Make FieldType and @@ abstract #807

joroKr21 · 2018-01-13T21:14:42Z

aka. "You shall not dealias"

Fixes a bunch of problems caused by:

the compiler normalizing types behind your back;
the compiler failing to unify the normalized types.

Also has the additional benefit of printing more nicely in the REPL.
This will always box however. Would that be a problem?

TODO: Check which bugs are fixed, reference them here and add regression tests:

Should close Look into type tagging and value classes #644 - check that value classes and arrays work (boxed)
Should fix Interaction between Lazy and tagged types #584 in a different way than Fix Lazy for tagged types #797 - add regression test
Should fix tag[Tag](value) doesnt work with explicitly declared target type #557 - add regression test
Fully fixes Interaction between shapeless FieldType, and Scalaz tagged types #309 - removed workaround
Should fix Runtime exception if the value of record field is Any or AnyRef #44 - similar to Look into type tagging and value classes #644 (with boxing)

joroKr21 · 2018-01-13T21:18:16Z

Sadly I think this is not allowed on Scala 2.10 so this has to wait for #655

joroKr21 · 2018-01-13T21:22:48Z

core/src/main/scala/shapeless/labelled.scala


 import scala.reflect.macros.whitebox

 object labelled {
  /**
   * The type of fields with keys of singleton type `K` and value type `V`.
   */
-  type FieldType[K, +V] = V with KeyTag[K, V]
+  type FieldType[K, +V] <: V with KeyTag[K, V]


What's the purpose of V in KeyTag?

How about calling FieldType[K, V] K ->> V?

codecov-io · 2018-01-14T12:49:01Z

Codecov Report

Merging #807 into master will decrease coverage by 2.01%.
The diff coverage is 10.00%.

@@            Coverage Diff             @@
##           master     #807      +/-   ##
==========================================
- Coverage   89.86%   87.85%   -2.02%     
==========================================
  Files          34       65      +31     
  Lines        1470     1498      +28     
  Branches        1        3       +2     
==========================================
- Hits         1321     1316       -5     
- Misses        149      182      +33

Impacted Files	Coverage Δ
core/src/main/scala/shapeless/generic.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/singletons.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/typeoperators.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/labelled.scala	`85.71% <66.66%> (-14.29%)`	⬇️
core/src/main/scala/shapeless/hlists.scala	`63.15% <0.00%> (-3.51%)`	⬇️
...re/src/main/scala/shapeless/syntax/coproduct.scala	`97.87% <0.00%> (-0.05%)`	⬇️
core/src/main/scala/shapeless/fin.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/annotation.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/poly.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/sybclass.scala	`0.00% <0.00%> (ø)`
... and 28 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 85fb65f...7192c0e. Read the comment docs.

joroKr21 · 2018-01-14T23:26:08Z

ClassTag magic doesn't work. HALP!

kwark · 2018-04-30T06:47:08Z

core/src/main/scala/shapeless/typeoperators.scala

-  def apply[U] = new Tagger[U]
-
+private[shapeless] trait Tag {
+  type @@[+T, U] <: T with Tagged[U]


I have been giving this encoding a spin, for encoding/decoding tagged types to/from json:

with the @@ part of the private Tag class, I run into a problem when using play-json's format macro:

No instance of play.api.libs.json.Format is available for xxx.Tag.$at$at[java.lang.String, TaggingTest.FooTag] in the implicit scope (Hint: if declared in the same file, make sure it's declared before)

If I move the @@ type to the tag object (just a few lines below), it does work. Check-out this scastie: https://scastie.scala-lang.org/kwark/rnWsgsTQQQ2Fmsh97NMRdQ

circe-generic seems to work with both variants

I introduced the trait only because Scala 2.10 doesn't support abstract types in objects / classes, but somehow it supports them when inherited.

Also it looks like type @@[+T, U] <: T with Tagged[U] now boxes tagged primitives, while type @@[+T, U] = T with Tagged[U] does not?

