Add clearer error message using `&str + &str` #39116

mgattozzi · 2017-01-17T04:02:16Z

This is the first part of #39018. One of the common things for new users
coming from more dynamic languages like JavaScript, Python or Ruby is to
use + to concatenate strings. However, this doesn't work that way in
Rust unless the first type is a String. This commit adds a check for
this use case and outputs a new error as well as a suggestion to guide
the user towards the desired behavior. It also adds a new test case to
test the output of the error.

rust-highfive · 2017-01-17T04:03:04Z

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @nrc (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 Due to the way GitHub handles out-of-date commits, this should also make it reasonably obvious what issues have or haven't been addressed.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.

Manishearth · 2017-01-17T04:13:31Z

Mostly looks okay? Would like to see if @jonathandturner has thoughts on this.

Feel free to try and add handling for the other forms. String is a struct so you can search for it.

I've been considering having lang-item-like annotations for types in the stdlib, which let diagnostics check for their existence. We do path matching extensively in clippy, and it would be useful to do something less hacky in rustc.

Manishearth · 2017-01-17T04:11:05Z

src/librustc_typeck/check/op.rs

+                        };
+                    err.span_suggestion(expr.span,
+                        &format!("Your first argument should be a `String`, not `&str`. \
+                                    Try using `to_string()`"),


.to_string()

Manishearth · 2017-01-17T04:11:35Z

src/test/compile-fail/E0369.rs

+    //~^ ERROR 12:18: 12:26: binary operation `+` cannot be applied to type `&'static str` [E0369]
+    //~| NOTE You can't use `+` between a `&str` and `&str`
+    //~| HELP Your first argument should be a `String`, not `&str`. Try using `to_string()`
+    //~| NOTE an implementation of `std::ops::Add` might be missing for `&'static str`


Have a thingy for the suggestion too

bjorn3 · 2017-01-17T07:13:41Z

src/librustc_typeck/check/op.rs

                            span_note!(&mut err, lhs_expr.span,
-                                       "an implementation of `{}` might be missing for `{}`",
+                                        "an implementation of `{}` might be missing for `{}`",


Please remove this extra space.

mgattozzi · 2017-01-17T15:59:56Z

I've updated the code with the above suggestions. @Manishearth how would I go about searching for a String so that I could match against it here.

Manishearth · 2017-01-17T16:19:42Z

This is what we do in clippy to match structs.

https://github.com/Manishearth/rust-clippy/blob/630e0ef96875e0958277674f81921bfaba30bc5f/clippy_lints/src/utils/mod.rs#L140-L179

I think we should just land as-is since that's a bit hacky. I'm filing an issue for the diagnostic-lang-item stuff.

r? @jonathandturner

Manishearth · 2017-01-17T16:45:53Z

Filed #39131 about the diagnostic item

mgattozzi · 2017-01-17T17:22:31Z

In the meantime I think it might be good to get the diagnostic of things like String + String if possible but that could be a separate PR if that's better.

Manishearth · 2017-01-17T17:55:32Z

Actually, you can make this work without the full framework of diagnostics items; just add a #[diagnostic_item = "String"] annotation and then in your code check that cx.tcx.get_attrs(adt_def.def_id) has that attribute.

mgattozzi · 2017-01-17T18:09:53Z

@Manishearth using the code from clippy or just putting it in the code now as is?

Manishearth · 2017-01-17T18:10:28Z

No, we don't want to use the clippy code, it's hacky. Use an attribute as shown above.

Manishearth · 2017-01-17T18:10:37Z

Maybe rustc_diagnostic_item

mgattozzi · 2017-01-17T18:29:20Z

I used ripgrep and git grep to search the code and neither of those attributes exist in the code base so I have no basis of comparison to do it the way you've mentioned.

Manishearth · 2017-01-17T18:32:17Z

I'm saying you invent a new attribute :)

Annotate String with #[rustc_diagnostic_item = "String"]. Then, using tcx.get_attrs(adt_def.def_id), you can check if your TyAdt has that attribute.

You will probably have to add it to libsyntax/feature_gate.rs as well, as a Gated attribute.

Manishearth · 2017-01-17T18:33:50Z

Anyway we can land as-is if you'd prefer too.

sophiajt · 2017-01-17T19:28:39Z

I like the idea.

