minor release breaks your “working” (albeit poor) code in non obvious ways.

Post #44

BartOtten

Great you came up with a simple demo.

Wonder what the reason was/is for the boolean missing notation. Imho saving a ‘true’ or ‘false’ is not worth the implicitness, but maybe I miss the point.

Post #43

Load more posts (41 remaining)