Richard Feldman
25518a76cf
Narrow down test failures
2020-01-18 08:16:59 -05:00
Richard Feldman
7e8ae25c2c
Fix some tests
2020-01-18 03:48:03 -05:00
Richard Feldman
99f99380da
Got tests compiling
2020-01-18 02:57:20 -05:00
Folkert
83dd87b121
add tests for ext_var and TagUnion matches
2020-01-18 01:12:10 +01:00
Folkert
17afa4bd1a
fix typo in tag unification logic
2020-01-17 15:17:14 +01:00
Folkert
4a4f2a2562
canonicalize Tags in types
2020-01-17 14:23:48 +01:00
Folkert
50f8bf1856
Recursive functions must be shared
2020-01-16 23:53:03 +01:00
Folkert
dbe615b4c5
Failing test for LinkedList.map
...
The rank of the ext_var of Cons is incorrect, it is to high triggering a
panic.
2020-01-16 23:44:08 +01:00
Folkert
3b0610130d
Optimize When typechecking
...
The old code generated constraints twice, which is inefficient. The new
version also sets the variable in the AST to the type of the full
expression in all cases (with and without an annotation).
In lifting annotations to Attr types, int and float are now
special-cased
2020-01-12 17:15:19 +01:00
Folkert
20ae31fb39
move variable testing code to tests/helpers
2020-01-12 11:19:16 +01:00
Folkert
e5a9a016cd
prevent duplicate definitions
...
patterns that bind multiple symbols, like `{ x, y } = someDef` would insert the definition twice (once for x, once for y).
that can lead to problems in codegen, and means the work for this definition is done twice.
Now we use the region of the pattern as a key, to ensure the pattern and its associated expression is only processed once.
2020-01-12 01:20:09 +01:00
Folkert
ceba17ed98
use all variables in uniq infer
2020-01-12 00:58:37 +01:00
Folkert
5bdf1d78cf
for infer tests, check declared vs. used variables
2020-01-11 23:37:32 +01:00
Folkert
542114acc4
Fix #134
...
Another case of a variable not being added to `flex_vars`, which means
it's not copied and may lead to principality issues
2020-01-11 23:37:32 +01:00
Folkert
3168998a2b
use pattern from annotation
...
The symbols from bare annotation patterns weren't added to scope, leading to a panic down the line in solve()
2020-01-11 23:37:32 +01:00
Folkert
1a5684e506
Merge branch 'trunk' into uniq-annotations
2020-01-11 11:04:23 +01:00
Folkert
ca9e86a56a
fix typo in tests
2020-01-09 13:48:37 +01:00
Folkert
ca41a0339d
tags with arguments in uniq infer
2020-01-09 13:38:47 +01:00
Folkert
dd240cf53a
typecheck tags with arguments
...
also fixes an issue where private tags in patterns would not have the @ symbol in their name
2020-01-09 13:20:21 +01:00
Richard Feldman
e019a4b662
Merge branch 'trunk' into fix-uniq-principality
2020-01-08 22:47:12 -05:00
Folkert
fc81a82472
merge ext fields into the record when printing
2020-01-08 23:45:59 +01:00
Folkert
6c363a7208
restore inference for guards in pattern matches
2020-01-08 20:42:29 +01:00
Folkert
37cf0f940b
possible fix for uniq principality
...
Apparently this fails because a closure would force the fn_var to be equal to the inferred type.
I'm really not sure why this would be a problem though. It works fine for 'normal' inference.
2020-01-08 14:30:14 +01:00
Folkert
34856e620d
Merge branch 'trunk' into refactor-uniq-infer
2020-01-07 20:16:41 +01:00
Folkert
2db7c0503f
fix uniq record destructure patterns
2020-01-07 20:09:52 +01:00
Richard Feldman
f00bf354aa
Update private tag test
2020-01-06 21:36:38 -05:00
Folkert
9c8e316135
add failing private tag application
...
Private tags are not currently parsed in applications, so in this case parsing fails after `@Foo`
2020-01-06 21:36:38 -05:00
Folkert
864aefb925
constrain If
2020-01-06 14:46:02 +01:00
Folkert
fe35e99518
remove If desugaring
...
To generate good error messages, this needs to happen after type checking. Otherwise we can get weird things like
if Foo then 1 else 0
Giving some error that the `Foo` tag is not covered in a pattern match.
2020-01-06 00:56:50 +01:00
Folkert
1211686797
canonicalize PrivateTag and If
...
not totally sure about the if, because we lose information that may be helpful for good error messages. Still it was fun to implement.
PrivateTag can be matched on, but can't currently be applied (an issue with parsing)
2020-01-06 00:44:49 +01:00
Folkert
5ad2823a23
support Tag arguments
2020-01-05 22:49:08 +01:00
Folkert
1d5ceeafe4
constraint Tag patterns
2020-01-05 21:37:06 +01:00
Folkert
26bd52c2a2
scaffolding for tag unions
2020-01-05 20:54:34 +01:00
Folkert
9e94241045
add test case for record update
...
fix bug in record unification: im-rs difference is symmetric, while in haskell it is asymmetric
2020-01-05 15:48:20 +01:00
Folkert
4147582738
remove RecordFieldLabel enum (now a type alias)
...
I've kept it a type alias so that we can more easily change this in the future. But removing the alias is also simple
2020-01-04 23:22:40 +01:00
Folkert
4f3f85f064
fix principality
2020-01-02 00:48:02 +01:00
Richard Feldman
a5a4141e7e
Reproduce principality bug
2020-01-01 17:02:22 -05:00
Richard Feldman
2f61455061
Merge remote-tracking branch 'origin/trunk' into improve-errors
2020-01-01 04:04:53 -05:00
Richard Feldman
b6ebfb8537
Merge remote-tracking branch 'origin/trunk' into constrain
2019-12-30 21:59:19 -05:00
Richard Feldman
ffd3cc4211
Constrain canonical exprs and modules
2019-12-30 20:33:05 -05:00
Folkert
d47abfb3a6
preparations for unifying optional fields
2019-12-30 16:52:57 +01:00
Folkert
13b3748996
fix record annotations
...
For literals, the extension variable was free. That is incorrect. so
foo : { x : Int } -> Int
Would actually have type
foo : { x : Int }* -> Int
Now the extension variable is `{}`, fixing this issue
2019-12-30 13:59:54 +01:00
Folkert
d8d045d122
parse optional fields
2019-12-30 12:42:24 +01:00
Folkert
9aee825391
remove EmptyRecord value from can/expr
2019-12-24 23:37:50 +01:00
Richard Feldman
5ffd62b9b3
Update record inference test
2019-12-23 23:44:08 -08:00
Richard Feldman
fd36e153ed
Merge branch 'trunk' into ct/case-when-to-when-is
2019-12-23 23:29:43 -08:00
Chad Stearns
844ca0a50e
Changed instances of 'Case' in the code to 'When'
2019-12-23 17:41:19 -05:00
Chad Stearns
62a004c103
Case is to when is
2019-12-23 17:17:04 -05:00
Chad Stearns
9a5b6a03b4
case when to case is
2019-12-23 17:08:53 -05:00
Folkert
c4ad367a8a
implement RecordField logic
2019-12-23 22:11:38 +01:00