Can you use a src/test/ui test instead of a compile_fail test? Just so that we can see what the final output looks like a bit better in the PR.

mgattozzi · 2017-01-17T19:37:57Z

@Manishearth ohhhhhh okay I think I see what you're talking about now. I can try when I'm off work.

@jonathandturner I'll switch it over to ui test when I'm off work!

nrc · 2017-01-18T01:00:18Z

src/librustc_typeck/check/op.rs

+            if let TyRef(_, r_ty) = rhs_ty.sty {
+                if l_ty.ty.sty == TyStr && r_ty.ty.sty == TyStr {
+                    span_note!(&mut err, lhs_expr.span,
+                            "You can't use `+` between a `&str` and `&str`");


We usually use the passive voice for error messages, i.e., something like "+ can't be used to concatenate two &str strings" and similarly later on.

Consistency is good. This one is a quick fix.

nrc · 2017-01-18T01:01:11Z

src/librustc_typeck/check/op.rs

+                            _ => format!("<expression>")
+                        };
+                    err.span_suggestion(expr.span,
+                        &format!("Your first argument should be a `String`, not `&str`. \


It would be nice to explain why a String is needed, otherwise this seems a little like boilerplate.

Yeah I can see why it's boilerplate like. I'll put a more descriptive message.

nrc · 2017-01-18T01:02:20Z

src/librustc_typeck/check/op.rs

+                        };
+                    err.span_suggestion(expr.span,
+                        &format!("Your first argument should be a `String`, not `&str`. \
+                                    Try using to_string()"),


I would prefer phrasing like "to_string can be used to create an owned string from a string reference", I actually prefer using to_owned rather than to_string because it better describes what is going on

to_owned or to_string is always a hot button topic. Now they're semantically equivalent at least. I can put to_owned. It's all the same to me and with the rest of the sentence you put it communicates the intent of the function call.

nrc · 2017-01-18T01:03:03Z

src/test/compile-fail/E0369.rs

+    //~| NOTE You can't use `+` between a `&str` and `&str`
+    //~| HELP Your first argument should be a `String`, not `&str`. Try using to_string()
+    //~| SUGGESTION let string = "Hello ".to_string() + "World" + "!"
+    //~| NOTE an implementation of `std::ops::Add` might be missing for `&'static str`


Would be nice if we could skip the NOTE in this case.

@nrc - I think he's going to move to a UI test, so we won't need them (and we'll be able to see what the actual error will look like)

I meant remove it from the error message, not from the test

Shouldn't be too hard. I have an idea of how to bypass this output.

mgattozzi · 2017-01-18T18:03:29Z

@nrc I've updated the wording, though I'm wondering if it could use some improvement. I didn't want to make it too verbose, but I wanted to get the point across as well. I've also added a way to bypass the NOTE you didn't want output here through some conditional logic short circuiting. It feels a little hackish and I was wondering if you had a better idea here as to what could be improved.

Also not since the test hasn't been updated yet this should fail testing.

bjorn3 · 2017-01-18T18:10:08Z

src/librustc_typeck/check/op.rs

+                            // and it was an addition with &str then skip the
+                            // span note below it. Otherwise just emit the span
+                            // note
+                            if missing_trait == "std::ops::Add" &&


This will also skip the note when the trait is not Add instead of only when Add + &str

You're right. I got the logic wrong here. I tested it just to make sure so this will need to get changed

nrc · 2017-01-23T21:43:07Z

src/librustc_typeck/check/op.rs

+                    err.span_suggestion(expr.span,
+                        &format!("to_owned() can be used to create an owned `String` \
+                                  from a string reference. This allows concatenation \
+                                  since the `String` is owned."), suggestion);


The second sentence doesn't quite explain why that is necessary. How about, "String concatenation appends the string on the right to the string on the left. That modifies the string on the left and may require reallocation, this requires ownership of the string on the left." It's a bit wordy, but tries to explain why you can't use a reference.

mgattozzi · 2017-01-24T04:40:04Z

@nrc and @jonathandturner these commits should cover both the ui test and message output that was brought up earlier.

