Implement the illegal_floating_point_literal_pattern compat lint #41293

est31 · 2017-04-14T01:11:36Z

Adds a future-compatibility lint for the [breaking-change] introduced by issue #41620 . cc issue #41255 .

rust-highfive · 2017-04-14T01:11:46Z

(rust_highfive has picked a reviewer for you, use r? to override)

bors · 2017-04-14T05:01:24Z

☔ The latest upstream changes (presumably #41294) made this pull request unmergeable. Please resolve the merge conflicts.

nikomatsakis · 2017-04-14T13:25:24Z

I'll spin up a crater run. Thanks @est31!

nagisa · 2017-04-14T13:33:02Z

src/librustc_const_eval/pattern.rs

+                    format!("floating point constants cannot be used in patterns"));
+            }
+            _ => {}
+        }
        match const_cx.eval(expr) {


This branch may be dead now.

nagisa · 2017-04-14T13:34:51Z

src/test/run-pass/issue-7222.rs


    match 0.0 {
-        0.0 ... FOO => (),
+        0.0 ... FOOFL => (),


This, i feel, should keep working just fine. It is just a range and testing that floats fall into some range is encouraged even.

I'm neutral about ranges, will adopt whatever the consensus is.

nagisa · 2017-04-14T13:36:05Z

src/test/run-pass/union/union-pat-refutability.rs

@@ -29,7 +29,7 @@ fn is_zero(v: Value) -> bool {
    unsafe {
        match v {
            Value { tag: Tag::I, u: U { i: 0 } } => true,
-            Value { tag: Tag::F, u: U { f: 0.0 } } => true,
+            Value { tag: Tag::F, u: U { u: 0 } } => true,


Similarly here. This code is already annotated with unsafe and all.

Although i guess one could argue that people should just match on the bit pattern (i.e. The integer variant)

It is annotated with unsafe, but unsafe doesn't allow everything, so IMO its okay to disallow it. Especially as making unsafe feel more comfortable to write than non unsafe code is not really a priority of Rust.

nikomatsakis · 2017-04-14T17:39:57Z

Crater report: https://gist.github.com/nikomatsakis/b482d33f058a8dbccc6a0def6545ed7f

8600 crates tested: 6541 working / 1929 broken / 35 regressed / 11 fixed / 84 unknown.

But the 11 fixed sounds a bit suspicious. I haven't looked in detail at the results, could well be a lot of false positives etc.

leoyvens · 2017-04-15T00:40:40Z

I count 15 legit regressions out of the 35. Those are:
fann-0.1.6 - Use of -1.0.
libstat-0.1.1 - Use of 0..
image-0.12.3 - Use of 0.0.
tfidf-0.3.0 - Use of 0.0f32.
clippy-lints-0.0.123 - Use of let Ok(0.0) = foo.
wankel-0.0.1 - Use of 0.0 and 1.0.
cargo-incremental-0.1.13 - Use of 0.0.
geodate-0.2.1 - Big user of matching float ranges, 22 errors in total.
capnpc-0.8.5 - Use of 0.0.
liquid-0.9.1 - Use of 1f32.
humansize-1.0.1 - Use of 0.0.
rand-0.3.15 - Use of 0.0 ... 1.0 range.
probability-0.15.6 - Use of 1.0.
rust-3d-0.5.4 - Use of 0.0.
basichll-0.3.1 -Use of 0.0.

nikomatsakis · 2017-04-18T14:06:40Z

@leodasvacas thanks for that! The majority, then, are "equality" matches, right?

nikomatsakis · 2017-04-18T14:07:01Z

i.e., as opposed to 0.0 ... 1.0-style ranges -- although I guess those have an equality component to them too.

nikomatsakis · 2017-04-18T14:21:07Z

cc @rust-lang/lang -- note that @est31 has kindly quantified the breakage here. I'm not sure how I feel about this. In some ways, I'd like to permit matches over floating points, which clearly have some utility, but definitely it raises some thorny questions (e.g., matching on 0.0 -- what should happen with -0?). (Once we have matching around associated constants and the like, we won't be able to rule out the ambiguous cases like NaN or 0.0 statically.)

