When normalizing, we need to stop processing on an NFC boundary, not an
NFD boundary, to support normalizations such as in Bengali, where
appending `U+09D7` to a context of `U+0995 U+09C7` should result in
`U+0995 U+09CC`.
The specification is unclear on this; see https://unicode-org.atlassian.net/browse/CLDR-19218
This also updates the ldml keyboard unit test suite to support running
in full NFC mode (used in all Engine implementations) as well retaining
the NFD mode (now only used by the debugger).
Side note: the Bengali normalization failure case was picked up by the
improvements to the unit test suite, proving once again that good tests
are so valuable.
Fixes: #15491Fixes: #15505
Follows: #15488
Relates-to: CLDR-19218
Ensure that when a single backspace decomposes the last NFC character in
the app context, the remainder of the 'cluster' is preserved, matching
the implication of the CLDR keyboard specification.
This addresses the behavior in #15487 where the cached context became
out of sync with the app context after deleting an entire NFC cluster
such as ê, which caused a loop ending up with the entire context being
deleted, at which point the loop exited with a fail-safe.
Note that the LDML keyboard tests (ldml.cpp) do not currently exercise
the normalization code; this is a gap that should be addressed to ensure
that we are testing final application behavior.
Fixes: #15487
- add a normalize_nfd() which takes a single codepoint
- temporarily keep ICU in actions_normalize.cpp and ldml_transforms.cpp
- expand wasm opts in unit tests
- always set to 0 for now (keep ICU around)
- set KMN_IN_LDML_TESTS in tests to keep ICU there for test and comparison
- add core_icu.cpp and put some utils there.
#9467
- output[0] was used as a boolean, but needed to check output.isEmpty() instead
- dead store to actions.code_points_to_delete
For: #11067 parts 1 and 2
Fixes#10582.
Returned struct from the `km_core_state_get_actions()` API is now owned
by the state object. This is a change in API contract. The corresponding
`km_core_actions_dispose()` API has been made private, because there is
never any need for API consumers to call it. As this change is happening
in alpha, we won't bump the ABI version.
`state->app_context` is now guaranteed to be in sync with
`state->context` after `km_core_process_event()`, with the actions
struct populated at that point.
The next and final step in this refactor is to remove the actions queue
altogether from the Core and make that a kmn-only concept, but that will
wait until 18.0. The only consumers of the actions queue are the
kmn-specific interactive debugger in Keyman Developer, and kmn-specific
Input Method eXtensions in Keyman Engine for Windows.
Fixes#10431.
The `km_core_` prefix has been removed from internal-only functions, and
these function declarations moved to context.hpp.
The functions have not been moved from km_core_context_api.cpp at this
stage.
Rewrote the function documentation in Javadoc style comments for the
internal use functions.
Per discussion in #10422, we can assume that input cached_context is
always NFD. However input actions->output may not start at a
normalization boundary, so we still need to backtrack to a normalization
boundary in order to get our NFC output. But cached_context never need
change.
This makes no change to the algorithm, but tweaks some of the unit tests
to adhere to this input assumption.
Note: we could consider adding a debug assertion that cached_context is
NFD.
Fixes#9999.
Note TODO items:
- [ ] Renormalize cached_context across action boundary. Blocked by #10369.
- [ ] Add extra tests for surrogate pairs
- [ ] Move set_context_from_string into helper module
- [ ] if we don't apply normalization, we still need to fixup the
app_context, to keep it coherent with cached_context (or at least
we need to verify that app_context is never used in this
situation)