mgattozzi · 2017-01-24T06:27:02Z

Looking at the raw log one failed on my test due to the output being different but it passed for all the other ones. I'm not sure why this would happen though

nrc · 2017-01-24T20:37:32Z

r = me with the test fixed. I'm not sure why that is failing, but it looks legit to me.

mgattozzi · 2017-01-24T22:16:31Z

@nrc looking at the logs further this seems to be the problem

expected stderr:
error[E0514]: found crate `std` compiled by an incompatible version of rustc
  |
  = help: please recompile that crate using this compiler (rustc 1.16.0-dev (b27d71cb1 2017-01-23))
  = note: crate `std` path #1: /home/michael/Code/rustc/build/x86_64-unknown-linux-gnu/stage2/lib/rustlib/x86_64-unknown-linux-gnu/lib/libstd-86f66de14176b4c2.rlib compiled by "rustc 1.16.0-dev (5c73f42db 2017-01-23)"
  = note: crate `std` path #2: /home/michael/Code/rustc/build/x86_64-unknown-linux-gnu/stage2/lib/rustlib/x86_64-unknown-linux-gnu/lib/libstd-86f66de14176b4c2.so compiled by "rustc 1.16.0-dev (5c73f42db 2017-01-23)"
  = note: crate `std` path #3: /home/michael/Code/rustc/build/x86_64-unknown-linux-gnu/stage2/lib/rustlib/x86_64-unknown-linux-gnu/lib/libstd_unicode-b421c85d3facafd6.rlib compiled by "rustc 1.16.0-dev (5c73f42db 2017-01-23)"
  = note: crate `std` path #4: /home/michael/Code/rustc/build/x86_64-unknown-linux-gnu/stage2/lib/rustlib/x86_64-unknown-linux-gnu/lib/libstd_shim-dac73ac8aa295bd7.rlib compiled by "rustc 1.16.0-dev (5c73f42db 2017-01-23)"

error: aborting due to previous error

I think I added in this file by accident. I think removing it should fix the build issue. If not I'm building the compiler again from scratch and reupdating the references

mgattozzi · 2017-01-25T01:16:59Z

@nrc I fixed the build!

nrc · 2017-01-25T01:49:18Z

@mgattozzi nice! Could you squash the commits please? Then r+

mgattozzi · 2017-01-25T01:49:49Z

@nrc squash and force push?

nrc · 2017-01-25T02:05:30Z

yes

mgattozzi · 2017-01-25T02:17:02Z

@nrc got it. Wasn't sure policy regarding force pushing. Alright it's all rebased and good to go!

nrc · 2017-01-25T03:26:10Z

@bors: r+

Thanks for sticking with it @mgattozzi , nice PR!

bors · 2017-01-25T03:26:11Z

📌 Commit 4da2994 has been approved by nrc

mgattozzi · 2017-01-25T03:45:26Z

@nrc I'm glad to have been able to contribute to the compiler itself finally! It was a joy and I'm hoping to tackle some more issues. Thanks for reviewing it!

bors · 2017-01-25T04:36:17Z

⌛ Testing commit 4da2994 with merge 27a80f7...

bors · 2017-01-25T05:14:48Z

💔 Test failed - status-travis

mgattozzi · 2017-01-25T05:26:25Z

@nrc looks like curl messed up trying to download the rustc binary to do compilation.

GuillaumeGomez · 2017-01-25T16:07:55Z

@bors: retry

bors · 2017-01-28T05:23:54Z

☔ The latest upstream changes (presumably #39353) made this pull request unmergeable. Please resolve the merge conflicts.

kevinmehall · 2017-01-28T18:59:43Z

Looks like this unintentionally added files under src/target/debug/ to git.

mgattozzi · 2017-01-28T22:24:32Z

@kevinmehall thanks I'll remove it

This is the first part of #39018. One of the common things for new users coming from more dynamic languages like JavaScript, Python or Ruby is to use `+` to concatenate strings. However, this doesn't work that way in Rust unless the first type is a `String`. This commit adds a check for this use case and outputs a new error as well as a suggestion to guide the user towards the desired behavior. It also adds a new test case to test the output of the error.

mgattozzi · 2017-01-29T01:23:16Z

@nrc it needs to get approved again so it can go through

nrc · 2017-02-02T03:36:53Z

@bors: r+

bors · 2017-02-02T03:36:54Z

📌 Commit b54f593 has been approved by nrc

bors · 2017-02-02T07:39:10Z

⌛ Testing commit b54f593 with merge 6abe648...

