[New DIP] Deprecation and removal of implicit integer and character literal conversion to bool by JinShil · Pull Request #99 · dlang/DIPs

JinShil · 2017-11-12T15:00:19Z

~~Still not quite ready, but should be in a few days.~~

PetarKirov · 2017-11-13T10:40:07Z

We also need to consider array literals. There's code out there that explicitly relies on the implicit int->bool conversion. For example, BitArray's examples would look considerably less readable if we disallow that.

JinShil · 2017-11-13T10:55:05Z

We also need to consider array literals. There's code out there that explicitly relies on the implicit int->bool conversion. For example, BitArray's examples would look considerably less readable if we disallow that.

Good point! I think we can overload the BitArray constructors to mitigate that specific case, but consider other cases like this:

bool[] values = [1,0,1,0,1,1];

Users can always do:

bool[] values = cast(bool[])[1,0,1,0,1,1];

That's not too bad, IMO, but I wonder how much disruption that would cause in the D ecosystem.

Anyway, I probably need to add this to the DIP. Thanks.

PetarKirov · 2017-11-13T10:59:38Z

I wonder if it's a good idea to allow implicit conversion for functions that are not overloaded.

JinShil · 2017-11-13T11:03:56Z

I wonder if it's a good idea to allow implicit conversion for functions that are not overloaded.

I assume you mean just for int-->bool, not in general, yes?

PetarKirov · 2017-11-13T11:47:49Z

I assume you mean just for int-->bool, not in general, yes?

Yes, only when a function is not overloaded (and not a template) and the parameter is of type bool or a type that has alias boolField this.

mdparker · 2017-11-14T09:44:36Z

Normally, I'd prefer for a DIP to be in a state that's ready to begin the Draft Review before the PR is submitted. No big deal, though, since that isn't documented anywhere at the moment. I'll just need you to let me know when this one and #97 are ready for Draft Review.

I also need to ask for your patience. There will likely be a bit of a delay before I initiate any new DIP reviews. We're currently looking at ways to improve upon the existing process. I don't want to start anything new until that's settled. Hopefully, it won't be too long before we get there.

Thanks!

JinShil · 2017-11-14T09:47:14Z

@mdparker Understood. #97 is ready to go whenever you are.

JinShil · 2017-11-14T09:48:42Z

Closing for now, as this needs quite a bit more work.

JinShil · 2017-11-16T03:52:06Z

This is ready for Draft Review.

… and character literals to bool

mdparker · 2018-03-07T13:40:24Z

@JinShil We can move on this one first if you prefer. Let me know. Whichever you pick to move on second will be delayed a bit longer. I'll want to see if any of the others are ready to go after moving your first one through.

mdparker · 2018-06-10T14:31:28Z

DIPs/DIP1xxy-mvf.md

+
+## Abstract
+
+This DIP proposes deprecation and removal of implicit conversion from integer and character literals to `bool` in order to close a hole in the type system and resolve overload resolution bugs.  For the purpose of this document, *Integer literals* are literals of type `int`, `uint`, `long` or `ulong` and *Character literals* are literals of type `char`, `wchar`, or `dchar`.


"proposes deprecation" => proposes thedeprecation

"to bool in order to close" => "to bool, thereby closing

"and resolve" => and resolving

mdparker · 2018-06-10T14:33:48Z

DIPs/DIP1xxy-mvf.md

+
+This DIP proposes deprecation and removal of implicit conversion from integer and character literals to `bool` in order to close a hole in the type system and resolve overload resolution bugs.  For the purpose of this document, *Integer literals* are literals of type `int`, `uint`, `long` or `ulong` and *Character literals* are literals of type `char`, `wchar`, or `dchar`.
+
+### Links


Please update the DIP to use the new template layout.

mdparker · 2018-06-10T14:37:13Z

DIPs/DIP1xxy-mvf.md

+    boolValue = 1UL;
+}
+```
+The implicit conversion of character literals to `bool` is questionable, as there is no intrinsic relationship between  characters `'\0'` and `'\1'`, and boolean states false and true.  This behavior may actually be simply an unintended consequence of [integer promotion in D](https://dlang.org/spec/type.html#integer-promotions) and implicit integer literal conversion to `bool`.  The reader of such code would be right to question whether such expressions are actually programming errors or the author's intent, as they circumvent the type system without an explicit cast.


