- Fixed a bug when multiple subjects in a case would be split even if not necessary. (Giacomo Cavalieri)
-
gleam removewill now present an error if a package being removed does not exist as a dependency in the project. (Changfeng Lou) -
gleam addnow takes an optional package version specifier, separated by a@, that resolves as follows:gleam add lustre@1.2.3 # "1.2.3" gleam add lustre@1.2 # ">= 1.2.0 and < 2.0.0" gleam add lustre@1 # ">= 1.0.0 and < 2.0.0"
-
Added more an informative error message for when attempting to use the
..syntax to append to a list rather than prepend.error: Syntax error ┌─ /src/parse/error.gleam:4:14 │ 4 │ [..rest, last] -> 1 │ ^^^^^^ I wasn't expecting elements after this I was expecting the end of the list. A spread can only be used to match on the entire end of a list. It is not possible to extract items from the end of a list using pattern matching because that would require walking through the entire list. Lists are immutable and singly-linked, so to match on the end of a list would require the whole list to be traversed. This would be slow, so there is no built-in syntax for it. Pattern match on the start of the list instead. -
The compiler now emits a warning for redundant function captures in a pipeline:
warning: Redundant function capture ┌─ /src/warning/wrn.gleam:5:17 │ 5 │ 1 |> wibble(_, 2) |> wibble(2) │ ^ You can safely remove this This function capture is redundant since the value is already piped as the first argument of this call. See: https://tour.gleam.run/functions/pipelines/ -
The syntax
[a..b]is now deprecated in favour of the[a, ..b]syntax. This was to avoid it being mistaken for a range syntax, which Gleam does not have. (Giacomo Cavalieri) -
Functions etc named
maybeare now escaped in generated Erlang as it is now a reserved word in Erlang/OTP 27. (Jake Barszcz) -
Functions, types and constructors named
maybeandelseare now escaped in generated Erlang to avoid clashing with Erlang's keywords. (Jake Barszcz) and (Giacomo Cavalieri) -
Non byte aligned arrays that use literals for size are now marked as an Unsupported feature for Javascript since they would already cause a runtime error on Javascript.
This means if you compile specifically for Javascript you will now recieve this error:
error: Unsupported feature for compilation target ┌─ /src/test/gleam_test.gleam:6:5 │ 6 │ <<1:size(5)>> │ ^^^^^^^^^ Non byte aligned array is not supported for JavaScript compilation.Else any functions which rely on this will not be compiled into Javascript. (Pi-Cla)
-
Compilation fault tolerance is now at a statement level instead of a function level. This means that the compiler will attempt to infer the rest of the statements in a function when there is an error in a previous one. (Ameen Radwan)
-
The JavaScript prelude is no-longer rewritten each time a project is compiled to JavaScript. (Ofek Doitch)
-
Compiler now supports arithmetic operations in guards. (Danielle Maywood)
-
Import cycles now show the location where the import occur. (Ameen Radwan)
-
Error messages resulting from unexpected tokens now include information on the found token's type. This updated message explains how the lexer handled the token, so as to guide the user towards providing correct syntax.
Following is an example, where the error message indicates that the name of the provided field conflicts with a keyword:
3 │ A(type: String) │ ^^^^ I was not expecting this Found the keyword `type`, expected one of: - `)` - a constructor argument name -
When compiling to JavaScript constants will now be annotated as
@__PURE__. (Giacomo Cavalieri)
-
The language server will now suggest the "Remove redundant tuple" action even if the case expression contains some catch all patterns:
case #(a, b) { #(1, 2) -> todo _ -> todo }Becomes:
case a, b { 1, 2 -> todo _, _ -> todo } -
LSP can now suggest completions for values and types from importable modules and adds the import to the top of the file. (Ameen Radwan)
-
Diagnostics with extra labels now show the diagnostic in all locations including across multiple files. (Ameen Radwan)
-
LSP completions now use the "text_edit" language server API resulting in better/more accurate insertions. (Ameen Radwan)
-
Completions are no longer provided inside comments. (Nicky Lim)
-
The language server will now show all the ignored fields when hovering over
..in a record pattern. (Giacomo Cavalieri)
-
Fixed a bug where the compiler would output a confusing error message when trying to use the spread syntax to append to a list. (Giacomo Cavalieri)
-
Fixed a bug where the formatter would strip empty lines out of the body of an anonymous function passed as an argument. (Giacomo Cavalieri)
-
Fixed a bug where the compiler would crash when a type was defined with the same name as an imported type. (Gears)
-
Fixed a bug where a horizontal scrollbar would appear on code blocks in built documentation when they contained lines 79 or 80 characters long. (Richard Viney)
-
Fixed a bug where importing a record constructor in an unqualified fashion and aliasing it and then using it in a constant expression would generate invalid JavaScript. (Louis Pilfold)
-
Fixed a bug where the compiler would crash because types weren't registered if they referenced a non-existent type. (Gears)
-
Fixed a bug where the compiler would generate invalid Erlang when pattern matching on strings with an
aspattern. (Giacomo Cavalieri) -
Fixed a bug where the compiler would warn that a module alias was unused when it was only used in case patterns. (Michael Jones)
-
Fixed a bug where the compiler could fail to detect modules that would clash with Erlang modules. (Louis Pilfold)
-
Fixed a bug where dependency version resolution could crash for certain release candidate versions. (Marshall Bowers)
-
Fixed a bug where trailing comments would be moved out of a bit array. (Giacomo Cavalieri)