Improve error message for non-exhaustive patterns #31020

regexident · 2016-01-19T07:43:34Z

Changes error message from displaying first found missing constructor witness to showing up to 10, if necessary.

Fixes issue #16884.

Changes error message from displaying first found missing constructor witness to showing up to 10, if necessary. Fixes issue #16884.

rust-highfive · 2016-01-19T07:43:49Z

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @brson (or someone else) soon.

If any changes to this PR are deemed necessary, please add them as extra commits. This ensures that the reviewer can see what has changed since they last reviewed the code. Due to the way GitHub handles out-of-date commits, this should also make it reasonably obvious what issues have or haven't been addressed. Large or tricky changes may require several passes of review and changes.

Please see the contribution instructions for more information.

mitaa · 2016-01-19T08:06:44Z

src/librustc/middle/check_match.rs

                    span_err!(cx.tcx.sess, sp, E0004,
                        "non-exhaustive patterns: `{}` not covered",
-                        pat_to_string(witness)
+                        pattern_strings.join("`, `")


Maybe delimit the last element with "and"? I think this reads a bit weird otherwise.

Good point. Was wondering about this, too.

Would something like this look good to you?

let joined_string = match pattern_strings { [pattern] => pattern, [head.., tail] => head.join("`, `") + "` and `" + tail }

This also uses slice patterns, and apparently requires the "advanced_slice_pattern" feature gate.

I'm not sure about the optimal solution, maybe something like this?

let joined_string = match pattern_strings.split_last() { Some((tail, head)) => { if head.is_empty() { tail.clone() } else { head.join("`, `") + "` and `" + tail } }, None => unreachable!() };

More human-readable error message showing ellipsis for excessively long witness lists.

brson · 2016-01-19T23:41:10Z

I'm guessing you are making this change because it's annoying to fix the error just to be given a nearly identical message. I think this is a fine reason.

I might suggest that 10 is too many though, based on the rule of thumb that people can generally keep 7 +/3 'things' straight at a time. @nikomatsakis do we have any precedent for how many 'things' to display before cutting errors off?

It may be worth saying "... and more (not shown)" or something for the case where we do cut them off. @nikomatsakis do we have existing precedent for what language to use for that?

regexident · 2016-01-20T02:05:07Z

Hi brson, thanks for the comment!

I totally agree, that 10 is probably quite a bit too much. I guess even three would be suffice in the end. After all if you're missing that many patterns in the first place, then you might want to consult the subject's type definition anyway, in order to proceed (instead of like copy/paste the missing patterns from the error message). As such, I suppose, even a limit as low as 3 should work here. After all it should do, is make it clear that there's more than one missing case, when there is.

I quickly searched the project for "and more" idioms but couldn't find anything that'd be of a similar enough scenario to imitate here. The phrases "and more", "not shown" as well as "..." (as a list ellipsis) appear to actually not even exist within src/test/….

Maybe @steveklabnik knows more about such a precedent or has suggestions for a proper fit from the overall point of view of the docs?

Further more I used an actual Unicode ellipsis glyph (…), but should be using three periods (...) instead, I suppose, as that's what's used everywhere else. Makes sense to stay within ASCII for mostly terminal/console use, I guess.

nikomatsakis · 2016-01-20T21:48:30Z

Hmm, I was trying to think of good precedent, but I'm coming up blank. I think there are other places we've added semi-arbitrary cutoffs, but I'm not sure where that is right now. In particular, I was looking at the suggestions for traits to import in method dispatch, but those seem to print all applicable options (those are in librustc_typeck/check/method/suggest.rs).

Still, I agree that just printing 2 or 3 is plenty. If we know the full set, you could print something like "...and 7 more possibilities." Otherwise, it feels like we ought to be able to just insert some wording to make it clear that these are just representative, and there are more possibilities than the ones we've printed.

…atterns for excessive cases.

regexident · 2016-01-21T21:34:04Z

I adjusted the error message to now print in either of these formats:

Print … non-exhaustive patterns: Red not covered. (1)
Print … non-exhaustive patterns: East, SouthandWest not covered. (1...3)
Print … non-exhaustive patterns: Second, Third, Fourth and 8 more not covered. (4…)

depending on the number of missing patterns.

regexident · 2016-01-22T14:54:27Z

@brson, @nikomatsakis: Any other changes you'd like me to make to this?
Or put differently: It's my first PR on /rust-lang/rust/, what's the next steps from here?

nikomatsakis · 2016-01-22T15:40:21Z

@regexident sounds awesome! :)

nikomatsakis · 2016-01-22T15:43:23Z

src/librustc/middle/check_match.rs

    let used_constructors: Vec<Constructor> = rows.iter()
        .flat_map(|row| pat_constructors(cx, row[0], left_ty, max_slice_length))
        .collect();
    all_constructors(cx, left_ty, max_slice_length)
        .into_iter()
-        .find(|c| !used_constructors.contains(c))
+        .filter(|c| !used_constructors.contains(c)).collect()


Nit: can you put the .collect() on the next line, to match the formatting above?

nikomatsakis · 2016-01-22T15:45:39Z

This looks good to me modulo the nits above. @regexident can you update with those small nits addressed (and then leave a comment, since otherwise I don't get any notifications)?

regexident · 2016-01-22T15:54:16Z

Sure, I'm at it already. Thanks for the comments!

regexident · 2016-01-22T16:16:25Z

@nikomatsakis:
All done.
👍

brson · 2016-01-27T01:59:30Z

@bors r+ Thanks @regexident

bors · 2016-01-27T01:59:31Z

📌 Commit 2c7a19a has been approved by brson

bors · 2016-01-27T07:32:01Z

⌛ Testing commit 2c7a19a with merge aba11b3...