Still, I kind of feel like maybe we should carve out special-case rules here around floats and deal with it, even if that introduces a measure of inconsistency. I guess we have to think about how this may interact with constant evaluation in general.

leoyvens · 2017-04-18T14:28:31Z

@nikomatsakis Yes, I think only geodate and rand used ranges.

aidanhs · 2017-04-18T22:36:09Z

Just throwing out that this will also regress ayzim which matches on 0f64 and 1f64 (I really need to keep that updated so I benefit from crater runs).

nagisa · 2017-04-24T21:27:25Z

The crater was run and the results posted, so removing the needs-crater tag.

aturon · 2017-04-25T23:37:32Z

Nominating for discussion in team meeting.

nikomatsakis · 2017-04-28T20:12:30Z

@est31 after some discussion in the @rust-lang/lang meeting, I think we decided we'd like to start issuing warnings (not deny-by-default) for this case. For that, we'll need a separate tracking issue etc. Do you think you can update this PR to do that?

withoutboats · 2017-05-04T19:15:34Z

It seems like we should have the same rules for now (no floats) and if its a problem we can expedite finding a permanent solution.

Imagining that we allow float consts in ranges, if we use PartialOrd NaN doesn't fall into NaN ... NaN, right? I guess you wouldn't be able to construct that range because its ends wouldn't be properly ordered?

These can all be transformed to conditionals right? Not saying its nice but it does work.

nikomatsakis · 2017-05-04T19:24:25Z

@withoutboats

Imagining that we allow float consts in ranges, if we use PartialOrd NaN doesn't fall into NaN ... NaN, right?

Yes. The main question is, is there any reason that this is bad? (I don't see one.)

withoutboats · 2017-05-04T19:31:26Z

just that float equality is a foot gun. I don't really have a strong opinion about what we do here. I guess procedurally I'm a bit worried we are capriciously reversing course on an RFC.

nikomatsakis · 2017-05-04T19:47:41Z

@withoutboats

Well, I don't think the RFC addressed range literals at all. The focus was purely on equality and nobody brought up ... patterns. I guess the main question is whether one should be able to expect that C...C is the same as C.

(The RFC also reported 6 regressions, because it was based on flawed data, since literals were accidentally overlooked.)

(My view is that in such cases, procedurally, the @rust-lang/lang team can make minor adjustments, within the spirit of the original consensus. If you don't feel this is such a case, I'd not necessarily be opposed to issuing the warnings, and perhaps opening another RFC to explicitly allow .. and ... patterns for floating point types -- and maybe any PartialOrd type?)

withoutboats · 2017-05-04T19:52:33Z

I guess the main question is whether one should be able to expect that C...C is the same as C.

A misguided user might also forget to handle NaN and write something like:

match x {
    0.0 ... INF => { ... }
    NEG_INF ... 0.0 => { ... }
    _ => unreachable!()
}

Of course depending on the context panicking might be better than whatever would happen if they instead wrote this conditional:

if x >= 0.0 { ... }
else { ... }

withoutboats · 2017-05-04T19:56:01Z

@nikomatsakis I guess one thing I do feel kind of strongly about is that we have the same policy for floats in ranges as we do for match in general (to make this easier to teach), so if we're going to allow those I'd prefer to reach a decision on all floats.

carols10cents · 2017-05-08T14:20:15Z

What are the next steps on this PR? @withoutboats @nikomatsakis

nikomatsakis · 2017-05-08T18:10:24Z

I think we should go forward with the PR for now. I'm still vaguely unhappy about the whole situation, but warnings seem ok and are a decent way to gain feedback too.

nikomatsakis · 2017-05-08T18:10:28Z

@bors r+

bors · 2017-05-08T18:10:29Z

📌 Commit 37fb676 has been approved by nikomatsakis

bors · 2017-05-08T21:38:56Z

⌛ Testing commit 37fb676 with merge 497d628...

bors · 2017-05-08T23:02:27Z

💔 Test failed - status-travis

alexcrichton · 2017-05-08T23:07:22Z

@bors: retry

Tracking issue for spurious network failures on bots #40474 - 500 error from crates.io

…omatsakis Implement the illegal_floating_point_literal_pattern compat lint Adds a future-compatibility lint for the [breaking-change] introduced by issue rust-lang#41620 . cc issue rust-lang#41255 .

