Provides elm-review
rules to ensure that anything (readily) sortable in Elm code is sorted in the
"proper" order.
- ๐ง
NoUnsortedCases
- Reports case patterns that are not in the "proper" order. - ๐ง
NoUnsortedLetDeclarations
- Reportslet
declarations that are not in the "proper" order. - ๐ง
NoUnsortedRecords
- Reports record fields that are not in the "proper" order. - ๐ง
NoUnsortedTopLevelDeclarations
- Reports top-level declarations that are not in the "proper" order.
module ReviewConfig exposing (config)
import NoUnsortedCases
import NoUnsortedLetDeclarations
import NoUnsortedRecords
import NoUnsortedTopLevelDeclarations
import Review.Rule exposing (Rule)
config : List Rule
config =
[ NoUnsortedCases.rule NoUnsortedCases.defaults
, NoUnsortedLetDeclarations.rule
(NoUnsortedLetDeclarations.sortLetDeclarations
|> NoUnsortedLetDeclarations.alphabetically
)
, NoUnsortedRecords.rule
(NoUnsortedRecords.defaults
|> NoUnsortedRecords.reportAmbiguousRecordsWithoutFix
)
, NoUnsortedTopLevelDeclarations.rule
(NoUnsortedTopLevelDeclarations.sortTopLevelDeclarations
|> NoUnsortedTopLevelDeclarations.portsFirst
|> NoUnsortedTopLevelDeclarations.exposedOrderWithPrivateLast
|> NoUnsortedTopLevelDeclarations.alphabetically
)
]
You can try the example configuration above out by running the following command:
elm-review --template SiriusStarr/elm-review-no-unsorted/example
-
1.1.8
-- ๐NoUnsortedRecords
now keeps record info from indirect dependencies. Previously, only information in direct imports was kept, due to an oversight in howelm-review
works. This should lead to fewer records being unknown now and thus correctly sorted. -
1.1.7
-- ๐ Fix bug inNoUnsortedRecords
where record patterns could be considered ambiguous due to fields in type annotations being ignored. For instance:type alias FBB = { foo : Int, bar : Int, baz : Int } type alias BBF = { bar : Int, baz : Int, foo : Int } fooBar : { bar : Int, foo : Int } -> Int fooBar { bar, foo } = foo + bar
{ bar, foo }
was considered ambiguous even though it can't matchFBB
orBBF
due to the type annotation specifying thatbaz
is absent. The fix works with extra fields and generic records as well, e.g.{ r | bar : Int, extra : Int, foo : Int }
also prevents ambiguity, but not{ r | bar : Int, foo : Int }
. -
1.1.6
-- ๐ Fix bug inNoUnsortedRecords
where multiple record matches were not considered identical orderings if the indices of the matching fields were different. For example, the following record update was considered ambiguous before this bug fix:type alias A = { a : Int, b : String, c : Bool } type alias B = { b : String, c : Bool } a r b c = { r | b = b, c = c }
Since
b
andc
had different indices in the matching records, even though the sort order was the same. -
1.1.5
-- ๐ Improve docs forNoUnsortedCases.doNotLookPastUnsortable
-
1.1.4
- โก๏ธ Improve performance when dealing with ignored files (~10% in one real-world test case).
- ๐ Bump
elm-review
to v2.12.1 for upstream bugfix.
