Allow the history URL provider to handle input of type QUERY. This helps in ...

Allow the history URL provider to handle input of type QUERY.  This helps in the case where the user types something that on its own isn't navigable, but might be the prefix of something else navigable.

While there are no tests for this directly, another change of mine to treat more inputs with explicit schemes as queries (e.g. "http:/") relies on this, and does unittest for it.

In order to fit the new relevance scores into the table, I went through and simplified the relevance scoring so that generally providers used the same scores for more input types.  The effects of this should be barely noticeable (it affects the ranking of past search queries that are old against results from the secondary search provider), and it simplifies the code noticeably.

This also fixes a "bug" that the NavSuggest results were incremented backwards, but since we only score one of these right now there's no visible effect.

BUG=none
TEST=none

[Paweł Hajdan Jr., 2009-10-17 06:42:24]

Drive-by with some test-style related comments.

http://codereview.chromium.org/291005/diff/1/6
File chrome/browser/autocomplete/history_url_provider_unittest.cc (right):

http://codereview.chromium.org/291005/diff/1/6#newcode215
Line 215: if (!matches_.empty())  // If this fails, RunTest() already printed
it.
Could you rather make it ASSERT_FALSE(matches_.empty()) or
ASSERT_NO_FATAL_FAILURE(RunTest(L"news", std::wstring(), ... ? This way we can
always be sure about the code path taken in test when it passes.

http://codereview.chromium.org/291005/diff/1/6#newcode350
Line 350: if (!matches_.empty())  // If this fails, RunTest() already printed
it.
Same here.

http://codereview.chromium.org/291005/diff/1/6#newcode362
Line 362: if (!matches_.empty())  // If this fails, RunTest() already printed
it.
And here.

[sky, 2009-10-19 15:51:50]

LGTM

[Peter Kasting, 2009-10-19 18:42:26]

On 2009/10/17 06:42:24, Paweł Hajdan Jr. wrote:
> http://codereview.chromium.org/291005/diff/1/6#newcode215
> Line 215: if (!matches_.empty())  // If this fails, RunTest() already printed
> it.
> Could you rather make it ASSERT_FALSE(matches_.empty()) or
> ASSERT_NO_FATAL_FAILURE(RunTest(L"news", std::wstring(), ... ? This way we can
> always be sure about the code path taken in test when it passes.

Sure, done.  Landing.