Rollup of 8 pull requests - Successful merges: #41293, #41520, #41827, #41828, #41833, #41836, #41838, #41842 - Failed merges:

bors · 2017-05-09T05:58:18Z

⌛ Testing commit 37fb676 with merge bedd7da...

@alexbool

Remove debug message It was added by me in rust-lang#41293. Sorry about that! Thanks goes to @alexbool for finding it.

@alexbool

Remove debug message It was added by me in rust-lang#41293. Sorry about that! Thanks goes to @alexbool for finding it.

rust-highfive assigned pnkfelix Apr 14, 2017

est31 changed the title ~~Floating literal match~~ Disallow floating point literals in match Apr 14, 2017

est31 mentioned this pull request Apr 14, 2017

Matching on floating-point literal values is totally allowed and shouldn't be #41255

Closed

est31 force-pushed the floating_literal_match branch from 6ec00d2 to 0e33daa Compare April 14, 2017 05:01

nikomatsakis added the S-waiting-on-crater Status: Waiting on a crater run to be completed. label Apr 14, 2017

nagisa reviewed Apr 14, 2017

View reviewed changes

shepmaster added the S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. label Apr 14, 2017

est31 force-pushed the floating_literal_match branch from 272fb0a to 14c1608 Compare April 18, 2017 00:42

nagisa removed the S-waiting-on-crater Status: Waiting on a crater run to be completed. label Apr 24, 2017

aturon added the I-nominated label Apr 25, 2017

nikomatsakis removed the T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. label Apr 27, 2017

arielb1 removed the I-nominated label Apr 27, 2017

est31 force-pushed the floating_literal_match branch from 14c1608 to e5f21bc Compare April 29, 2017 05:01

frewsxcv mentioned this pull request May 9, 2017

Rollup of 10 pull requests #41845

Closed

frewsxcv mentioned this pull request May 9, 2017

Rollup of 8 pull requests #41846

Merged

bors added a commit that referenced this pull request May 9, 2017

Auto merge of #41846 - frewsxcv:rollup, r=frewsxcv

bedd7da

Rollup of 8 pull requests - Successful merges: #41293, #41520, #41827, #41828, #41833, #41836, #41838, #41842 - Failed merges:

bors merged commit 37fb676 into rust-lang:master May 9, 2017

Arnavion mentioned this pull request May 10, 2017

ERROR:rustc_lint::builtin: span mc spanspam when compiling clippy_lints rust-lang/rust-clippy#1743

Closed

SimonSapin mentioned this pull request May 10, 2017

Tracking issue for illegal_floating_point_constant_pattern compatibility lint #36890

Closed

3 tasks

est31 mentioned this pull request May 10, 2017

Remove debug message #41889

Merged

frewsxcv added a commit to frewsxcv/rust that referenced this pull request May 11, 2017

Rollup merge of rust-lang#41889 - est31:master, r=estebank

42a3522

Remove debug message It was added by me in rust-lang#41293. Sorry about that! Thanks goes to @alexbool for finding it.

frewsxcv added a commit to frewsxcv/rust that referenced this pull request May 11, 2017

Rollup merge of rust-lang#41889 - est31:master, r=estebank

00f6e39

Remove debug message It was added by me in rust-lang#41293. Sorry about that! Thanks goes to @alexbool for finding it.