Yes, that's true unfortunately.

aka. "You shall not dealias" Fixes a bunch of problems caused by: 1. the compiler normalizing types behind your back; 2. the compiler failing to unify the normalized types.

milessabin · 2020-01-21T11:01:38Z

Continuing the discussion from scala/community-build#1047 (comment).

I think it's worth investigating which approach most smoothly transfers to Dotty. At least some uses of tagged types will be replaced by opaque types, but probably not all. It'd be good to see what works best for the cases which aren't handled well by opaque types.

joroKr21 · 2020-02-15T20:13:13Z

Progress!

The boxing issue is solved by scala/scala#8720 so we can have our cake and eat it too.

Dotty has better type inference and some simple cases that I tried work better than in Scala 2 with the current encoding of tagged types. I still need to come up with a better comparison (and also check how they are erased in dotty).

I agree that perhaps tagged types will not be completely replaced (because they are subtypes) but as a native solution I expect opaque types to become much more widely used, perhaps even as much as to make tagged types on shapeless 3 unwarranted.

joroKr21 · 2020-03-16T14:12:21Z

Changed milestone to 2.5.0 which will be 2.13 only.
If scala/scala#8720 is merged (I see no reason why not) then this has no downsides.

codecov-commenter · 2020-08-16T07:03:01Z

Codecov Report

Merging #807 into master will decrease coverage by 2.01%.
The diff coverage is 10.00%.

@@            Coverage Diff             @@
##           master     #807      +/-   ##
==========================================
- Coverage   89.86%   87.85%   -2.02%     
==========================================
  Files          34       65      +31     
  Lines        1470     1498      +28     
  Branches        1        3       +2     
==========================================
- Hits         1321     1316       -5     
- Misses        149      182      +33

Impacted Files	Coverage Δ
core/src/main/scala/shapeless/generic.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/singletons.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/typeoperators.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/labelled.scala	`85.71% <66.66%> (-14.29%)`	⬇️
core/src/main/scala/shapeless/hlists.scala	`63.15% <0.00%> (-3.51%)`	⬇️
...re/src/main/scala/shapeless/syntax/coproduct.scala	`97.87% <0.00%> (-0.05%)`	⬇️
core/src/main/scala/shapeless/annotation.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/fin.scala	`0.00% <0.00%> (ø)`
...re/src/main/scala/shapeless/hlistconstraints.scala	`0.00% <0.00%> (ø)`
core/src/main/scala/shapeless/zipper.scala	`0.00% <0.00%> (ø)`
... and 28 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e8b6028...edc47fa. Read the comment docs.

joroKr21 · 2021-04-21T18:05:04Z

Changed milestone to 2.5.0 which will be 2.13 only.
If scala/scala#8720 is merged (I see no reason why not) then this has no downsides.

In the end it was not merged. In Scala 3 we have opaque types.

joroKr21 commented Jan 13, 2018

View reviewed changes

joroKr21 force-pushed the type-aliases branch 3 times, most recently from f1a4c77 to 602ce99 Compare January 14, 2018 12:48

joroKr21 force-pushed the type-aliases branch 4 times, most recently from b2a53cd to a42e14f Compare January 14, 2018 19:52

joroKr21 force-pushed the type-aliases branch 2 times, most recently from 56469d0 to 59edbc8 Compare January 16, 2018 01:36

joroKr21 force-pushed the type-aliases branch 3 times, most recently from c8d15d9 to e7c72d6 Compare February 28, 2018 14:33

kwark reviewed Apr 30, 2018

View reviewed changes

joroKr21 force-pushed the type-aliases branch from e7c72d6 to c58af22 Compare May 26, 2018 20:13

joroKr21 force-pushed the type-aliases branch from c58af22 to 5eaf2fa Compare October 13, 2018 21:04

joroKr21 force-pushed the type-aliases branch from 5eaf2fa to eb7f006 Compare November 30, 2019 20:27

joroKr21 force-pushed the type-aliases branch from eb7f006 to 4944f42 Compare December 24, 2019 12:18

joroKr21 added this to the shapeless-2.4.0 milestone Jan 12, 2020

Make FieldType and @@ abstract

7517d9a

aka. "You shall not dealias" Fixes a bunch of problems caused by: 1. the compiler normalizing types behind your back; 2. the compiler failing to unify the normalized types.