Add clearer error message using `&str + &str` This is the first part of #39018. One of the common things for new users coming from more dynamic languages like JavaScript, Python or Ruby is to use `+` to concatenate strings. However, this doesn't work that way in Rust unless the first type is a `String`. This commit adds a check for this use case and outputs a new error as well as a suggestion to guide the user towards the desired behavior. It also adds a new test case to test the output of the error.

bors · 2017-02-02T10:03:57Z

☀️ Test successful - status-appveyor, status-travis
Approved by: nrc
Pushing 6abe648 to master...

Manishearth · 2017-02-02T10:11:58Z

Finally!

mgattozzi · 2017-02-02T16:55:35Z

@Manishearth and @nrc thanks a lot! This great that it's finally merged!

rust-highfive assigned nrc Jan 17, 2017

mgattozzi mentioned this pull request Jan 17, 2017

Better diagnostics when concatenating strings #39018

Closed

Manishearth approved these changes Jan 17, 2017

View reviewed changes

bjorn3 requested changes Jan 17, 2017

View reviewed changes

Manishearth mentioned this pull request Jan 17, 2017

Add "diagnostic items" which behave like lang items #39131

Closed

nrc suggested changes Jan 18, 2017

View reviewed changes

bjorn3 requested changes Jan 18, 2017

View reviewed changes

nrc reviewed Jan 23, 2017

View reviewed changes

nrc approved these changes Jan 24, 2017

View reviewed changes

bors merged commit b54f593 into rust-lang:master Feb 2, 2017

mgattozzi deleted the better-string-message branch February 2, 2017 16:55

eddyb mentioned this pull request Mar 27, 2017

ICE when trying to Add a Vec and vec![] on 1.17 nightly #40610

Closed

bluthej mentioned this pull request Apr 28, 2023

Make String a diagnostic item #110949

Closed

Add clearer error message using &str + &str #39116

Add clearer error message using &str + &str #39116

Conversation

mgattozzi commented Jan 17, 2017

rust-highfive commented Jan 17, 2017

Manishearth commented Jan 17, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mgattozzi commented Jan 17, 2017

Manishearth commented Jan 17, 2017

Manishearth commented Jan 17, 2017

mgattozzi commented Jan 17, 2017

Manishearth commented Jan 17, 2017

mgattozzi commented Jan 17, 2017

Manishearth commented Jan 17, 2017

Manishearth commented Jan 17, 2017

mgattozzi commented Jan 17, 2017

Manishearth commented Jan 17, 2017

Manishearth commented Jan 17, 2017

sophiajt commented Jan 17, 2017

mgattozzi commented Jan 17, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mgattozzi commented Jan 18, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mgattozzi commented Jan 24, 2017

mgattozzi commented Jan 24, 2017

nrc commented Jan 24, 2017

mgattozzi commented Jan 24, 2017 • edited Loading

mgattozzi commented Jan 25, 2017

nrc commented Jan 25, 2017

mgattozzi commented Jan 25, 2017

nrc commented Jan 25, 2017

mgattozzi commented Jan 25, 2017

nrc commented Jan 25, 2017

bors commented Jan 25, 2017

mgattozzi commented Jan 25, 2017

bors commented Jan 25, 2017

bors commented Jan 25, 2017

mgattozzi commented Jan 25, 2017

GuillaumeGomez commented Jan 25, 2017

bors commented Jan 28, 2017

kevinmehall commented Jan 28, 2017

mgattozzi commented Jan 28, 2017

mgattozzi commented Jan 29, 2017

nrc commented Feb 2, 2017

bors commented Feb 2, 2017

bors commented Feb 2, 2017

bors commented Feb 2, 2017

Manishearth commented Feb 2, 2017 • edited Loading

mgattozzi commented Feb 2, 2017

Add clearer error message using `&str + &str` #39116

Add clearer error message using `&str + &str` #39116

mgattozzi commented Jan 24, 2017 •

edited

Loading

Manishearth commented Feb 2, 2017 •

edited

Loading