Inference of lambda type and array element type from an antecedent (RFC 45). #2168

jemc · 2017-08-15T05:41:45Z

Resolves #2094.

jemc · 2017-08-15T05:43:10Z

This one is in the final stages of cleanup work before preparing to merge.

I've tried to imagine and test for all the cases of this, but invariably I will have missed some cases. We can clean up the cases I missed later, as they come up.

jemc · 2017-08-16T02:44:36Z

Hm, nevermind - I've come across a case that I didn't handle already, so I'm going to take a look to see if it can be handled.

Praetonus · 2017-08-16T13:16:49Z

src/libponyc/expr/array.c

@@ -27,6 +272,13 @@ bool expr_array(pass_opt_t* opt, ast_t** astp)
    told_type = true;
  }

+  if(!told_type && (ast_childcount(elements) == 0))


This can use the size local instead of the duplicate call to ast_childcount.

Good call. Fixed in 943c1ba.

Praetonus · 2017-08-16T13:18:29Z

src/libponyc/expr/array.c

+#include "../type/lookup.h"
+#include "ponyassert.h"
+
+static ast_t* build_array_type(ast_t* scope, ast_t* elem_type, token_id cap)


Is there a particular reason to use this new function instead of type_builtin_args?

I didn't know about that function.

In 943c1ba I refactored build_array_type to use type_builtin_args as part of its implementation.

Praetonus · 2017-08-16T14:18:44Z

src/libponyc/expr/array.c

+    {
+      for(ast_t* c = ast_child(ast); c != NULL; c = ast_sibling(c))
+        find_possible_element_types(opt, c, list);
+    }


I'd feel more comfortable with a break or return here, to avoid problems in future refactorings.

Good call. Fixed in 943c1ba.

jemc · 2017-08-27T20:14:53Z

I found and fixed some more edge cases.

I also addressed @Praetonus' review comments.

This is ready for a final review.

SeanTAllen · 2017-08-27T21:25:59Z

@jemc looks like this needs a rebase?

jemc · 2017-08-27T23:28:09Z

@SeanTAllen - I'm not sure what you mean. As long as we merge with the squash-and-merge feature of GitHub, it will only show as one commit.

In other news, I'm applying the "DO NOT MERGE" label on this again because it's hitting an issue in the changelog tool compilation.

SeanTAllen · 2017-08-27T23:55:20Z

@jemc i was referring to the "This branch cannot be rebased due to conflicts"

Praetonus · 2017-08-30T17:58:41Z

The Linux LLVM 3.9 builds (both debug and release) are hitting an assertion failure in the type inference code when building pony-stable.

These calls were incorrect anyway, because it wasn't a subtree - the opt frames were not the right frames, causing subtle issues in some cases.

jemc · 2017-09-08T07:15:29Z

Removing the "DO NOT MERGE" label - if this passes CI, it's good to merge.

Theodus · 2017-09-08T10:04:40Z

@jemc Travis is showing the following error on all builds:

Error:
/home/travis/build/ponylang/ponyc/packages/cli/command_parser.pony:162:19: can't reuse name 'v'
            | let v: _Value => v
                  ^
    Info:
    /home/travis/build/ponylang/ponyc/packages/cli/command_parser.pony:160:11: previous use of 'v'
              let v: _Value =
              ^

jemc · 2017-09-10T16:24:47Z

This now passed all Travis builds except the last build for OSX, which had a Travis-caused error. I restarted that one build, so here's hoping it finally passes this time.

jemc · 2017-09-10T18:24:14Z

🎉

SeanTAllen · 2017-09-12T12:49:37Z

@jemc can you write up release notes for this so i can include with the release?

jemc · 2017-09-13T05:12:14Z

@SeanTAllen - here's a shot at some helpful release notes:

This release includes significant improvements to type inference for array literals and lambda expressions when the type of the expression has an unambiguous antecedent (such as the "left side" of an assignment, the parameter signature of a method call, or the return type of a method body), including the following improvements:

Empty array literals are now valid syntax: let a: Array[U8] = []
Non-homogenous concrete elements in an array literal can be treated as a trait or interface instead of as a union: let a: Array[Stringable] = [true; None; "string"]
Array literals can have an implied capability recovery to iso or val: let a: Array[String] val = ["foo"; "bar"; "baz"]
Lambda expressions can have implied parameter and return types: let fn: {(U64, U64): U64} = {(x, y) => x + y }
Unused lambda parameters can use the "don't care" symbol (_) instead of a parameter name and type: let fn: {(U64, U64): U64} = {(_, y) => y * 2 }
The receiver and object capability of lambda expressions can be inferred from context, instead of being inferred from the statefulness of the lamba.

For more information, see RFC #45.

SeanTAllen · 2017-09-13T12:48:03Z

@jemc is this a breaking change? would we be doing 0.19.1 or 0.20.0?

jemc · 2017-09-13T13:34:02Z

@SeanTAllen - it's not a breaking change. If this breaks some working code, that would be a bug to be fixed.

That said, I think we've been bumping the second number for both breaking changes and new features, here in the pre-1.0.0 days. That is, I think we use the third number for representing bug fixes. I could be wrong though.

SeanTAllen · 2017-09-13T13:44:10Z

@jemc i havent been bumping for new features at this point, only breaking changes.

Test for specific errors in lambda.cc.

6a44eaa

jemc added the do not merge This PR should not be merged at this time label Aug 15, 2017

jemc self-assigned this Aug 15, 2017

jemc force-pushed the feature/lambda-inference branch 2 times, most recently from 0d65065 to 9a2145c Compare August 16, 2017 02:26

jemc removed the do not merge This PR should not be merged at this time label Aug 16, 2017

Implement array and lambda inference.