joroKr21 force-pushed the type-aliases branch from 4944f42 to 7517d9a Compare January 21, 2020 06:34

joroKr21 mentioned this pull request Jan 21, 2020

2.12: advance project SHAs scala/community-build#1047

Merged

joroKr21 added 3 commits March 9, 2020 10:11

Merge branch 'master' into type-aliases

3544a63

Merge branch 'master' into type-aliases

8c28fed

Merge branch 'master' into type-aliases

ec0d207

joroKr21 modified the milestones: shapeless-2.4.0, shapeless-2.5.0 Mar 16, 2020

joroKr21 added 6 commits March 27, 2020 08:22

Merge branch 'master' into type-aliases

833fbf5

Merge branch 'master' into type-aliases

6572778

Merge branch 'master' into type-aliases

43a6dd1

Merge branch 'master' into type-aliases

c820625

Merge branch 'master' into type-aliases

7192c0e

Merge branch 'master' into type-aliases

bf8c199

joroKr21 added 4 commits September 19, 2020 01:01

Merge branch 'master' into type-aliases

edc47fa

Merge branch 'master' into type-aliases

47a5d0c

Merge branch 'master' into type-aliases

5185245

Merge branch 'master' into type-aliases

344dc3c

Base automatically changed from master to main March 19, 2021 12:59

joroKr21 closed this Apr 21, 2021

joroKr21 mentioned this pull request Nov 15, 2022

Shouldn't KeyTag be an abstract type rather than a trait? #1285

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Make FieldType and @@ abstract #807

Make FieldType and @@ abstract #807

joroKr21 commented Jan 13, 2018 •

edited

Loading

joroKr21 commented Jan 13, 2018

joroKr21 Jan 13, 2018

joroKr21 Jan 13, 2018

codecov-io commented Jan 14, 2018 •

edited

Loading

joroKr21 commented Jan 14, 2018

kwark Apr 30, 2018

joroKr21 Apr 30, 2018

kwark Apr 30, 2018

joroKr21 Apr 30, 2018

milessabin commented Jan 21, 2020

joroKr21 commented Feb 15, 2020

joroKr21 commented Mar 16, 2020

codecov-commenter commented Aug 16, 2020 •

edited

Loading

joroKr21 commented Apr 21, 2021

Make FieldType and @@ abstract #807

Make FieldType and @@ abstract #807

Conversation

joroKr21 commented Jan 13, 2018 • edited Loading

joroKr21 commented Jan 13, 2018

joroKr21 Jan 13, 2018

Choose a reason for hiding this comment

joroKr21 Jan 13, 2018

Choose a reason for hiding this comment

codecov-io commented Jan 14, 2018 • edited Loading

Codecov Report

joroKr21 commented Jan 14, 2018

kwark Apr 30, 2018

Choose a reason for hiding this comment

joroKr21 Apr 30, 2018

Choose a reason for hiding this comment

kwark Apr 30, 2018

Choose a reason for hiding this comment

joroKr21 Apr 30, 2018

Choose a reason for hiding this comment

milessabin commented Jan 21, 2020

joroKr21 commented Feb 15, 2020

joroKr21 commented Mar 16, 2020

codecov-commenter commented Aug 16, 2020 • edited Loading

Codecov Report

joroKr21 commented Apr 21, 2021

joroKr21 commented Jan 13, 2018 •

edited

Loading

codecov-io commented Jan 14, 2018 •

edited

Loading

codecov-commenter commented Aug 16, 2020 •

edited

Loading