dwrensha mentioned this pull request May 13, 2017

update for latest rustc rust-lang/miri#167

Merged

jashank mentioned this pull request Jun 26, 2017

Quiesce float literal lint on beta/nightly. xurtis/atoms#10

Merged

saethlin mentioned this pull request Oct 3, 2023

make matching on NaN a hard error, and remove the rest of illegal_floating_point_literal_pattern #116284

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Implement the illegal_floating_point_literal_pattern compat lint #41293

Implement the illegal_floating_point_literal_pattern compat lint #41293

est31 commented Apr 14, 2017 •

edited

Loading

rust-highfive commented Apr 14, 2017

bors commented Apr 14, 2017

nikomatsakis commented Apr 14, 2017

nagisa Apr 14, 2017

nagisa Apr 14, 2017

est31 Apr 14, 2017

nagisa Apr 14, 2017

nagisa Apr 14, 2017

est31 Apr 14, 2017

nikomatsakis commented Apr 14, 2017

leoyvens commented Apr 15, 2017 •

edited

Loading

nikomatsakis commented Apr 18, 2017

nikomatsakis commented Apr 18, 2017

nikomatsakis commented Apr 18, 2017

leoyvens commented Apr 18, 2017

aidanhs commented Apr 18, 2017

nagisa commented Apr 24, 2017

aturon commented Apr 25, 2017

nikomatsakis commented Apr 28, 2017

withoutboats commented May 4, 2017 •

edited

Loading

nikomatsakis commented May 4, 2017

withoutboats commented May 4, 2017 •

edited

Loading

nikomatsakis commented May 4, 2017 •

edited

Loading

withoutboats commented May 4, 2017

withoutboats commented May 4, 2017

carols10cents commented May 8, 2017

nikomatsakis commented May 8, 2017

nikomatsakis commented May 8, 2017

bors commented May 8, 2017

bors commented May 8, 2017

bors commented May 8, 2017

alexcrichton commented May 8, 2017

bors commented May 9, 2017

Implement the illegal_floating_point_literal_pattern compat lint #41293

Implement the illegal_floating_point_literal_pattern compat lint #41293

Conversation

est31 commented Apr 14, 2017 • edited Loading

rust-highfive commented Apr 14, 2017

bors commented Apr 14, 2017

nikomatsakis commented Apr 14, 2017

nagisa Apr 14, 2017

Choose a reason for hiding this comment

nagisa Apr 14, 2017

Choose a reason for hiding this comment

est31 Apr 14, 2017

Choose a reason for hiding this comment

nagisa Apr 14, 2017

Choose a reason for hiding this comment

nagisa Apr 14, 2017

Choose a reason for hiding this comment

est31 Apr 14, 2017

Choose a reason for hiding this comment

nikomatsakis commented Apr 14, 2017

leoyvens commented Apr 15, 2017 • edited Loading

nikomatsakis commented Apr 18, 2017

nikomatsakis commented Apr 18, 2017

nikomatsakis commented Apr 18, 2017

leoyvens commented Apr 18, 2017

aidanhs commented Apr 18, 2017

nagisa commented Apr 24, 2017

aturon commented Apr 25, 2017

nikomatsakis commented Apr 28, 2017

withoutboats commented May 4, 2017 • edited Loading

nikomatsakis commented May 4, 2017

withoutboats commented May 4, 2017 • edited Loading

nikomatsakis commented May 4, 2017 • edited Loading

withoutboats commented May 4, 2017

withoutboats commented May 4, 2017

carols10cents commented May 8, 2017

nikomatsakis commented May 8, 2017

nikomatsakis commented May 8, 2017

bors commented May 8, 2017

bors commented May 8, 2017

bors commented May 8, 2017

alexcrichton commented May 8, 2017

bors commented May 9, 2017

est31 commented Apr 14, 2017 •

edited

Loading

leoyvens commented Apr 15, 2017 •

edited

Loading

withoutboats commented May 4, 2017 •

edited

Loading

withoutboats commented May 4, 2017 •

edited

Loading

nikomatsakis commented May 4, 2017 •

edited

Loading