| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This is a breaking change.
* fromJS now returns `Opt[void]' (effectively a `bool'), and only sets
the result if the conversion was successful. An unsuccessful
conversion may leave the result in an unspecified state.
This makes fromJS much lighter, both in terms of generated code as
well as unnecessary overhead introduced by copying. It's also not much
less ergonomic than the previous signature, since the ? syntax still
works.
* Improved JS_UNDEFINED/global object handling: instead of special-
casing the global object, we special-case the `this' value conversion
so that it fixes up `undefined' automatically.
* fromJS on JSDict now throws on missing properties. Default values can be set
through the new .jsdefault pragma.
|
| |
| |
| |
| |
| | |
* remove bounds check if actualMinArgs == 0
* skip name collision checks (the compiler will do it for us anyway)
|
| | |
|
| |
| |
| |
| | |
Treat it as a 0-copy slice of the QJS argp.
|
| | |
|
| | |
|
| |
| |
| |
| | |
It was importing the wrong header
|
| | |
|
| |
| |
| |
| | |
This was causing memory corruption.
|
| |
| |
| |
| | |
Mainly for documentation purposes.
|
| | |
|
| | |
|
| |
| |
| |
| | |
this should be all of them
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| | |
* fix headers
* move libunicode & cutils compilation commands to libunicode wrapper
* replace lre_is_space with lre_is_space_non_ascii (the inlined variant
doesn't work if the module importing the wrapper is in a different
directory)
|
| | |
|
| |
|
|\
| |
| |
| |
| |
| | |
git-subtree-dir: lib/chagashi0
git-subtree-mainline: ddd639280d93da1ae51af18fe22fdd315fc2d42d
git-subtree-split: f5cf26b044a82a2e3cffe83bb2342b7b029f2ab8
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| | |
Makes it possible to pass in string slices without copying.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| | |
Chagashi never throws; this is just to make consumers' jobs easier
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* remove unused JIS X 0212 encoding table
* skip JIS X 0208 empty rows
* merge Shift_JIS encoding table with JIS X 0208 table
* skip EUC-KR empty rows & columns
* RLE EUC-KR & GBK top + bottom left sections
* skip GB18030 PUA section
* use RLE EUC-KR and GB18030 decoder ranges for encoding too
* fix gitignore
shaves off ~154K from the binary
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| | |
* simplify decode iterator: only yield once, etc.
* fix decode iterator done not resetting `n'
* fix partial UTF-8 at the end of a string eating the internal buffer
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The validator/decoder split was an abomination, and having to buffer
your input queue for UTF-8 inputs was kind of annoying too.
Now we have an extra instruction that also instructs the reader to use
data from its input queue. Currently this is only used for UTF-8, but we
may extend it to other ASCII-compatible charsets too in the future.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| | |
+ remove an artifact
|
| |
| |
| |
| | |
just a sanity check for myself, it's a bug in Chawan
|
| | |
|
| |
| |
| |
| | |
seems like it doesn't like `out'
|