460d5e4

jemc force-pushed the feature/lambda-inference branch from 9a2145c to 460d5e4 Compare August 16, 2017 02:28

jemc changed the title ~~Feature: Lambda and array inference~~ Inference of lambda type and array element type from an antecedent (RFC 0045). Aug 16, 2017

jemc changed the title ~~Inference of lambda type and array element type from an antecedent (RFC 0045).~~ Inference of lambda type and array element type from an antecedent (RFC 45). Aug 16, 2017

jemc added the changelog - added Automatically add "Added" CHANGELOG entry on merge label Aug 16, 2017

jemc mentioned this pull request Aug 16, 2017

Arrays of lambdas build a union typearg; could infer a common interface #808

Closed

jemc added the do not merge This PR should not be merged at this time label Aug 16, 2017

Praetonus reviewed Aug 16, 2017

View reviewed changes

This was referenced Aug 17, 2017

lambdas are being treated as concrete types in the array literal constructor #568

Open

Lambda capture types aren't inferred correctly #1441

Open

jemc added 8 commits August 22, 2017 19:19

Allow inferring Arrays from an values interface antecedent.

0c22363

Allow inferring Arrays from a values call with Iterator antecedent.

39333d8

Update standard library to use array inference where applicable.

070cf82

Fix some edge cases of lambda inference involving type parameters.

858beef

Update standard library to use lambda inference where applicable.

14f909e

Fix PASS_EXPR invocation for default args in lookup_nominal.

7e89179

Some array inference refactoring based on review from @Praetonus.

943c1ba

Merge branch 'master' into feature/lambda-inference

5c12f9d

jemc removed the do not merge This PR should not be merged at this time label Aug 27, 2017

jemc added the do not merge This PR should not be merged at this time label Aug 27, 2017

jemc added 4 commits September 7, 2017 14:57

Swap order of TK_ASSIGN children so that left precedes right.

92eeec2

Swap order of TK_CALL children so that receiver comes first.

6cc7428

Remove obsolete calls to ast_passes_subtree in find_antecedent_type.

aa5b46a

These calls were incorrect anyway, because it wasn't a subtree - the opt frames were not the right frames, causing subtle issues in some cases.

Merge branch 'master' into feature/lambda-inference

638ed60

jemc removed the do not merge This PR should not be merged at this time label Sep 8, 2017

Fix order of children for TK_ASSIGN and TK_CALL in the scope pass.

750e155

jemc mentioned this pull request Sep 11, 2017

adapt to new String.join - now accepting Iterator jemc/ponycc#6

Closed

Praetonus approved these changes Sep 12, 2017

View reviewed changes

Fix lambda inference when an argument to a call on a callable object.

4fb0253

SeanTAllen merged commit 2bca9dc into master Sep 13, 2017

ponylang-main added a commit that referenced this pull request Sep 13, 2017

Update CHANGELOG for PR #2168 [skip ci]

0f5de57

SeanTAllen deleted the feature/lambda-inference branch September 13, 2017 19:34

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Inference of lambda type and array element type from an antecedent (RFC 45). #2168

Inference of lambda type and array element type from an antecedent (RFC 45). #2168

jemc commented Aug 15, 2017

jemc commented Aug 15, 2017

jemc commented Aug 16, 2017

Praetonus Aug 16, 2017

jemc Aug 27, 2017

Praetonus Aug 16, 2017

jemc Aug 27, 2017

Praetonus Aug 16, 2017

jemc Aug 27, 2017

jemc commented Aug 27, 2017 •

edited

Loading

SeanTAllen commented Aug 27, 2017

jemc commented Aug 27, 2017

SeanTAllen commented Aug 27, 2017

Praetonus commented Aug 30, 2017

jemc commented Sep 8, 2017

Theodus commented Sep 8, 2017

jemc commented Sep 10, 2017

jemc commented Sep 10, 2017

SeanTAllen commented Sep 12, 2017

jemc commented Sep 13, 2017 •

edited

Loading

SeanTAllen commented Sep 13, 2017

jemc commented Sep 13, 2017

SeanTAllen commented Sep 13, 2017

Inference of lambda type and array element type from an antecedent (RFC 45). #2168

Inference of lambda type and array element type from an antecedent (RFC 45). #2168

Conversation

jemc commented Aug 15, 2017

jemc commented Aug 15, 2017

jemc commented Aug 16, 2017

Praetonus Aug 16, 2017

Choose a reason for hiding this comment

jemc Aug 27, 2017

Choose a reason for hiding this comment

Praetonus Aug 16, 2017

Choose a reason for hiding this comment

jemc Aug 27, 2017

Choose a reason for hiding this comment

Praetonus Aug 16, 2017

Choose a reason for hiding this comment

jemc Aug 27, 2017

Choose a reason for hiding this comment

jemc commented Aug 27, 2017 • edited Loading

SeanTAllen commented Aug 27, 2017

jemc commented Aug 27, 2017

SeanTAllen commented Aug 27, 2017

Praetonus commented Aug 30, 2017

jemc commented Sep 8, 2017

Theodus commented Sep 8, 2017

jemc commented Sep 10, 2017

jemc commented Sep 10, 2017

SeanTAllen commented Sep 12, 2017

jemc commented Sep 13, 2017 • edited Loading

SeanTAllen commented Sep 13, 2017

jemc commented Sep 13, 2017

SeanTAllen commented Sep 13, 2017

jemc commented Aug 27, 2017 •

edited

Loading

jemc commented Sep 13, 2017 •

edited

Loading