Fix TTS extension API language matching.

Fix TTS extension API language matching.

Now if an extension registers a voice with lang 'fr-FR', a speech
request with lang 'fr' will match, and vice versa. Strict matches
still take precedence.

BUG=133470
TEST=adds new browser test.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=143320

[David Tseng, 2012-06-20 17:18:29]

http://codereview.chromium.org/10581011/diff/1/chrome/browser/speech/extensio...
File chrome/browser/speech/extension_api/tts_engine_extension_api.cc (right):

http://codereview.chromium.org/10581011/diff/1/chrome/browser/speech/extensio...
chrome/browser/speech/extension_api/tts_engine_extension_api.cc:133: for (size_t
i = 0; i < tts_voices.size(); ++i) {
Use a const iterator here.

http://codereview.chromium.org/10581011/diff/1/chrome/browser/speech/extensio...
chrome/browser/speech/extension_api/tts_engine_extension_api.cc:140: if
(!voice.lang.empty() && !utterance->lang().empty()) {
Can you only make two passes to resolve the lang code (and not iterate through
all voices again)?

[dmazzoni, 2012-06-20 22:51:12]

http://codereview.chromium.org/10581011/diff/1/chrome/browser/speech/extensio...
File chrome/browser/speech/extension_api/tts_engine_extension_api.cc (right):

http://codereview.chromium.org/10581011/diff/1/chrome/browser/speech/extensio...
chrome/browser/speech/extension_api/tts_engine_extension_api.cc:133: for (size_t
i = 0; i < tts_voices.size(); ++i) {
On 2012/06/20 17:18:30, David Tseng wrote:
> Use a const iterator here.

We're using *i* below - we need to return the voice index. Switching to an
iterator would mean we'd need to keep a separate index variable.

http://codereview.chromium.org/10581011/diff/1/chrome/browser/speech/extensio...
chrome/browser/speech/extension_api/tts_engine_extension_api.cc:140: if
(!voice.lang.empty() && !utterance->lang().empty()) {
On 2012/06/20 17:18:30, David Tseng wrote:
> Can you only make two passes to resolve the lang code (and not iterate through
> all voices again)?

I think it's necessary to iterate over all extensions twice, because I want to
make sure that an exact match is always preferred over a fuzzy match.

It would be possible to iterate over all voices just once and keep track of just
possible matches, then do two quick passes over that shorter list to pick the
one with the best language match, but I don't think that would improve
readability, and the number of voices is small so performance shouldn't matter
here.

Does that make sense? Or do you see another approach?

[David Tseng, 2012-06-20 23:38:40]

lgtm

http://codereview.chromium.org/10581011/diff/1/chrome/browser/speech/extensio...
File chrome/browser/speech/extension_api/tts_engine_extension_api.cc (right):

http://codereview.chromium.org/10581011/diff/1/chrome/browser/speech/extensio...
chrome/browser/speech/extension_api/tts_engine_extension_api.cc:133: for (size_t
i = 0; i < tts_voices.size(); ++i) {
On 2012/06/20 22:51:12, Dominic Mazzoni wrote:

std::distance

But, I can go either way.

http://codereview.chromium.org/10581011/diff/1/chrome/browser/speech/extensio...
chrome/browser/speech/extension_api/tts_engine_extension_api.cc:140: if
(!voice.lang.empty() && !utterance->lang().empty()) {
On 2012/06/20 22:51:12, Dominic Mazzoni wrote:

We perform strict matching via voice_name right? Are there extensions that have
no voice_name? I was thinking only of the case when we match against only one
voice name (and so we would only need to strict/fuzzy match against its lang).

But, if some extensions have no voice name, then I see why you need to do this.

[dmazzoni, 2012-06-20 23:42:53]

On 2012/06/20 23:38:40, David Tseng wrote:
> We perform strict matching via voice_name right? Are there extensions that
have
> no voice_name? I was thinking only of the case when we match against only one
> voice name (and so we would only need to strict/fuzzy match against its lang).

Every engine will have a voice name, but the client is allowed to not specify
the voice name. So a client can specify just the language and nothing else - in
which case we should make two complete passes, once to see if there's an exact
match for the language, and again to see if there's a fuzzy match.

[commit-bot: I haz the power, 2012-06-20 23:44:00]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dmazzoni@chromium.org/10581011/1

[commit-bot: I haz the power, 2012-06-21 01:00:07]

Change committed as 143320