Uri.decodeFull Invalid argument(s): Illegal percent encoding in URI

[yarnping]

hi team, I'm trying to decode url, seems not working

Uri.decodeFull("text/33%20-%20â.html")
Uri.decodeComponent("text/33%20-%20â.html")
expect
text/33 - â.html

actually throw error:
Invalid argument(s): Illegal percent encoding in URI

refer to
https://www.urldecoder.org/

[dart-github-bot]

Summary: The user is encountering an error when attempting to decode a URL using Uri.decodeFull(). The provided URL contains an invalid percent encoding, causing the function to throw an error.

[lrhn]

The decode functions only accept valid URI text, the output of encoding, which means it must be ASCII only.
The error message is confusing though. It should probably say "Invalid unencoded character".

(One of the reason for not just silently accepting non-ASCII characters is that the function also decodes UTF-8 encoded byte sequences, so fx, "%C3%98" decodes to "Ø" (U+00D8).
It's not clear whether an input of "\xC3\x98" should be decoded as UTF-8 or not. Rather than guess in either direction, it's just an error to have any non-ASCII in the input to the decode functions.

[yarnping]

got it, thanks for explaination

but how can I decode str like this
text/33%20-%20â.html

I use java before, seems that it's has a difference choice
https://github.com/openjdk/jdk/blob/688e92e7f5febddd2935cb7f500dd3f10fbd9401/src/java.base/share/classes/java/net/URLDecoder.java#L170

[lrhn]

Maybe Dart should just do what Java does: Only worry about decoding %xx sequences, and leave everything else unchanged.
There is absolutely no difference between decodeComponent and decodeFull, they do the exact same thing.

We could change one of them to only decode percentages. Maybe even allow passing in an Encoding that defaults to utf8. I'd use decodeComponent for that. Or change both, and the rest of the decode... functions, because then it's just one function.

(The one thing Dart does is to also decode + to space in some cases, but that's only decodeQueryComponent, which also allows providing a decoding.)

No promises, but I think it'd be a reasonable and non-breaking change.

[yarnping]

ok thanks for your time

I'll make a temporary solution for this, leave everything else unchanged.

[lrhn]

You can perhaps use something like:

final _escapeRE = RegExp(r'(?:%[\da-fA-F]{2})+');
String unescapeEscapes(String input) =>
    input.replaceAllMapped(_escapeRE, (m) => Uri.decodeComponent(m[0]!));

That replaces only sequences of %HH escapes with the UTF-8 decoding of their values.
(And is what the function decodeComponent would do if I make the change suggested above.)