Fix for #9243, nre returns "" instead of nil for missing matches #9755
Conversation
|
\cc @Araq @timotheecour |
|
I think this API breakage is fine and unavoidable given that |
|
/cc @flaviut what's the status on this? |
|
@timotheecour it's in good shape, I'd just like to get one other person to review it before I push the button. |
| @@ -31,6 +31,25 @@ | |||
|
|
|||
| - `options.UnpackError` is no longer a ref type and inherits from `System.Defect` instead of `System.ValueError`. | |||
|
|
|||
| - nre's `RegexMatch.{captureBounds,captures}[]` no longer return `Option` or | |||
| `nil`/`""`, respectivly. Use the newly added `n in p.captures` method to | |||
| check if a group is captured, otherwise you'll recieve an exception. | |||
There was a problem hiding this comment.
haven't thought about it much but is it possible to get a CT (instead of RT) error?
There was a problem hiding this comment.
Not unless the halting problem is solved. It's unknown whether a certain variable corresponds to a certain pattern. With constants it might be possible, but only when a group isn't optional (i.e. (foo)? will never fail at CT), and only if nim starts supporting compile-time FFI.
There was a problem hiding this comment.
/cc @flaviut
that's not what I meant, but I worded it poorly; see https://github.com/nim-lang/Nim/issues/9941
I'd like community feedback on this PR because it contains some significant breaking nre changes.