"between characters" => between the characters

"and boolean" => and the boolean

"false and true" => false and true

"may actually be simply an" => may be an

mdparker · 2018-06-10T14:40:03Z

DIPs/DIP1xxy-mvf.md

+```
+The implicit conversion of character literals to `bool` is questionable, as there is no intrinsic relationship between  characters `'\0'` and `'\1'`, and boolean states false and true.  This behavior may actually be simply an unintended consequence of [integer promotion in D](https://dlang.org/spec/type.html#integer-promotions) and implicit integer literal conversion to `bool`.  The reader of such code would be right to question whether such expressions are actually programming errors or the author's intent, as they circumvent the type system without an explicit cast.
+
+It is not clear how the implicit conversion of integer literals to `bool` came to be a feature in D.  It may be a historical leftover from a time when [D had a 1-bit integer, `bit`, that was later replaced by `bool`](https://issues.dlang.org/show_bug.cgi?id=9999#c9).  It may also be a carry-over from D's roots in C, the way conditionals are evaluated in C, and the conventions that were employed over the years to use integer values to compensate for C's lack of a boolean type.  Regardless, unlike C, D has a boolean type `bool`, boolean literals `false` and `true`, and also supports explicit casting of integers to `bool`, so implicit conversion from integer literals to `bool` is unnecessary, and even problematic as demonstrated below.


I think everything before and including "Regardless" can be deleted. Speculation on possible origins is superfluous. If you had something concrete against which you needed to justify the change, then that would be different. Maybe something will come up in review.

mdparker · 2018-06-10T14:43:29Z

DIPs/DIP1xxy-mvf.md

+D's existing implementation implicitly converts integer and character literals that evaluate to 0 and 1 to `false` and `true` respectively.
+
+### Example A
+The following Example currently compiles in the existing implementation.


"Example" => example

mdparker · 2018-06-10T15:17:01Z

DIPs/DIP1xxy-mvf.md

+
+Also, like the former, it would still break code relying on the existing overload resolution behavior illustrated in [Example B](#example-b) and [Example C](#example-c), resulting in a deprecation process similar, if not the same, to that of the proposed solution, but would not close the hole in the type system.
+
+This solution was also not analyzed in enough detail to determine whether it would even scale.


"would even scale" => would scale

mdparker · 2018-06-10T15:18:14Z

DIPs/DIP1xxy-mvf.md

+
+The proposed implementation is to be rolled out through a sequence of steps to minimize any disruption.
+
+1.  If and when this DIP is approved, an announcement will be made on the D programming language forum to raise awareness of the change in the community.  Users will have a link to this DIP and the implementation so they can understand how to mitigate breakage in their code, and monitor the timing of the subsequent releases.


"If and when this DIP is approved, an"=> An

mdparker · 2018-06-10T15:19:42Z

DIPs/DIP1xxy-mvf.md

+
+## Copyright & License
+
+Copyright (c) 2017 by the D Language Foundation


"2017" => 2018

mdparker · 2018-06-10T15:20:23Z

DIPs/DIP1xxy-mvf.md

+
+2.  After approval of this DIP, code in DMD's test suite, Phobos, and DRuntime will be updated with the appropriate remedies documented in [Description](#description) to avoid deprecation warnings or errors in the stage 1 and stage 2 releases.  Preemptive pull requests anticipating the change proposed in this DIP have already been merged for [Phobos](https://github.com/dlang/phobos/pull/5019) and [DRuntime](https://github.com/dlang/druntime/pull/1732), but another pass will need to be made.
+
+4.  After the changes in step 2 have been pulled, stage 1 can be pulled.  From that point on the compiler will emit deprecation warnings when an implicit conversion from an integer literal to `bool` is encountered.  This will precisely identify instances in users' code that will no longer compile when stage 2 is released, but will still perform the conversion, buying users time to update their code. The deprecation will remain in place for a time period to be decided by the language authors at the time of this DIP's formal review.


"been pulled, stage 1 can be pulled" => been merged, state 1 can be merged

"the language authors" => the language maintainers
That's the "official" terminology now.