Changes error message from displaying first found missing constructor witness to showing up to 10, if necessary. Fixes issue #16884.

bors · 2016-01-27T10:17:47Z

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-debug-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-x-android-t, auto-linux-cross-opt, auto-linux-musl-64-opt, auto-mac-32-opt, auto-mac-64-nopt-t, auto-mac-64-opt, auto-win-gnu-32-nopt-t, auto-win-gnu-32-opt, auto-win-gnu-64-nopt-t, auto-win-gnu-64-opt, auto-win-msvc-32-opt, auto-win-msvc-64-opt

regexident · 2016-01-27T10:35:39Z

✋ @brson ✋ @nikomatsakis

🙇

the assert was invalidated by PR rust-lang#31020 Fixes rust-lang#31561

@eddyb

the assert was invalidated by PR #31020 Fixes #31561 r? @eddyb

the assert was invalidated by PR rust-lang#31020 Fixes rust-lang#31561

frewsxcv · 2016-09-12T01:42:29Z

An issue was just opened relevant to this: #36410

Improve error message for non-exhaustive patterns

6100743

Changes error message from displaying first found missing constructor witness to showing up to 10, if necessary. Fixes issue #16884.

rust-highfive assigned brson Jan 19, 2016

regexident mentioned this pull request Jan 19, 2016

Display all witnesses for non-exhaustive patterns #16884

Closed

mitaa reviewed Jan 19, 2016
View reviewed changes

Refined error message.

48e8326

More human-readable error message showing ellipsis for excessively long witness lists.

Refined error message to truncate at 3 and hint at number of hidden p…

70692ce

…atterns for excessive cases.

nikomatsakis reviewed Jan 22, 2016
View reviewed changes

Adjusted PR to better match project's coding style.

2c7a19a

bors added a commit that referenced this pull request Jan 27, 2016

Auto merge of #31020 - regexident:fix_16884, r=brson

aba11b3

Changes error message from displaying first found missing constructor witness to showing up to 10, if necessary. Fixes issue #16884.

bors merged commit 2c7a19a into rust-lang:master Jan 27, 2016

regexident deleted the fix_16884 branch January 27, 2016 10:35

brson added the relnotes Marks issues that should be documented in the release notes of the next release. label Jan 29, 2016

mitaa mentioned this pull request Mar 4, 2016

rustc nightly panics on refutable pattern #31561

Closed

arielb1 added a commit to arielb1/rust that referenced this pull request Mar 4, 2016

remove wrong assert in check_match

26a0afb

the assert was invalidated by PR rust-lang#31020 Fixes rust-lang#31561

arielb1 mentioned this pull request Mar 4, 2016

remove wrong assert in check_match #32041

Merged

bors added a commit that referenced this pull request Mar 6, 2016

Auto merge of #32041 - arielb1:multiply-inexhaustive, r=eddyb

80922a7

the assert was invalidated by PR #31020 Fixes #31561 r? @eddyb

alexcrichton pushed a commit to alexcrichton/rust that referenced this pull request Mar 16, 2016

remove wrong assert in check_match

d66fabb

the assert was invalidated by PR rust-lang#31020 Fixes rust-lang#31561

apasel422 mentioned this pull request May 16, 2016

Release note PR link incorrect (suggest enum variants for non-exhaustive match patterns) #32975

Closed

jfirebaugh mentioned this pull request Sep 12, 2016

E0004 lists incorrect secondary witnesses #36410

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Improve error message for non-exhaustive patterns #31020

Improve error message for non-exhaustive patterns #31020

regexident commented Jan 19, 2016

rust-highfive commented Jan 19, 2016

mitaa Jan 19, 2016

regexident Jan 19, 2016

mitaa Jan 19, 2016

brson commented Jan 19, 2016

regexident commented Jan 20, 2016

nikomatsakis commented Jan 20, 2016

regexident commented Jan 21, 2016

regexident commented Jan 22, 2016

nikomatsakis commented Jan 22, 2016

nikomatsakis Jan 22, 2016

nikomatsakis commented Jan 22, 2016

regexident commented Jan 22, 2016

regexident commented Jan 22, 2016

brson commented Jan 27, 2016

bors commented Jan 27, 2016

bors commented Jan 27, 2016

bors commented Jan 27, 2016

regexident commented Jan 27, 2016

frewsxcv commented Sep 12, 2016

Improve error message for non-exhaustive patterns #31020

Improve error message for non-exhaustive patterns #31020

Conversation

regexident commented Jan 19, 2016

rust-highfive commented Jan 19, 2016

mitaa Jan 19, 2016

Choose a reason for hiding this comment

regexident Jan 19, 2016

Choose a reason for hiding this comment

mitaa Jan 19, 2016

Choose a reason for hiding this comment

brson commented Jan 19, 2016

regexident commented Jan 20, 2016

nikomatsakis commented Jan 20, 2016

regexident commented Jan 21, 2016

regexident commented Jan 22, 2016

nikomatsakis commented Jan 22, 2016

nikomatsakis Jan 22, 2016

Choose a reason for hiding this comment

nikomatsakis commented Jan 22, 2016

regexident commented Jan 22, 2016

regexident commented Jan 22, 2016

brson commented Jan 27, 2016

bors commented Jan 27, 2016

bors commented Jan 27, 2016

bors commented Jan 27, 2016

regexident commented Jan 27, 2016

frewsxcv commented Sep 12, 2016