-
1.1.3
-- Bumpelm-review
to v2.11.1 and mark rule as providing fixes. -
1.1.2
-- ๐ Fix a bug where, when dealing with cyclical sorting conditions, some low priority edges associated with the cycle but not actually responsible for it would be disregarded instead of only those responsible. In practice, this meant that, very rarely, sortablecase
patterns would not be sorted at all (or would be sorted unstably) when wildcard (_
) patterns lead to cyclic ordering relationships. Since the problem at the core of this is NP-hard, it is not guaranteed to be fixed but should nevertheless now occur substantially less often (if at all). Please open an issue if you encounter unstable sorting behavior orcase
patterns that should be able to be sorted but are not with this version. -
1.1.1
-- ๐ Fix critical bug caused by upstream bugfix inelm-syntax
. Version7.2.9
ofelm-syntax
no longer incorrectly parses record field type signature ranges, soNoUnsortedRecords
(which repaired the previously incorrect behavior) was producing in invalid ranges for fixes, causing them to mangle code.elm-syntax
new lower bound is set to7.2.9
and the workaround removed, fixing the issue. -
1.1.0
- New Features:
- โจ -- Disable typechecking of unambiguous records by
NoUnsortedRecords
. The old default can be re-enabled withtypecheckAllRecords
- โจ -- Add control over subrecord support for
NoUnsortedRecords
. Default behavior is to sort them when they appear in context (e.g. as part of their larger record) but not when they appear alone. The old behavior did this unreliably and also treated custom type argument records as always canonical; this old behavior may be re-enabled (without the unreliability) withtreatCustomTypeRecordsAsCanonical
. New settings for this behavior are also available withtreatSubrecordsAsUnknown
andtreatAllSubrecordsAsCanonical
.
- โจ -- Disable typechecking of unambiguous records by
- Bugfixes:
- ๐ -- Fix critical bug causing control flow to sometimes be altered by
NoUnsortedCases
due toList.sort
assuming transitivity. New sorting methodology renders such issues impossible in the future. - ๐ -- Fix doc comments not moving for ports with
NoUnsortedTopLevelDeclarations
(possibly leading to invalid code after fixes). Doc comment support for ports was added manually, aselm-syntax
does not parse them, and they now behave like doc comments for all other declarations. - ๐ -- Improve handling of subrecords by
NoUnsortedRecords
. Previously, they were sometimes sorted and sometimes not (depending on what type information was available). (See above for new configuration options controlling this behavior.) - ๐ -- Fix bad type checking by
NoUnsortedRecords
assigning independent type vars to the same type, e.g. assigning allNothing
s to the sameMaybe a
value, even if they were different. - ๐ -- Fix bad type checking by
NoUnsortedRecords
preserving type var assignment between fields (for type vars not in the known record), causing e.g. allNothing
s in a record to be required to be the same type. - ๐ -- Fix bad type inference of lambda functions by
NoUnsortedRecords
. - ๐ -- Fix non-functions being considered dependencies/helpers by
NoUnsortedTopLevelDeclarations
. This brings actual rule behavior in line with that stated in the documentation.
- ๐ -- Fix critical bug causing control flow to sometimes be altered by
- Other Changes:
- โก๏ธ -- Significantly improve performance of
NoUnsortedRecords
(2x faster on some real-world codebases). Previously, a significant amount of unnecessary type inference and duplicated recursion was being performed due to lack of laziness.
- โก๏ธ -- Significantly improve performance of
- New Features:
-
1.0.6
-- ๐ Fix a bug causingNoUnsortedRecords
to recurse infinitely in certain rare cases where a type variable was assigned to a value containing an identically-named type variable (e.g. the type variablea
ended up being assigned to a value defined by the generic record{ a | field : Int }
). -
1.0.5
-- ๐ Fix critical bug causingNoUnsortedRecords
to rarely generate valid code that was missing fields for record updates. It is recommended that you check any record update expressions that contained comments that were fixed by this rule prior to this version, as one or more fields may have been appended to the end of a comment in sorting. -
1.0.4
-- Update toelm-syntax
v7.2.8 to fix upstream issue with lambda ranges generating invalid code in fixes. -
1.0.3
-- Fix crash with--fix
or--watch
due to comparison of function types inProjectContext
ofNoUnsortedRecords
. -
1.0.2
-- Fix crash in handling of record type comparison. -
1.0.1
-- Fix crash in handling of generic records when not all fields must be present. -
1.0.0
-- Initial release