Helpers for null-terminated Utf8 #3
Conversation
|
LGTM overall, Lasse will do a style review |
lib/src/utf8.dart
Outdated
| } | ||
|
|
||
| static Pointer<Utf8> toUtf8(String s) { | ||
| final List<int> units = Utf8Encoder().convert(s); |
There was a problem hiding this comment.
I believe Utf8Encoder.convert this is now typed to return Uint8List. You may want to use that as type.
You can use const Utf8Encoder() to avoid allocating a new object each time (if you care).
That's the same object returned by utf8.encoder as well, and you could also just write utf8.encode(s) for exactly the same result.
Easier to read.
There was a problem hiding this comment.
I don't see the new type:
abstract class Encoding extends Codec<String, List<int>> {
|
Working on better tests. |
|
PTAL |
lib/src/utf8.dart
Outdated
| int i = 0; | ||
| while (nativeString[i] != 0) { | ||
| if (++i == count) { | ||
| count *= 2; |
There was a problem hiding this comment.
Could we use a precise bound here?
There was a problem hiding this comment.
No, because the length of the string can be only be detected by scanning until the first NULL byte.
There was a problem hiding this comment.
But, why times 2 every time? Why not scan until null and use that as count?
There was a problem hiding this comment.
This is scanning until null.
There was a problem hiding this comment.
Ah, I was confused by array.asExternalTypedData(count: count) where that count is potentially 2x the string length. We need an oversized TypedData wrapper in order to not create a new wrapper every character for scanning in the first place. (I hope we can switch to the normal Pointer<Int8>.load() soon, that would make it easier to read this code.)
No description provided.