In this test, it's impossible that start < 0 so the test is useless.
I'm not sure what the cursor test was for, but it's very old code, and
it seems the assumption was either misled or doesn't hold any more:
testing for the absolute cursor position against the length of the
word against the cursor makes no sense.

The net result of this was that when the cursor index got large
enough, resuming suggestion would not work any more.

Bug: 7586467
Change-Id: I3462082374fe9579bec7698f4d424de6ff5f2ded

multi project commit with I5847d4aba97a7224ae170a7153811c5a11816449

Change-Id: Ie6fbf6f19f0193451fe6f3ddf9d4ef1637a35db1

multi-space logging should look like single-space logging, missing a few minor log statements
(SuggestionUpdates, SetComposingText)

multi-project commit with I2af842348c2f2b8f7271ac5b63def245e83df24d

Change-Id: Icd3187c0d0377255f82787afffea657c14345803

Change-Id: I7defaf1f577fd67e678cac83ff935e8181dd0a48

When the user edits a word before adding it to the user
dictionary, the keyboard should replace whatever was
committed before with the amended version.

Bug: 7725834
Change-Id: I1a417be6c5a86d6a96bc2c76aca314ad8f1202a9

Change-Id: I0c29c5d2abcbf80759b996d34b534deb083cd7d3

Not sure when this happens exactly, but it is possible that
InputConnection#getTextBeforeCursor returns null. This
happens for example upon rotating the screen with the
composing field empty in Gmail.
In this case, StringBuilder#append will convert the null
pointer into the string "null", which is sure better than a
crash, but can have a number of bad side-effects, like
auto-caps not working.

Bug: 7533034
Change-Id: Ia1cfab432c13a12ff1c2f013c59bac05a587f553

Change-Id: Iee01d4d2b916d0b584531104ac865ae6e6370a3d

Change-Id: Idc478f901185ee1b4912acc82d0cbc54fee4e991

Bug: 7359291
Bug: 7357758
Bug: 7197651
Change-Id: Ia805a87e922739ae0a06978a3bf00d91c94b6c51

Bug: 7226098
Change-Id: Ibacb740022a6af56c9d555d6797dd402a89f6ca4

Change-Id: I6009b3c1950ba32b7f1e205a3db2307fe0cd688e

Bug: 6950087
Change-Id: I198699f6e687ac74c35b9edfe469dce880d9ecaf

Bug: 6950087
Change-Id: I8a1bca24db64b4dd54db8ac74d90cf43cbdddab6

Bug: 4967874
Change-Id: Ic7ce7b2de088308fa00865c81246c84c605db1e5

Since the function has to be modified heavily but does a lot
of non-trivial work, add a wealth of comments explaining what
it does and why so as to facilitate understanding the changes
to come.

Bug: 4967874
Change-Id: I6c21aea15f161d807035f279dfb7d1b98b3e9144

This should have on effect at all on behavior,
except an increase in performance.

Bug: 4967874
Bug: 6950087
Change-Id: Ie2b51efefe84ca767f5dc8e3b80bfef7e1faab3d

This is pretty much as strong as it gets. It should be
impossible to get false positives and nearly impossible to
get true negatives with this new code.

Bug: 6981089
Change-Id: Ia32ab62f89c5943f0be169b979abab652e67bf5b

This is a preparatory change for
Bug: 4967874
Bug: 6617760
Bug: 6950087

Change-Id: I3abf8e45c0d02c42491421f108370220134b9602

Move many ResearchLogger data collection points to RichInputConnection.
By collecting data here, developers do not have to remember to keep the
ResearchLog consistent with calls to the RichInputConnection.

In addition, some unnecessary log points were removed, and the ResearchLogger
is now independent of LatinImeLogger.

multi-project change with I05496cfd762e9a41c42c50099143f1efc0a3d8a2

Bug: 6188932
Change-Id: I424abb2e437901262a9620255493916b7c3ff74b

This reverts commit d0c41e0c

Change-Id: I5fd8a67a9fdadb7e6532cafecd00c8f33a00ec72

Move many ResearchLogger data collection points to RichInputConnection.
By collecting data here, developers do not have to remember to keep the
ResearchLog consistent with calls to the RichInputConnection.

In addition, some unnecessary log points were removed, and the ResearchLogger
is now independent of LatinImeLogger.

multi-project change with Ifbabefc008d32d99f89e8289e504188e7232ebd4

Bug: 6188932
Change-Id: I4a42d963ba952e703d425acd3a9871d5fca1f03a

Change-Id: Id79c70310dd774dd5e7ccda6d4b8c36b88c67744

This reverts commit 1964ffb3

Change-Id: Ie43e7ac828faa91e1824a255bb3ecf6c8222d4e8

Move many ResearchLogger data collection points to RichInputConnection.
By collecting data here, developers do not have to remember to keep the
ResearchLog consistent with calls to the RichInputConnection.

In addition, some unnecessary log points were removed, and the ResearchLogger
is now independent of LatinImeLogger.

multi-project change with Ib71f841044ec1072610ab5638a5edfce29b7c05b

DO NOT MERGE

Bug: 6188932
Change-Id: I9ec7500a5b18bfe1a5a5bb1e5bf21c43351fc59e

Move many ResearchLogger data collection points to RichInputConnection.
By collecting data here, developers do not have to remember to keep the
ResearchLog consistent with calls to the RichInputConnection.

In addition, some unnecessary log points were removed, and the ResearchLogger
is now independent of LatinImeLogger.

multi-project change with I67bbf5fe8fd821da967a50b69e467caafc7fd388

Bug: 6188932
Change-Id: I49ccee1dcffe8d41401b42a637dd0da8dd11d90d

- new package: com.android.inputmethod.research

multi-project commit with Ic0a5744f3160d13218addd589890623c0d120ffc

Bug: 6188932

Change-Id: Icf8d4a40a5725401799be6e209a640d99a5f34c4

Change-Id: I6e7c3aba2b40ecefa483c9680978bc806368e96e

Bug: 6766059
Change-Id: I378f9d35c4904c4f373260bda5863235d833eb31

This has a good, although small, impact on performance : it removes
a two-way IPC call in a most frequent case, while possibly adding
one in a rather unfrequent and less critical case.

Also, this fixes a bug with surrogate pairs. This specific branch
of code now correctly handles surrogate pairs.
Aside from this, it should have no impact on behavior.

However, since it does delay access to the previous character in
the text view by a two-way IPC call, it actually goes a long way
toward fixing bug#6668226. It is not really a fix and the race
condition still exists, but this change makes it much, much
harder to hit.

Bug: 6668226
Change-Id: Id11cc6a0b7488d6bd392227cafdcf3a8d4c62f6c

This is harmless, but against policy.
Also, rework the checking code to be more readable, give more
information, and be called for all relevant methods - and not
for informative methods, which are not required to be in a
batch edit.

Change-Id: I03fa8b2e7d68a6a133f86be8a214671750c29256

Change-Id: I512b04e23490413a44b1ca0517102fe2d9138df3

Change-Id: Ie143ad008ca8deb1c5034cb80ec1cb28a6b0682d

Change-Id: If8d03bea1b6369cb3043da6321048d052995d664

Change-Id: I8527776a4bd8236a85bb77f9791f9b76977ad181

Change-Id: I9982ff325bef56694402caef28a77683c52ccd71

The goal is to simplify the code in LatinIME.java as well as having
a handy place to put debug calls to see interaction with TextView.

Change-Id: I255227e7e7343e0c2f3dcd1f185e5020d6186732