"formal review" => formal assessment
Ditto.

mdparker · 2018-06-10T15:21:52Z

DIPs/DIP1xxy-mvf.md

+
+This solution was also not analyzed in enough detail to determine whether it would even scale.
+
+### Breaking changes / deprecation process


Expected breakage needs to be iterated here, even when it is mentioned elsewhere in the document.

JinShil closed this Nov 14, 2017

JinShil reopened this Nov 16, 2017

[New DIP] Deprecation and removal of implicit conversion from integer…

12eb442

… and character literals to bool

JinShil mentioned this pull request Nov 20, 2017

Fix Issues 9999 & 10560 - [DIP1015] Deprecate & Remove implicit integer & character conversion to bool dlang/dmd#7310

Closed

JinShil mentioned this pull request Jan 4, 2018

Fix Issue 8006: Property binary assignment operators dlang/dmd#7079

Closed

mdparker requested changes Jun 10, 2018

View reviewed changes

JinShil added 2 commits June 11, 2018 18:50

Incorporate suggested edits from @mdparker

d827652

Updated to conform to new DIP template

70313f3

mdparker approved these changes Jun 14, 2018

View reviewed changes

mdparker merged commit 19ca5a6 into dlang:master Jun 20, 2018


		## Abstract

		This DIP proposes deprecation and removal of implicit conversion from integer and character literals to `bool` in order to close a hole in the type system and resolve overload resolution bugs. For the purpose of this document, Integer literals are literals of type `int`, `uint`, `long` or `ulong` and Character literals are literals of type `char`, `wchar`, or `dchar`.


		This DIP proposes deprecation and removal of implicit conversion from integer and character literals to `bool` in order to close a hole in the type system and resolve overload resolution bugs. For the purpose of this document, Integer literals are literals of type `int`, `uint`, `long` or `ulong` and Character literals are literals of type `char`, `wchar`, or `dchar`.

		### Links


		Also, like the former, it would still break code relying on the existing overload resolution behavior illustrated in [Example B](#example-b) and [Example C](#example-c), resulting in a deprecation process similar, if not the same, to that of the proposed solution, but would not close the hole in the type system.

		This solution was also not analyzed in enough detail to determine whether it would even scale.


		The proposed implementation is to be rolled out through a sequence of steps to minimize any disruption.

		1. If and when this DIP is approved, an announcement will be made on the D programming language forum to raise awareness of the change in the community. Users will have a link to this DIP and the implementation so they can understand how to mitigate breakage in their code, and monitor the timing of the subsequent releases.


		## Copyright & License

		Copyright (c) 2017 by the D Language Foundation


		2. After approval of this DIP, code in DMD's test suite, Phobos, and DRuntime will be updated with the appropriate remedies documented in [Description](#description) to avoid deprecation warnings or errors in the stage 1 and stage 2 releases. Preemptive pull requests anticipating the change proposed in this DIP have already been merged for [Phobos](https://github.com/dlang/phobos/pull/5019) and [DRuntime](https://github.com/dlang/druntime/pull/1732), but another pass will need to be made.

		4. After the changes in step 2 have been pulled, stage 1 can be pulled. From that point on the compiler will emit deprecation warnings when an implicit conversion from an integer literal to `bool` is encountered. This will precisely identify instances in users' code that will no longer compile when stage 2 is released, but will still perform the conversion, buying users time to update their code. The deprecation will remain in place for a time period to be decided by the language authors at the time of this DIP's formal review.


		This solution was also not analyzed in enough detail to determine whether it would even scale.

		### Breaking changes / deprecation process

Uh oh!

Conversation

JinShil commented Nov 12, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

PetarKirov commented Nov 13, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

JinShil commented Nov 13, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

PetarKirov commented Nov 13, 2017

Uh oh!

JinShil commented Nov 13, 2017

Uh oh!

PetarKirov commented Nov 13, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mdparker commented Nov 14, 2017

Uh oh!

JinShil commented Nov 14, 2017

Uh oh!

JinShil commented Nov 14, 2017

Uh oh!

JinShil commented Nov 16, 2017

Uh oh!

mdparker commented Mar 7, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

JinShil commented Nov 12, 2017 •

edited

Loading

PetarKirov commented Nov 13, 2017 •

edited

Loading

JinShil commented Nov 13, 2017 •

edited

Loading

PetarKirov commented Nov 13, 2017 •

edited

Loading