This is because prediction can't be easily distinguished in
style. This fixes the bug by simulating the right members,
but some refactoring should be done to remove useless
booleans.
Bug: 17271923
Change-Id: Ib88f3fb95678021624e59535492926dd315d26fb
This is a groundwork for subsequent CLs whith need to
do certain things for and only for the typed word.
Change-Id: I374b79cbc79470f57f523c996010301ea5701a61
They are pretty extensive already and require an implementation
that works pretty well. However ours is based on an old architecture
and updating them will take time, so for the time being it's best
to just deactivate the tests.
Change-Id: I649a220d13fc455b4306706bb964602c27a5a28c
Formerly "Show correction suggestions" had three options, "always
show", "show in portrait mode", and "always hide". The reason behind
"show in portrait mode" was that there may not be enough screen estate
in landscape mode to show suggestions. Because recent phone devices
have relatively large screen, we decide to remove "show in portrait
mode" option.
Bug: 15780939
Change-Id: I896d737452c3893d43ce20bd88127f10c1eb3d83
This is a groundwork for enabling LocaleSpan.
To enable LocaleSpan everywhere, we need to switch to
CharSequence from String so that Span infromation can be
preserved end-to-end. To achieve this, we need to have
CharSequence version of String#split.
BUG: 16029304
Change-Id: I0dd103185dcf62fb1e25054a374340790e6a4678
This CL removes following methods from StringUtils, which are
no longer used since Id3c262386a7bc7ed75966b1395a50171abe550d3.
- #joinCommaSplittableText
- #appendToCommaSplittableTextIfNotExists
BUG: 16000850
Change-Id: Ibdc9b7e463a19cdfac788b24c0a6322d0c5850c5
This CL consolidates the initialization logic for additional
subtypes so that each additional subtypes can have predictable
subtype ID regardless of OS version.
Previously subtype IDs for additional subtypes are calculated
differently depending on the running OS version with hoping it
minimizes the risk of compatibility issues. However, it is
getting harder and harder to maintain slightly different
logic between OSes. Thus we decided to unify the logic into
that in KitKat even though it may causes some breaking changes.
Note that the actual extra values that are used to instantiate
InputMethodSubtype object are still determined on the fly
depending on the running OS version. However these actual
extra values are no longer used for the subtype ID calculation.
BUG: 16000850
Change-Id: Id3c262386a7bc7ed75966b1395a50171abe550d3
Special case <valid word>.<valid word> to send as a suggestion
the same string where the periods is replaced by a space.
Bug: 10780091
Change-Id: I43c94675977f9ab5d7ee5671486cb742b39f3974
This member has outlived its usefulness. It's not clear now that it
makes things really faster, but it does bring a lot of complexity
that we can avoid by removing it.
Change-Id: Ifbc8094a45b56b958fe165b1930f4cc358a97721
Rename `initialize' to `start' and `deactivate' to `stop' for ease
of understanding. It's completely bizarre that you can "deactivate"
a recapitalization and that will just restart a new one the next time
you press shift. Start and stop are easier to understand.
Change-Id: I455175aebfcfe11198c6c61bf2c6b19efc3f82df
DEBUG in LatinIME has been changed by Ie52007dc to mean both
DebugSettings.PREF_DEBUG_MODE and Settings.PREF_KEY_IS_INTERNAL
instead of just DEBUG_MODE as used to be the case. Tests depend
on the debug mode to test many things and they make sure they run
in debug mode or they fail. This has tests understand the change in
policy.
Change-Id: Ie9f29e6a1d85e19704c8940d3fcf97c13c3dbaa6
This reverts commit d941ea18 and builds on it to fix the behavior
of the broken cases.
It also fixes a small, related bug that probably has existed for
a very long time: predictions not displayed when cancelling
double-space-to-period.
Bug: 15148015
Change-Id: I1f9358f8b6f5804f831643611576be347e83999d
