Add Latin-1 converters to convert library.

Add Latin-1 converters to convert library.

I plan to modify this to a generic version and make Ascii and Latin1
special cases that only differ in the mask (0x7f vs 0xff), or maybe
a few other cases.

R=floitsch@google.com

Committed: https://code.google.com/p/dart/source/detail?r=26508

[floitsch, 2013-08-21 14:07:33]

LGTM.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart
File sdk/lib/convert/latin1.dart (right):

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:31: * returned by [decoder] will default ti allowing
invalid values. Invalid
to

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:44: * will eventually thrown a [FormatException].
throw

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:50: if ((allowInvalid == null) ? _allowInvalid :
allowInvalid) {
split into two lines:
if (allowInvalid == null) allowInvalid = _allowInvalid;
if (allowInvalid) ...

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:81: // TODO: Use Uint8List when possible.
There should be a bug-number.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:82: return new List<int>(string.length)..setAll(0,
string.codeUnits);
You think it's faster this way?
I would allocate the List first, and then fill it while verifying that they are
0xFF.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:111: // TODO(lrn): Use Uint8List when available.
There is probably a bug-number.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:217: class _Latin1DecoderSink implements
ByteConversionSink {
extends ByteConversionSinkBase
We want to be able to add "asXXSink" and not need to change _Latin1DecoderSink

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:233: _sink.add(new
String.fromCharCodes(source.getRange(start, end)));
I would avoid the getRange if start and end are 0 and length.
If the source is a Uint8List we don't want to hide that from
String.fromCharCodes.
(Although this is mostly temporarily anyways since the StringConversionSink
should deal with it.)

https://codereview.chromium.org/23224014/diff/1/tests/lib/convert/latin1_test...
File tests/lib/convert/latin1_test.dart (right):

https://codereview.chromium.org/23224014/diff/1/tests/lib/convert/latin1_test...
tests/lib/convert/latin1_test.dart:8: var latin1Strings = [
It would be good to create additional tests from this.
longer (duplications), for example.
See unicode_tests.dart for some examples.

[Lasse Reichstein Nielsen, 2013-08-22 08:28:37]

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart
File sdk/lib/convert/latin1.dart (right):

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:31: * returned by [decoder] will default ti allowing
invalid values. Invalid
On 2013/08/21 14:07:33, floitsch wrote:
> to

Done.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:44: * will eventually thrown a [FormatException].
On 2013/08/21 14:07:33, floitsch wrote:
> throw

Done.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:50: if ((allowInvalid == null) ? _allowInvalid :
allowInvalid) {
On 2013/08/21 14:07:33, floitsch wrote:
> split into two lines:
> if (allowInvalid == null) allowInvalid = _allowInvalid;
> if (allowInvalid) ...

Done.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:81: // TODO: Use Uint8List when possible.
On 2013/08/21 14:07:33, floitsch wrote:
> There should be a bug-number.

Done.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:82: return new List<int>(string.length)..setAll(0,
string.codeUnits);
Good point, optimize for the valid case.
Done.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:111: // TODO(lrn): Use Uint8List when available.
On 2013/08/21 14:07:33, floitsch wrote:
> There is probably a bug-number.

Done.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:217: class _Latin1DecoderSink implements
ByteConversionSink {
On 2013/08/21 14:07:33, floitsch wrote:
> extends ByteConversionSinkBase
> We want to be able to add "asXXSink" and not need to change _Latin1DecoderSink

Done.

https://codereview.chromium.org/23224014/diff/1/sdk/lib/convert/latin1.dart#n...
sdk/lib/convert/latin1.dart:233: _sink.add(new
String.fromCharCodes(source.getRange(start, end)));
Yes, this was kept deliberately simple because it's temporary. Making it too
fancy might make someone think it's actually reasonable code.

I don't think String.fromCharCodes recognizes UInt8List anyway.

No change, kept deliberately sucky.

https://codereview.chromium.org/23224014/diff/1/tests/lib/convert/latin1_test...
File tests/lib/convert/latin1_test.dart (right):

https://codereview.chromium.org/23224014/diff/1/tests/lib/convert/latin1_test...
tests/lib/convert/latin1_test.dart:8: var latin1Strings = [
Wouldn't it be great if String had an operator*, so I could just write 
  "blåbærgrød" * 100
?!?

Added longer concatenations of the example strings as well.

[Lasse Reichstein Nielsen, 2013-08-22 08:28:47]

Committed patchset #1 manually as r26508 (presubmit successful).